The first implementation #7456 did not include `index_part.json` changes
in an attempt to keep amount of changes down. Tracks the historic
reparentings and earlier detach in `index_part.json`.
- `index_part.json` receives a new field `lineage: Lineage`
- `Lineage` is queried through RemoteTimelineClient during basebackup,
creating `PREV LSN: none` for the invalid prev record lsn just as it
would had been created for a newly created timeline
- as `struct IndexPart` grew, it is now boxed in places
Cc: #6994
Split checkpoint_stats into two separate metrics: checkpoints_req and
checkpoints_timed
Fixes commit
21e1a496a3
---------
Co-authored-by: Peter Bendel <peterbendel@neon.tech>
The old test based on the immutable `target_file_size` that was a
parameter to the function.
It makes no sense to go further once `current_level_target_height` has
reached `u64::MAX`, as lsn's are u64 typed. In practice, we should only
run into this if there is a bug, as the practical lsn range usually ends
much earlier.
Testing on `target_file_size` makes less sense, it basically implements
an invocation mode that turns off the looping and only runs one
iteration of it.
@hlinnaka agrees that `current_level_target_height` is better here.
Part of #7554
While switching to use nextest with the repository in f28bdb6, we had
not noticed that it doesn't yet support running doctests. Run the doc
tests before other tests.
in addition to layer names, expand the input vocabulary to recognize
lines in the form of:
${kind}:${lsn}
where:
- kind in `gc_cutoff` or `branch`
- lsn is accepted in Lsn display format (x/y) or hex (as used in layer
names)
gc_cutoff and branch have different colors.
We had accidentally left two endpoints for `tenant`: `/synthetic_size`
and `/size`. Size had the more extensive description but has returned
404 since renaming. Remove the `/size` in favor of the working one and
describe the `text/html` output.
We didn't check permission in `"/v1/failpoints"` endpoint, it means that
everyone with per-tenant token could modify the failpoints. This commit
fixes that.
## Problem
Move from aws based arm64 runners to bare-metal based
## Summary of changes
Changes in GitHub action workflows where `runs-on: arm64` used. More
parallelism added, build time for `neon with extra platform builds`
workflow reduced from 45m to 25m
## Problem
This caused a variation of the stats bug fixed by
https://github.com/neondatabase/neon/pull/7662. That PR also fixed this
case, but we still shouldn't make redundant get calls.
## Summary of changes
- Only call get in the create image layers loop at the end of a range if
some keys have been accumulated
## Problem
There is no global per-ep rate limiter in proxy.
## Summary of changes
* Return global per-ep rate limiter back.
* Rename weak compute rate limiter (the cli flags were not used
anywhere, so it's safe to rename).
## Problem
https://github.com/neondatabase/neon/pull/7637 breaks forward compat
test.
On commit ea531d448e.
https://neon-github-public-dev.s3.amazonaws.com/reports/main/8988324349/index.html
```
test_create_snapshot
2024-05-07T16:03:11.331883Z INFO version: git-env:ea531d448eb65c4f58abb9ef7d8cd461952f7c5f failpoints: true, features: ["testing"] launch_timestamp: 2024-05-07 16:03:11.316131763 UTC build_tag: build_tag-env:5159
test_forward_compatibility
2024-05-07T16:07:02.310769Z INFO version: git-env:ea531d448eb65c4f58abb9ef7d8cd461952f7c5f failpoints: true, features: ["testing"] launch_timestamp: 2024-05-07 16:07:02.294676183 UTC build_tag: build_tag-env:5159
```
The forward compatibility test is actually using the same tag as the
current build.
The commit before that,
https://neon-github-public-dev.s3.amazonaws.com/reports/main/8988126011/index.html
```
test_create_snapshot
2024-05-07T15:47:21.900796Z INFO version: git-env:2dbd1c1ed5cd0458933e8ffd40a9c0a5f4d610b8 failpoints: true, features: ["testing"] launch_timestamp: 2024-05-07 15:47:21.882784185 UTC build_tag: build_tag-env:5158
test_forward_compatibility
2024-05-07T15:50:48.828733Z INFO version: git-env:c4d7d5982553d2cf66634d1fbf85d95ef44a6524 failpoints: true, features: ["testing"] launch_timestamp: 2024-05-07 15:50:48.816635176 UTC build_tag: build_tag-env:release-5434
```
This pull request patches the bin path so that the new neon_local will
use the old binary.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
This PR does two things:
First, it fixes a bug with tiered compaction's k-merge implementation.
It ignored the lsn of a key during ordering, so multiple updates of the
same key could be read in arbitrary order, say from different layers.
For example there is layers `[(a, 2),(b, 3)]` and `[(a, 1),(c, 2)]` in
the heap, they might return `(a,2)` and `(a,1)`.
Ultimately, this change wasn't enough to fix the ordering issues in
#7296, in other words there is likely still bugs in the k-merge. So as
the second thing, we switch away from the k-merge to an in-memory based
one, similar to #4839, but leave the code around to be improved and
maybe switched to later on.
Part of #7296
## Problem
Storage controller was using a zero layer count in SecondaryProgress as
a proxy for "not initialized". However, in tenants with zero timelines
(a legitimate state), the layer count remains zero forever.
This caused https://github.com/neondatabase/neon/pull/7583 to
destabilize the storage controller scale test, which creates lots of
tenants, some of which don't get any timelines.
## Summary of changes
- Use a None mtime instead of zero layer count to determine if a
SecondaryProgress should be ignored.
- Adjust the test to use a shorter heatmap upload period to let it
proceed faster while waiting for scheduling optimizations to complete.
Before this PR, using the AWS SDK profile feature for running against
minio didn't work because
* our SDK versions were too old and didn't include
https://github.com/awslabs/aws-sdk-rust/issues/1060 and
* we didn't massage the s3 client config builder correctly.
This PR
* udpates all the AWS SDKs we use to, respectively, the latest version I
could find on crates.io (Is there a better process?)
* changes the way remote_storage constructs the S3 client, and
* documents how to run the test suite against real S3 & local minio.
Regarding the changes to `remote_storage`: if one reads the SDK docs, it
is clear that the recommended way is to use `aws_config::from_env`, then
customize.
What we were doing instead is to use the `aws_sdk_s3` builder directly.
To get the `local-minio` in the added docs working, I needed to update
both the SDKs and make the changes to the `remote_storage`. See the
commit history in this PR for details.
Refs:
* byproduct: https://github.com/smithy-lang/smithy-rs/pull/3633
* follow-up on deprecation:
https://github.com/neondatabase/neon/issues/7665
* follow-up for scrubber S3 setup:
https://github.com/neondatabase/neon/issues/7667
For aux file keys (v1 or v2) the vectored read path does not return an
error when they're missing. Instead they are omitted from the resulting
btree (this is a requirement, not a bug). Skip updating the metric in
these cases to avoid infinite results.
## Problem
See #6714, #6967
## Summary of changes
Completely ignore page header when comparing VM pages.
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
- Rename "filename" types which no longer map directly to a filename
(LayerFileName -> LayerName)
- Add a -v1- part to local layer paths to smooth the path to future
updates (we anticipate a -v2- that uses checksums later)
- Rename methods that refer to the string-ized version of a LayerName to
no longer be called "filename"
- Refactor reconcile() function to use a LocalLayerFileMetadata type
that includes the local path, rather than carrying local path separately
in a tuple and unwrap()'ing it later.
The test utils should only be used during tests. Users should not be
able to create this extension on their own.
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
We currently have no way to see what the current LSN of a compute its,
and in case of read replicas, we don't know what the difference in LSNs
is.
## Summary of changes
Adds these metrics
Fixes flaky test `test_gc_of_remote_layers`, which was failing because
of the `Nothing to GC` pageserver log.
I looked into the fails, it seems that backround `gc_loop` sometimes
started GC for initial tenant, which wasn't
configured to disable GC. The fix is to not create initial tenant with
enabled gc at all.
Fixes#7538
Before this PR, `neon_local` would store a copy of a subset of the
initial `pageserver.toml` in its `.neon/config`, e.g, `listen_pg_addr`.
That copy is represented as `struct PageServerConf`.
This copy was used to inform e.g., `neon_local endpoint` and other
commands that depend on Pageserver about which port to connect to.
The problem with that scheme is that the duplicated information in
`.neon/config` can get stale if `pageserver.toml` is changed.
This PR fixes that by eliminating populating `struct PageServerConf`
from the `pageserver.toml`s.
The `[[pageservers]]` TOML table in the `.neon/config` is obsolete.
As of this PR, `neon_local` will fail to start and print an error
informing about this change.
Code-level changes:
- Remove the `--pg-version` flag, it was only used for some checks
during `neon_local init`
- Remove the warn-but-continue behavior for when auth key creation fails
but auth keys are not required. It's just complexity that is unjustified
for a tool like `neon_local`.
- Introduce a type-system-level distinction between the runtime state
and the two (!) toml formats that are almost the same but not quite.
- runtime state: `struct PageServerConf`, now without `serde` derives
- toml format 1: the state in `.neon/config` => `struct OnDiskState`
- toml format 2: the `neon_local init --config TMPFILE` that, unlike
`struct OnDiskState`, allows specifying `pageservers`
- Remove `[[pageservers]]` from the `struct OnDiskState` and load the
data from the individual `pageserver.toml`s instead.
Improves the tiered compaction tests:
* Adds a new test that is a simpler version of the ignored
`test_many_updates_for_single_key` test.
* Reduces the amount of data that `test_many_updates_for_single_key`
processes to make it execute more quickly.
* Adds logging support.
Adds ordering asserts to the output of the delta key iterator
`MergeDeltaKeys` that implements a k-merge.
Part of #7296 : the asserts added by this PR get hit in the reproducers
of #7296 as well, but they are earlier in the pipeline.
## Problem
If a permit cannot be acquired to connect to compute, the cache is
invalidated. This had the observed affect of sending more traffic to
ProxyWakeCompute on cplane.
## Summary of changes
Make sure that permit acquire failures are marked as "should not
invalidate cache".
This does to `neon_local` what
https://github.com/neondatabase/aws/pull/1322 does to our production
deployment.
After both are merged, there are no users of `pageserver --init` /
`pageserver --config-override` left, and we can remove those flags
eventually.
Part of applying the changes from #7600. This piece *technically* can
change the semantics because now the context guard is held before
process_cli, but... the difference is likely quite small.
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
This commit is intentionally designed to have as small a diff as
possible. To that end, the basic idea is that each distinct "chunk" of
the previous main() has been wrapped in its own function, with the
return values from each function being passed directly into the next.
The structure of main() is now visible from its contents, which have a
handful of smaller functions.
There's a lot of other work that can / should(?) be done beyond this,
but I figure that's more opinionated, and this should be a solid start.
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
## Problem
Some HTTP client connections can stay open for quite a long time.
## Summary of changes
When there are too many HTTP client connections, pick a random
connection and gracefully cancel it.
## Problem
In https://github.com/neondatabase/neon/pull/7531, we would like to be
able to rewrite layers safely. One option is to make `Layer` able to
rewrite files in place safely (e.g. by blocking evictions/deletions for
an old Layer while a new one is created), but that's relatively fragile.
It's more robust in general if we simply never overwrite the same local
file: we can do that by putting the generation number in the filename.
## Summary of changes
- Add `local_layer_path` (counterpart to `remote_layer_path`) and
convert all locations that manually constructed a local layer path by
joining LayerFileName to timeline path
- In the layer upload path, construct remote paths with
`remote_layer_path` rather than trying to build them out of a local
path.
- During startup, carry the full path to layer files through
`init::reconcile`, and pass it into `Layer::for_resident`
- Add a test to make sure we handle upgrades properly.
- Comment out the generation part of `local_layer_path`, since we need
to maintain forward compatibility for one release. A tiny followup PR
will enable it afterwards.
We could make this a bit simpler if we bulk renamed existing layers on
startup instead of carrying literal paths through init, but that is
operationally risky on existing servers with millions of layer files. We
can always do a renaming change in future if it becomes annoying, but
for the moment it's kind of nice to have a structure that enables us to
change local path names again in future quite easily.
We should rename `LayerFileName` to `LayerName` or somesuch, to make it
more obvious that it's not a literal filename: this was already a bit
confusing where that type is used in remote paths. That will be a
followup, to avoid polluting this PR's diff.
This pull request adds the new basebackup read path + aux file write
path. In the regression test, all logical replication tests are run with
matrix aux_file_v2=false/true.
Also fixed the vectored get code path to correctly return missing key
error when being called from the unified sequential get code path.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
Preceding PR https://github.com/neondatabase/neon/pull/7613 reduced the
usage of `--pageserver-config-override`.
This PR builds on top of that work and fully removes the `neon_local
--pageserver-config-override`.
Tests that need a non-default `pageserver.toml` control it using two
options:
1. Specify `NeonEnvBuilder.pageserver_config_override` before
`NeonEnvBuilder.init_start()`. This uses a new `neon_local init
--pageserver-config` flag.
2. After `init_start()`: `env.pageserver.stop()` +
`NeonPageserver.edit_config_toml()` + `env.pageserver.start()`
A few test cases were using
`env.pageserver.start(overrides=("--pageserver-config-override...",))`.
I changed them to use one of the options above.
Future Work
-----------
The `neon_local init --pageserver-config` flag still uses `pageserver
--config-override` under the hood. In the future, neon_local should just
write the `pageserver.toml` directly.
The `NeonEnvBuilder.pageserver_config_override` field should be renamed
to `pageserver_initial_config`. Let's save this churn for a separate
refactor commit.
RemoteTimelineClient has a lot of mandatory cloning. By using a single
way of creating IndexPart out of UploadQueueInitialized we can simplify
things and also avoid cloning the latest files for each
`index_part.json` upload (the contents will still be cloned).
The `test_forward_compatibility` test runs the old production binaries,
but is supposed to always run the latest neon_local binary.
I think commit 6acbee23 broke that by accident because in that commit,
`from_repo_dir` is introduced and runs an `init_start()` before the
`test_forward_compatibility` gets a chance to patch up the
neon_local_binpath.
## Problem
Ref https://neondb.slack.com/archives/C036U0GRMRB/p1688122168477729
## Summary of changes
- Add sha from postgres repo into postgres version string (via
`--with-extra-version`)
- Add a test that Postgres version matches the expected one
- Remove build-time hard check and allow only related tests to fail
## Problem
Timelines cannot be deleted if they have children. In many production
cases, a branch or a timeline has been created off the main branch for
various reasons to the effect of having now a "new main" branch. This
feature will make it possible to detach a timeline from its ancestor by
inheriting all of the data before the branchpoint to the detached
timeline and by also reparenting all of the ancestor's earlier branches
to the detached timeline.
## Summary of changes
- Earlier added copy_lsn_prefix functionality is used
- RemoteTimelineClient learns to adopt layers by copying them from
another timeline
- LayerManager adds support for adding adopted layers
-
`timeline::Timeline::{prepare_to_detach,complete_detaching}_from_ancestor`
and `timeline::detach_ancestor` are added
- HTTP PUT handler
Cc: #6994
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
After a shard split of a large existing tenant, child tenants can end up
with oversized historic layers indefinitely, if those layers are
prevented from being GC'd by branchpoints.
This PR is followed by https://github.com/neondatabase/neon/pull/7531
Related issue: https://github.com/neondatabase/neon/issues/7504
## Summary of changes
- Add a new compaction phase `compact_shard_ancestors`, which identifies
layers that are no longer needed after a shard split.
- Add a Timeline->LayerMap code path called `rewrite_layers` , which is
currently only used to drop layers, but will later be used to rewrite
them as well in https://github.com/neondatabase/neon/pull/7531
- Add a new test that compacts after a split, and checks that something
is deleted.
Note that this doesn't have much impact on a tenant's resident size
(since unused layers would end up evicted anyway), but it:
- Makes index_part.json much smaller
- Makes the system easier to reason about: avoid having tenants which
are like "my physical size is 4TiB but don't worry I'll never actually
download it", instead have tenants report the real physical size of what
they might download.
Why do we remove these layers in compaction rather than during the
split? Because we have existing split tenants that need cleaning up. We
can add it to the split operation in future as an optimization.
## Problem
Usually, the connection itself is quite fast (bellow 10ms for p999:
https://neonprod.grafana.net/goto/aOyn8vYIg?orgId=1).
It doesn't make a lot of sense to wait for a lot of time for the lock,
if it takes a lot of time to acquire it, probably, something goes wrong.
We also spawn a lot of retries, but they are not super helpful (0 means
that it was connected successfully, 1, most probably, that it was
re-request of the compute node address
https://neonprod.grafana.net/goto/J_8VQvLIR?orgId=1). Let's try to keep
a small number of retries.
## Problem
There's allegedly a bug where if we connect a subscriber before WAL is
downloaded from the safekeeper, it creates an error.
## Summary of changes
Adds support for pausing safekeepers from sending WAL to computes, and
then creates a compute and attaches a subscriber while it's in this
paused state. Fails to reproduce the issue, but probably a good test to
have
---------
Co-authored-by: Arseny Sher <sher-ars@yandex.ru>
Changes parameters to fix the flakiness of `test_ts_of_lsn_api`. Already
now, the amount of flakiness of the test is pretty low. With this, it's
even lower.
cc #5768
## Problem
The current `tenant_slots` metric becomes less useful once we have lots
of secondaries, because we can't tell how many tenants are really
attached (without doing a sum() on some other metric).
## Summary of changes
- Add a `mode` label to this metric
- Update the metric with `slot_added` and `slot_removed` helpers that
are called at all the places we mutate the tenants map.
- Add a debug assertion at shutdown that checks the metrics add up to
the right number, as a cheap way of validating that we're calling the
metric hooks in all the right places.
## Problem
In testing of the earlier fix for OOMs under heavy write load
(https://github.com/neondatabase/neon/pull/7218), we saw that the limit
on ephemeral layer size wasn't being reliably enforced. That was
diagnosed as being due to overwhelmed compaction loops: most tenants
were waiting on the semaphore for background tasks, and thereby not
running the function that proactively rolls layers frequently enough.
Related: https://github.com/neondatabase/neon/issues/6939
## Summary of changes
- Create a new per-tenant background loop for "ingest housekeeping",
which invokes maybe_freeze_ephemeral_layer() without taking the
background task semaphore.
- Downgrade to DEBUG a log line in maybe_freeze_ephemeral_layer that had
been INFO, but turns out to be pretty common in the field.
There's some discussion on the issue
(https://github.com/neondatabase/neon/issues/6939#issuecomment-2083554275)
about alternatives for calling this maybe_freeze_epemeral_layer
periodically without it getting stuck behind compaction. A whole task
just for this feels like kind of a big hammer, but we may in future find
that there are other pieces of lightweight housekeeping that we want to
do here too.
Why is it okay to call maybe_freeze_ephemeral_layer outside of the
background tasks semaphore?
- this is the same work we would do anyway if we receive writes from the
safekeeper, just done a bit sooner.
- The period of the new task is generously jittered (+/- 5%), so when
the ephemeral layer size tips over the threshold, we shouldn't see an
excessively aggressive thundering herd of layer freezes (and only layers
larger than the mean layer size will be frozen)
- All that said, this is an imperfect approach that relies on having a
generous amount of RAM to dip into when we need to freeze somewhat
urgently. It would be nice in future to also block compaction/GC when we
recognize resource stress and need to do other work (like layer
freezing) to reduce memory footprint.
Previously its segment header and page header of first record weren't
initialized because compute streams data only since first record LSN. Also, fix
a bug in the existing code for initialization: xlp_rem_len must not include page
header.
These changes make first segment pg_waldump'able.
- On a non-pooled start, do not reset the 'start_time' after launching
the HTTP service. In a non-pooled start, it's fair to include that in
the total startup time.
- When setting wait_for_spec_ms and resetting start_time, call
Utc::now() only once. It's a waste of cycles to call it twice, but also,
it ensures the time between setting wait_for_spec_ms and resetting
start_time is included in one or the other time period.
These differences should be insignificant in practice, in the
microsecond range, but IMHO it seems more logical and readable this way
too. Also fix and clarify some of the surrounding comments.
(This caught my eye while reviewing PR #7577)
Part of neondatabase/cloud#12047. Resolves#7239.
In short, this PR:
1. Adds `ComputeSpec.swap_size_bytes: Option<u64>`
2. Adds a flag to compute_ctl: `--resize-swap-on-bind`
3. Implements running `/neonvm/bin/resize-swap` with the value from the
compute spec before starting postgres, if both the value in the spec
*AND* the flag are specified.
4. Adds `sudo` to the final image
5. Adds a file in `/etc/sudoers.d` to allow `compute_ctl` to resize swap
Various bits of reasoning about design decisions in the added comments.
In short: We have both a compute spec field and a flag to make rollout
easier to implement. The flag will most likely be removed as part of
cleanups for neondatabase/cloud#12047.
This is the first step towards representing all of Pageserver
configuration as clean `serde::Serialize`able Rust structs in
`pageserver_api`.
The `neon_local` code will then use those structs instead of the crude
`toml_edit` / string concatenation that it does today.
refs https://github.com/neondatabase/neon/issues/7555
---------
Co-authored-by: Alex Chi Z <iskyzh@gmail.com>
## Problem
Too many connect_compute attempts can overwhelm postgres, getting the
connections stuck.
## Summary of changes
Limit number of connection attempts that can happen at a given time.
This pull request adds the scan interface. Scan operates on a sparse
keyspace and retrieves all the key-value pairs from the keyspaces.
Currently, scan only supports the metadata keyspace, and by default do
not retrieve anything from the ancestor branch. This should be fixed in
the future if we need to have some keyspaces that inherits from the
parent.
The scan interface reuses the vectored get code path by disabling the
missing key errors.
This pull request also changes the behavior of vectored get on aux file
v1/v2 key/keyspace: if the key is not found, it is simply not included in the
result, instead of throwing a missing key error.
TODOs in future pull requests: limit memory consumption, ensure the
search stops when all keys are covered by the image layer, remove
`#[allow(dead_code)]` once the code path is used in basebackups / aux
files, remove unnecessary fine-grained keyspace tracking in vectored get
(or have another code path for scan) to improve performance.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
The logic in Service::optimize_all would sometimes choose to migrate a
tenant to a secondary location that was only recently created, resulting
in Reconciler::live_migrate hitting its 5 minute timeout warming up the
location, and proceeding to attach a tenant to a location that doesn't
have a warm enough local set of layer files for good performance.
Closes: #7532
## Summary of changes
- Add a pageserver API for checking download progress of a secondary
location
- During `optimize_all`, connect to pageservers of candidate
optimization secondary locations, and check they are warm.
- During shard split, do heatmap uploads and start secondary downloads,
so that the new shards' secondary locations start downloading ASAP,
rather than waiting minutes for background downloads to kick in.
I have intentionally not implemented this by continuously reading the
status of locations, to avoid dealing with the scale challenge of
efficiently polling & updating 10k-100k locations status. If we
implement that in the future, then this code can be simplified to act
based on latest state of a location rather than fetching it inline
during optimize_all.
We keep the practice of keeping the compiler up to date, pointing to the
latest release. This is done by many other projects in the Rust
ecosystem as well.
Release notes: https://blog.rust-lang.org/2024/05/02/Rust-1.78.0.html
Prior update was in #7198
## Problem
After some time the load from heatmap uploads gets rather spiky. They're
unintentionally synchronising.
Chart (does this make a _boing_ sound in anyone else's head?):
## Summary of changes
- Add a helper `period_jitter` and apply a 5% jitter from downloader and
heatmap_uploader when updating the next runtime at the end of an
interation.
- Refactor existing places that we pick a startup interval into
`period_warmup`, so that the intent is obvious.
The current implementation of finding timeline gc cutoff Lsn(s) is done
while holding `Tenant::gc_cs`. In recent incidents long create branch
times were caused by holding the `Tenant::gc_cs` over extremely long
`Timeline::find_lsn_by_timestamp`. The fix is to find the GC cutoff
values before taking the `Tenant::gc_cs` lock. This change is safe to do
because the GC cutoff values and the branch points have no dependencies
on each other. In the case of `Timeline::find_gc_cutoff` taking a long
time with this change, we should no longer see `Tenant::gc_cs`
interfering with branch creation.
Additionally, the `Tenant::refresh_gc_info` is now tolerant of timeline
deletions (or any other failures to find the pitr_cutoff). This helps
with the synthetic size calculation being constantly completed instead
of having a break for a timely timeline deletion.
Fixes: #7560Fixes: #7587
## Problem
We were matching on `/tenant/:tenant_id` and
`/tenant/:tenant_id/timeline*`, but not non-timeline tenant sub-paths.
There aren't many: this was only noticeable when using the
synthetic_size endpoint by hand.
## Summary of changes
- Change the wildcard from `/tenant/:tenant_id/timeline*` to
`/tenant/:tenant_id/*`
- Add test lines that exercise this
## Problem
This test triggers layer download failures on demand. It is possible to
modify the failpoint
during a `Timeline::get_vectored` right between the vectored read and
it's validation read.
This means that one of the reads can fail while the other one succeeds
and vice versa.
## Summary of changes
These errors are expected, so allow them to happen.
Split `GcInfo` and replace `Timeline::update_gc_info` with a method that
simply finds gc cutoffs `Timeline::find_gc_cutoffs` to be combined as
`Timeline::gc_info` at the caller.
This change will be followed up with a change that finds the GC cutoff
values before taking the `Tenant::gc_cs` lock.
Cc: #7560
## Problem
Issues around operation and tenant locks would have been hard to debug
since there was little observability around them.
## Summary of changes
- As suggested in the issue, a wrapper was added around
`OwnedRwLockWriteGuard` called `IdentifierLock` that removes the
operation currently holding the exclusive lock when it's dropped.
- The value in `IdLockMap` was extended to hold a pair of locks and
operations that can be accessed and locked independently.
- When requesting an exclusive lock besides returning the lock on that
resource, an operation is changed if the lock is acquired.
Closes https://github.com/neondatabase/neon/issues/7108
The main challenge was in the second commit, as `DownloadStream`
requires the inner to be Sync but the stream returned by the Azure SDK
wasn't Sync.
This left us with three options:
* Change the Azure SDK to return Sync streams. This was abandoned after
we realized that we couldn't just make `TokenCredential`'s returned
future Sync: it uses the `async_trait` macro and as the
`TokenCredential` trait is used in dyn form, one can't use Rust's new
"async fn in Trait" feature.
* Change `DownloadStream` to not require `Sync`. This was abandoned
after it turned into a safekeeper refactoring project.
* Put the stream into a `Mutex` and make it obtain a lock on every poll.
This adds some performance overhead but locks that actually don't do
anything should be comparatively cheap.
We went with the third option in the end as the change still represents
an improvement.
Follow up of #5446 , fixes#5563
As discussed in https://github.com/neondatabase/autoscaling/pull/895, we
want to have a separate sql_exporter for simple metrics to avoid
overload the database because the autoscaling agent needs to scrape at a
higher interval. The new exporter is exposed at port 9499. I didn't do
any testing for this pull request but given it's just a configuration
change I assume this works.
Signed-off-by: Alex Chi Z <chi@neon.tech>
This allows passing a humantime string in the CLI to configure the
initial wait for the database.
It defaults to the previously hard-coded value of 5 seconds.
Some part of the code requires missing key error to be propagated to the
code path correctly (i.e., aux key range scan). Currently, it's an
anyhow error.
* remove `stuck_lsn` from the missing key error.
* as a result, when matching missing key, we do not distinguish the case
`stuck_lsn = false/true`.
* vectored get now use the unified missing key error.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
Instead of showing the full path of layer traversal, we now only show
tenant (in tracing context)+timeline+filename.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
We had an incident where pageserver requests timed out because
pageserver couldn't fetch WAL from safekeepers. This incident was caused
by a bug in safekeeper logic for timeline activation, which prevented
pageserver from finding safekeepers.
This bug was since fixed, but there is still a chance of a similar bug
in the future due to overall complexity.
We add a new broker message to "signal interest" for timeline. This
signal will be sent by pageservers `wait_lsn`, and safekeepers will
receive this signal to start broadcasting broker messages. Then every
broker subscriber will be able to find the safekeepers and connect to
them (to start fetching WAL).
This feature is not limited to pageservers and any service that wants to
download WAL from safekeepers will be able to use this discovery
request.
This commit changes pageserver's connection_manager (walreceiver) to
send a SafekeeperDiscoveryRequest when there is no information about
safekeepers present in memory. Current implementation will send these
requests only if there is an active wait_lsn() call and no more often
than once per 10 seconds.
Add `test_broker_discovery` to test this: safekeepers started with
`--disable-periodic-broker-push` will not push info to broker so that
pageserver must use a discovery to start fetching WAL.
Add task_stats in safekeepers broker module to log a warning if there is
no message received from the broker for the last 10 seconds.
Closes#5471
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
The current Makefile assumes that homebrew is used on macos. There are
other ways to install dependencies on MacOS (nix, macports, "manually").
It would be great to allow the one who wants to use other options to
disable homebrew integration.
## Summary of changes
It adds DISABLE_HOMEBREW variable that if set skips extra
homebrew-specific configuration steps.
## Problem
This test became flaky recently with failures like:
```
AssertionError: Log errors on storage_controller: (129, '2024-04-29T16:41:03.591506Z ERROR request{method=PUT path=/control/v1/tenant/b38c0447fbdbcf4e1c023f00b0f7c221/shard_split request_id=34df4975-2ef3-4ed8-b167-2956650e365c}: Error processing HTTP request: InternalServerError(Reconcile error on shard b38c0447fbdbcf4e1c023f00b0f7c221-0002: Cancelled\n')
```
Likely due to #7508 changing how errors are reported from Reconcilers.
## Summary of changes
- Tolerate `Reconcile error.*Cancelled` log errors
## Problem
Storage controller was observed to have unexpectedly large memory
consumption when loaded with many thousands of shards.
This was recently fixed:
- https://github.com/neondatabase/neon/pull/7493
...but we need a general test that the controller is well behaved with
thousands of shards.
Closes: https://github.com/neondatabase/neon/issues/7460
Closes: https://github.com/neondatabase/neon/issues/7463
## Summary of changes
- Add test test_storage_controller_many_tenants to exercise the system's
behaviour with a more substantial workload. This test measures memory
consumption and reproduces #7460 before the other changes in this PR.
- Tweak reconcile_all's return value to make it nonzero if it spawns no
reconcilers, but _would_ have spawned some reconcilers if they weren't
blocked by the reconcile concurrency limit. This makes the test's
reconcile_until_idle behave as expected (i.e. not complete until the
system is nice and calm).
- Fix an issue where tenant migrations would leave a spurious secondary
location when migrated to some location that was not already their
secondary (this was an existing low-impact bug that tripped up the
test's consistency checks).
On the test with 8000 shards, the resident memory per shard is about
20KiB. This is not really per-shard memory: the primary source of memory
growth is the number of concurrent network/db clients we create.
With 8000 shards, the test takes 125s to run on my workstation.
- pageserver_id in project details is now is optional, fix it
- add active_timeline_count guard/stat similar to active_tenant_count
- fix safekeeper prefix
- count and log deleted keys
It works by listing postgres table with memory dump of safekeepers state. s3
contents for each timeline are checked then against timeline_start_lsn and
backup_lsn. If inconsistency is found, before complaining timeline (branch) is
checked at control plane; it might have been deleted between the dump take and
s3 check.
Makes two of the tests work with the tiered compaction that I had to
ignore in #7283.
The issue was that tiered compaction actually created image layers, but
the keys didn't appear in them as `collect_keyspace` didn't include
them. Not a compaction problem, but due to how the test is structured.
Fixes#7287
## Problem
`init_tenant_mgr` blocks the rest of pageserver startup, including
starting the admin API.
This was noticeable in #7475 , where the init_tenant_mgr runtime could
be long enough to trip the controller's 30 second heartbeat timeout.
## Summary of changes
- When detaching tenants during startup, spawn the background deletes as
background tasks instead of doing them inline
- Write all configs before spawning any tenants, so that the config
writes aren't fighting tenants for system resources
- Write configs with some concurrency (16) rather than writing them all
sequentially.
extracted (and tested) from
https://github.com/neondatabase/neon/pull/7468, part of
https://github.com/neondatabase/neon/issues/7462.
The current codebase assumes the keyspace is dense -- which means that
if we have a keyspace of 0x00-0x100, we assume every key (e.g., 0x00,
0x01, 0x02, ...) exists in the storage engine. However, the assumption
does not hold any more in metadata keyspace. The metadata keyspace is
sparse. It is impossible to do per-key check.
Ideally, we should not have the assumption of dense keyspace at all, but
this would incur a lot of refactors. Therefore, we split the keyspaces
we have to dense/sparse and handle them differently in the code for now.
At some point in the future, we should assume all keyspaces are sparse.
## Summary of changes
* Split collect_keyspace to return dense+sparse keyspace.
* Do not allow generating image layers for sparse keyspace (for now --
will fix this next week, we need image layers anyways).
* Generate delta layers for sparse keyspace.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
Not sure if this should actually be a link pointing to the
`persistence.rs` file but following the conventions of the rest of the
file, change `persistence.rs` reference to simply be a file name
mention.
## Problem
Followup to https://github.com/neondatabase/neon/pull/6776
While #6776 makes compaction safe on sharded tenants, the logic for
keyspace partitioning remains inefficient: it assumes that the size of
data on a pageserver can be calculated simply as the range between start
and end of a Range -- this is not the case in sharded tenants, where
data within a range belongs to a variety of shards.
Closes: https://github.com/neondatabase/neon/issues/6774
## Summary of changes
I experimented with using a sharding-aware range type in KeySpace to
replace all the Range<Key> uses, but the impact on other code was quite
large (many places use the ranges), and not all of them need this
property of being able to approximate the physical size of data within a
key range.
So I compromised on expressing this as a ShardedRange type, but only
using that type selctively: during keyspace repartition, and in tiered
compaction when accumulating key ranges.
- keyspace partitioning methods take sharding parameters as an input
- new `ShardedRange` type wraps a Range<Key> and a shard identity
- ShardedRange::page_count is the shard-aware replacement for
key_range_size
- Callers that don't need to be shard-aware (e.g. vectored get code that
just wants to count the number of keys in a keyspace) can use
ShardedRange::raw_size to get the faster, shard-naive code (same as old
`key_range_size`)
- Compaction code is updated to carry a shard identity so that it can
use shard aware calculations
- Unit tests for the new fragmentation logic.
- Add a test for compaction on sharded tenants, that validates that we
generate appropriately sized image layers (this fails before fixing
keyspace partitioning)
previously in https://github.com/neondatabase/neon/pull/7375, we
observed that for in-memory layers, we will need to iterate every key in
the key space in order to get the result. The operation can be more
efficient if we use BTreeMap as the in-memory layer representation, even
if we are doing vectored get in a dense keyspace. Imagine a case that
the in-memory layer covers a very little part of the keyspace, and most
of the keys need to be found in lower layers. Using a BTreeMap can
significantly reduce probes for nonexistent keys.
## Summary of changes
* Use BTreeMap as in-memory layer representation.
* Optimize the vectored get flow to utilize the range scan functionality
of BTreeMap.
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
Sequential get runs after vectored get, so it is possible for the later
to time out while waiting for its ancestor's Lsn to become ready and for
the former to succeed (it essentially has a doubled wait time).
## Summary of Changes
Relax the validation to allow for such rare cases.
## Problem
Benchmarks don't use the vectored read path.
## Summary of changes
* Update the benchmarks to use the vectored read path for both singular
and vectored gets.
* Disable validation for the benchmarks
## Problem
It's not possible to get the duration of the session from proxy events.
## Summary of changes
* Added a separate events folder in s3, to record disconnect events.
* Disconnect events are exactly the same as normal events, but also have
`disconnect_timestamp` field not empty.
* @oruen suggested to fill it with the same information as the original
events to avoid potentially heavy joins.
## Problem
Downloading tenant data for analysis/debug with `aws s3 cp` works well
for small tenants, but for larger tenants it is unlikely that one ends
up with an index that matches layer files, due to the time taken to
download.
## Summary of changes
- Add a `tenant-snapshot` command to the scrubber, which reads timeline
indices and then downloads the layers referenced in the index, even if
they were deleted. The result is a snapshot of the tenant's remote
storage state that should be usable when imported (#7399 ).
## Problem
Previously, we try to send compute notifications in startup_reconcile
before completing that function, with a time limit. Any notifications
that don't happen within the time limit result in tenants having their
`pending_compute_notification` flag set, which causes them to spawn a
Reconciler next time the background reconciler loop runs.
This causes two problems:
- Spawning a lot of reconcilers after startup caused a spike in memory
(this is addressed in https://github.com/neondatabase/neon/pull/7493)
- After https://github.com/neondatabase/neon/pull/7493, spawning lots of
reconcilers will block some other operations, e.g. a tenant creation
might fail due to lack of reconciler semaphore units while the
controller is busy running all the Reconcilers for its startup compute
notifications.
When the code was first written, ComputeHook didn't have internal
ordering logic to ensure that notifications for a shard were sent in the
right order. Since that was added in
https://github.com/neondatabase/neon/pull/7088, we can use it to avoid
waiting for notifications to complete in startup_reconcile.
Related to: https://github.com/neondatabase/neon/issues/7460
## Summary of changes
- Add a `notify_background` method to ComputeHook.
- Call this from startup_reconcile instead of doing notifications inline
- Process completions from `notify_background` in `process_results`, and
if a notification failed then set the `pending_compute_notification`
flag on the shard.
The result is that we will only spawn lots of Reconcilers if the compute
notifications _fail_, not just because they take some significant amount
of time.
Test coverage for this case is in
https://github.com/neondatabase/neon/pull/7475
## Problem
Sometimes we have test data in the form of S3 contents that we would
like to run live in a neon_local environment.
## Summary of changes
- Add a storage controller API that imports an existing tenant.
Currently this is equivalent to doing a create with a high generation
number, but in future this would be something smarter to probe S3 to
find the shards in a tenant and find generation numbers.
- Add a `neon_local` command that invokes the import API, and then
inspects timelines in the newly attached tenant to create matching
branches.
extracted from https://github.com/neondatabase/neon/pull/7468, part of
https://github.com/neondatabase/neon/issues/7462.
In the page server, we use i128 (instead of u128) to do the integer
representation of the key, which indicates that the highest bit of the
key should not be 1. This constraints our keyspace to <= 0x7F.
Also fix the bug of `to_i128` that dropped the highest 4b. Now we keep
3b of them, dropping the sign bit.
And on that, we shrink the metadata keyspace to 0x60-0x7F for now, and
once we add support for u128, we can have a larger metadata keyspace.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
Changing metadata format is not easy. This pull request adds a
tenant-level flag on whether to enable aux file v2. As long as we don't
roll this out to the user and guarantee our staging projects can persist
tenant config correctly, we can test the aux file v2 change with setting
this flag. Previous discussion at
https://github.com/neondatabase/neon/pull/7424.
Signed-off-by: Alex Chi Z <chi@neon.tech>
Implements an approach different from the one #7488 chose: We now return
`past` instead of `present` (or`future`) when encountering the edge case
where commit_lsn < min_lsn. In my opinion, both `past` and `present` are
correct responses, but past is slightly better as the lsn returned by
`present` with #7488 is one too "new". In practice, this shouldn't
matter much, but shrug.
We agreed in slack that this is the better approach:
https://neondb.slack.com/archives/C03F5SM1N02/p1713871064147029
## Problem
PR #7230 attempted to introduce a WAL ingest threshold for checking
whether enough deltas are stacked to warrant creating a new image layer.
However, this check was incorrectly performed at the compaction
partition level instead of the timeline level. Hence, it inhibited GC
for any keys outside of the first partition.
## Summary of Changes
Hoist the check up to the timeline level.
part of https://github.com/neondatabase/neon/issues/7124
# Problem
(Re-stating the problem from #7124 for posterity)
The `test_bulk_ingest` benchmark shows about 2x lower throughput with
`tokio-epoll-uring` compared to `std-fs`.
That's why we temporarily disabled it in #7238.
The reason for this regression is that the benchmark runs on a system
without memory pressure and thus std-fs writes don't block on disk IO
but only copy the data into the kernel page cache.
`tokio-epoll-uring` cannot beat that at this time, and possibly never.
(However, under memory pressure, std-fs would stall the executor thread
on kernel page cache writeback disk IO. That's why we want to use
`tokio-epoll-uring`. And we likely want to use O_DIRECT in the future,
at which point std-fs becomes an absolute show-stopper.)
More elaborate analysis:
https://neondatabase.notion.site/Why-test_bulk_ingest-is-slower-with-tokio-epoll-uring-918c5e619df045a7bd7b5f806cfbd53f?pvs=4
# Changes
This PR increases the buffer size of `blob_io` and `EphemeralFile` from
PAGE_SZ=8k to 64k.
Longer-term, we probably want to do double-buffering / pipelined IO.
# Resource Usage
We currently do not flush the buffer when freezing the InMemoryLayer.
That means a single Timeline can have multiple 64k buffers alive, esp if
flushing is slow.
This poses an OOM risk.
We should either bound the number of frozen layers
(https://github.com/neondatabase/neon/issues/7317).
Or we should change the freezing code to flush the buffer and drop the
allocation.
However, that's future work.
# Performance
(Measurements done on i3en.3xlarge.)
The `test_bulk_insert.py` is too noisy, even with instance storage. It
varies by 30-40%. I suspect that's due to compaction. Raising amount of
data by 10x doesn't help with the noisiness.)
So, I used the `bench_ingest` from @jcsp 's #7409 .
Specifically, the `ingest-small-values/ingest 128MB/100b seq` and
`ingest-small-values/ingest 128MB/100b seq, no delta` benchmarks.
| | | seq | seq, no delta |
|-----|-------------------|-----|---------------|
| 8k | std-fs | 55 | 165 |
| 8k | tokio-epoll-uring | 37 | 107 |
| 64k | std-fs | 55 | 180 |
| 64k | tokio-epoll-uring | 48 | 164 |
The `8k` is from before this PR, the `64k` is with this PR.
The values are the throughput reported by the benchmark (MiB/s).
We see that this PR gets `tokio-epoll-uring` from 67% to 87% of `std-fs`
performance in the `seq` benchmark. Notably, `seq` appears to hit some
other bottleneck at `55 MiB/s`. CC'ing #7418 due to the apparent
bottlenecks in writing delta layers.
For `seq, no delta`, this PR gets `tokio-epoll-uring` from 64% to 91% of
`std-fs` performance.
part of https://github.com/neondatabase/neon/issues/7124
Changes
-------
This PR replaces the `EphemeralFile::write_blob`-specifc `struct Writer`
with re-use of `owned_buffers_io::write::BufferedWriter`.
Further, it restructures the code to cleanly separate
* the high-level aspect of EphemeralFile's write_blob / read_blk API
* the page-caching aspect
* the aspect of IO
* performing buffered write IO to an underlying VirtualFile
* serving reads from either the VirtualFile or the buffer if it hasn't
been flushed yet
* the annoying "feature" that reads past the end of the written range
are allowed and expected to return zeroed memory, as long as one remains
within one PAGE_SZ
These are testability/logging improvements spun off from #7475
- Don't log warnings for shutdown errors in compute hook
- Revise logging around heartbeats and reconcile_all so that we aren't
emitting such a large volume of INFO messages under normal quite
conditions.
- Clean up the `last_error` of TenantShard to hold a ReconcileError
instead of a String, and use that properly typed error to suppress
reconciler cancel errors during reconcile_all_now. This is important for
tests that iteratively call that, as otherwise they would get 500 errors
when some reconciler in flight was cancelled (perhaps due to a state
change on the tenant shard starting a new reconciler).
Before PR #7377, on-demand SLRU download always used the basebackup's
LSN in the SLRU download, but that LSN might get garbage-collected away
in the pageserver. We should request the latest LSN, like with GetPage
requests, with the LSN just indicating that we know that the page hasn't
been changed since the LSN (since the basebackup in this case).
Add test to demonstrate the problem. Without the fix, it fails with
"tried to request a page version that was garbage collected" error from
the pageserver.
I wrote this test as part of earlier PR #6693, but that fell through
the cracks and was never applied. PR #7377 superseded the fix from
that older PR, but the test is still valid.
Instead of thinking in terms of 'latest' and 'lsn' of the request,
each request has two LSNs: the request LSN and 'not_modified_since'
LSN. The request is nominally made at the request LSN, that determines
what page version we want to see. But as a hint, we also include
'not_modified_since'. It tells the pageserver that the page has not
been modified since that LSN, which allows the pageserver to skip
waiting for newer WAL to arrive, and could allow more optimizations in
the future.
Refactor the internal functions to calculate the request LSN to
calculate both LSNs.
Sending two LSNs to the pageserver requires using the new protocol
version 2. The previous commit added the server support for it, but we
still default to the old protocol for compatibility with old
pageservers. The 'neon.protocol_version' GUC can be used to use the
new protocol.
The new protocol addresses one cause of issue #6211, although you can
still get the same error if you have a standby that is lagging behind
so that the page version it needs is genuinely GC'd away.
In the old protocol version, the client sent with each request:
- latest: bool. If true, the client requested the latest page
version, and the 'lsn' was just a hint of when the page was last
modified
- lsn: Lsn, the page version to return
This protocol didn't allow requesting a page at a particular
non-latest LSN and *also* sending a hint on when the page was last
modified. That put a read only compute into an awkward position where
it had to either request each page at the replay-LSN, which could be
very close to the last LSN written in the primary and therefore
require the pageserver to wait for it to arrive, or an older LSN which
could already be garbage collected in the pageserver, resulting in an
error. The new protocol version fixes that by allowing a read only
compute to send both LSNs.
To use the new protocol version, use "pagestream_v2" command instead
of just "pagestream". The old protocol version is still supported, for
compatibility with old computes (and in fact there is no client
support yet, it is added by the next commit).
The 'latest' argument was passed to the functions in
pgdatadir_mapping.rs to know when they can update the relsize
cache. Commit e69ff3fc00 changed how the relsize cache is updated,
making the 'latest' argument unused.
## Problem
Start switching from the global redis to the regional one
## Summary of changes
* Publish cancellations to the regional redis
* Listen notifications from both: global and regional
## Problem
We are currently supporting two read paths. No bueno.
## Summary of changes
High level: use vectored read path to serve get page requests - gated by
`get_impl` config
Low level:
1. Add ps config, `get_impl` to specify which read path to use when
serving get page requests
2. Fix base cached image handling for the vectored read path. This was
subtly broken: previously we
would not mark keys that went past their cached lsn as complete. This is
a self standing change which
could be its own PR, but I've included it here because writing separate
tests for it is tricky.
3. Fork get page to use either the legacy or vectored implementation
4. Validate the use of vectored read path when serving get page requests
against the legacy implementation.
Controlled by `validate_vectored_get` ps config.
5. Use the vectored read path to serve get page requests in tests (with
validation).
## Note
Since the vectored read path does not go through the page cache to read
buffers, this change also amounts to a removal of the buffer page cache. Materialized page cache
is still used.
## Problem
The `WithClientIp` AsyncRead/Write abstraction never filled me with much
joy. I would just rather read the protocol header once and then get the
remaining buf and reader.
## Summary of changes
* Replace `WithClientIp::wait_for_addr` with `read_proxy_protocol`.
* Replace `WithClientIp` with `ChainRW`.
* Optimise `ChainRW` to make the standard path more optimal.
## Problem
Storage controller memory can spike very high if we have many tenants
and they all try to reconcile at the same time.
Related:
- https://github.com/neondatabase/neon/issues/7463
- https://github.com/neondatabase/neon/issues/7460
Not closing those issues in this PR, because the test coverage for them
will be in https://github.com/neondatabase/neon/pull/7475
## Summary of changes
- Add a CLI arg `--reconciler-concurrency`, defaulted to 128
- Add a semaphore to Service with this many units
- In `maybe_reconcile_shard`, try to acquire semaphore unit. If we can't
get one, return a ReconcileWaiter for a future sequence number, and push
the TenantShardId onto a channel of delayed IDs.
- In `process_result`, consume from the channel of delayed IDs if there
are semaphore units available and call maybe_reconcile_shard again for
these delayed shards.
This has been tested in https://github.com/neondatabase/neon/pull/7475,
but will land that PR separately because it contains other changes &
needs the test stabilizing. This change is worth merging sooner, because
it fixes a practical issue with larger shard counts.
Currently we move data to the intended storage class via lifecycle
rules, but those are a daily batch job so data first spends up to a day
in standard storage.
Therefore, make it possible to specify the storage class used for
uploads to S3 so that the data doesn't have to be migrated
automatically.
The advantage of this is that it gives cleaner billing reports.
Part of https://github.com/neondatabase/cloud/issues/11348
## Problem
If the previous step of the vectored left no further keyspace to
investigate (i.e. keyspace remains empty after removing keys completed in the previous step),
then we'd still grab the layers lock, potentially add an in-mem layer to the fringe
and at some further point read its index without reading any values from it.
## Summary of changes
If there's nothing left in the current keyspace, then skip the search
and just select the next item from the fringe as usual.
When running `test_pg_regress[release-pg16]` with the vectored read path
for singular gets this improved perf drastically (see PR cover letter).
## Correctness
Since no keys remained from the previous range (i.e. we are on a leaf
node) there's nothing that search can find in deeper nodes.
## Problem
Vectored and non-vectored read paths don't publish the same set of
metrics. Metrics parity is needed for coalescing the read paths.
## Summary of changes
* Publish reconstruct time and fetching data for reconstruct time from
the vectored read path
* Remove pageserver_getpage_reconstruct_seconds{res="err"} - wasn't used
anyway
As seen with a recent incident, eviction tasks can cause pageserver-wide
permit starvation on the background task semaphore when synthetic size
calculation takes a long time for a tenant that has more than our permit
number of timelines or multiple tenants that have slow synthetic size
and total number of timelines exceeds the permits. Metric links can be
found in the internal [slack thread].
As a solution, release the permit while waiting for the state guarding
the synthetic size calculation. This will most likely hurt the eviction
task eviction performance, but that does not matter because we are
hoping to get away from it using OnlyImitiate policy anyway and rely
solely on disk usage-based eviction.
[slack thread]:
https://neondb.slack.com/archives/C06UEMLK7FE/p1713810505587809?thread_ts=1713468604.508969&cid=C06UEMLK7FE
There was an edge case where
`get_lsn_by_timestamp`/`find_lsn_for_timestamp` could have returned an
lsn that is before the limits we enforce: when we did find SLRU entries
with timestamps before the one we search for.
The API contract of `get_lsn_by_timestamp` is to not return something
before the anchestor lsn.
cc https://neondb.slack.com/archives/C03F5SM1N02/p1713871064147029
## Problem
We already made a change in #6407 to make pitr_interval authoritative
for synthetic size calculations (do not charge users for data retained
due to gc_horizon), but that change didn't cover the case where someone
entirely disables time-based retention by setting pitr_interval=0
Relates to: https://github.com/neondatabase/neon/issues/6374
## Summary of changes
When pitr_interval is zero, do not set `pitr_cutoff` based on
gc_horizon.
gc_horizon is still enforced, but separately (its value is passed
separately, there was never a need to claim pitr_cutoff to gc_horizon)
## More detail
### Issue 1
Before this PR, we would skip the update_gc_info for timelines with
last_record_lsn() < gc_horizon.
Let's call such timelines "tiny".
The rationale for that presumably was that we can't GC anything in the
tiny timelines, why bother to call update_gc_info().
However, synthetic size calculation relies on up-to-date
update_gc_info() data.
Before this PR, tiny timelines would never get an updated
GcInfo::pitr_horizon (it remained Lsn(0)).
Even on projects with pitr_interval=0d.
With this PR, update_gc_info is always called, hence
GcInfo::pitr_horizon is always updated, thereby
providing synthetic size calculation with up-to-data data.
### Issue 2
Before this PR, regardless of whether the timeline is "tiny" or not,
GcInfo::pitr_horizon was clamped to at least last_record_lsn -
gc_horizon, even if the pitr window in terms of LSN range was shorter
(=less than) the gc_horizon.
With this PR, that clamping is removed, so, for pitr_interval=0, the
pitr_horizon = last_record_lsn.
## Problem
Split off from https://github.com/neondatabase/neon/pull/7399, which is
the first piece of code that does a WithDelimiter object listing using a
prefix that isn't a full directory name.
## Summary of changes
- Revise list function to not append a `/` to the prefix -- prefixes
don't have to end with a slash.
- Fix local_fs implementation of list to not assume that WithDelimiter
case will always use a directory as a prerfix.
- Remove `list_files`, `list_prefixes` wrappers, as they add little
value and obscure the underlying list function -- we need callers to
understand the semantics of what they're really calling (listobjectsv2)
Extracted from https://github.com/neondatabase/neon/pull/7375. We assume
everything >= 0x80 are metadata keys. AUX file keys are part of the
metadata keys, and we use `0x90` as the prefix for AUX file keys.
The AUX file encoding is described in the code comment. We use xxhash128
as the hash algorithm. It seems to be portable according to the
introduction,
> xxHash is an Extremely fast Hash algorithm, processing at RAM speed
limits. Code is highly portable, and produces hashes identical across
all platforms (little / big endian).
...though whether the Rust version follows the same convention is
unknown and might need manual review of the library. Anyways, we can
always change the hash algorithm before rolling it out in
staging/end-user, and I made a quick decision to use xxhash here because
it generates 128b hash + portable. We can save the discussion of which
hash algorithm to use later.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
In some dev/test environments, there aren't health checks to guarantee
the database is available before starting the controller. This creates
friction for the developer.
## Summary of changes
- Wait up to 5 seconds for the database to become available on startup
## Problem
We recently went through an incident where compaction was inhibited by a
bug. We didn't observe this until quite late because we did not have alerting
on deep reads.
## Summary of changes
+ Tweak an existing metric that tracks the depth of a read on the
non-vectored read path:
* Give it a better name
* Track all layers
* Larger buckets
+ Add a similar metric for the vectored read path
+ Add a compaction smoke test which uses these metrics. This test would
have caught
the compaction issue mentioned earlier.
Related https://github.com/neondatabase/neon/issues/7428
## Problem
Vectored get would descend into ancestor timelines for aux files.
This is not the behaviour of the legacy read path and blocks cutting
over to the vectored read path.
Fixes https://github.com/neondatabase/neon/issues/7379
## Summary of Changes
Treat non inherited keys specially in vectored get. At the point when
we want to descend into the ancestor mark all pending non inherited keys
as errored out at the key level. Note that this diverges from the
standard vectored get behaviour for missing keys which is a top level
error. This divergence is required to avoid blocking compaction in case
such an error is encountered when compaction aux files keys. I'm pretty
sure the bug I just described predates the vectored get implementation,
but it's still worth fixing.
## Problem
The `export_import_between_pageservers` script us to do major storage format changes
in the past. If we have to do such breaking changes in the future this approach
wouldn't be suitable because:
1. It doesn't scale to the current size of the fleet
2. It loses history
## Summary of changes
Remove the script and its associated test.
Keep `fullbasebackup` and friends because it's useful for debugging.
Closes https://github.com/neondatabase/cloud/issues/11648
Instead of trusting that a request with latest == true means that the
request LSN was at least last_record_lsn, remember explicitly when the
relation cache was initialized.
Incidentally, this allows updating the relation size cache also on reads
from read-only endpoints, when the endpoint is at a relatively recent
LSN (more recent than the end of the timeline when the timeline was
loaded in the pageserver).
Add a comment to wait_or_get_last_lsn() that it might be better to use
an older LSN when possible. Note that doing that would be unsafe,
without the relation cache changes in this commit!
As part of https://github.com/neondatabase/neon/pull/7375 and to improve
the current vectored get implementation, we separate the missing key
error out. This also saves us several Box allocations in the get page
implementation.
## Summary of changes
* Create a caching field of layer traversal id for each of the layer.
* Remove box allocations for layer traversal id retrieval and implement
MissingKey error message as before. This should be a little bit faster.
* Do not format error message until `Display`.
* For in-mem layer, the descriptor is different before/after frozen. I'm
using once lock for that.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
- #7451
INIT_FORKNUM blocks must be stored on shard 0 to enable including them
in basebackup.
This issue can be missed in simple tests because creating an unlogged
table isn't sufficient -- to repro I had to create an _index_ on an
unlogged table (then restart the endpoint).
Closes: #7451
## Summary of changes
- Add a reproducer for the issue.
- Tweak the condition for `key_is_shard0` to include anything that isn't
a normal relation block _and_ any normal relation block whose forknum is
INIT_FORKNUM.
- To enable existing databases to recover from the issue, add a special
case that omits relations if they were stored on the wrong INITFORK.
This enables postgres to start and the user to drop the table and
recreate it.
## Problem
Currently we cannot configure retries, also, we don't really have
visibility of what's going on there.
## Summary of changes
* Added cli params
* Improved logging
* Decrease the number of retries: it feels like most of retries doesn't
help. Once there would be better errors handling, we can increase it
back.
Before, we asserted that a layer would only be loaded by the timeline
that initially created it. Now, with the ancestor detach, we will want
to utilize remote copy as much as possible, so we will need to open
other timeline layers as our own.
Cc: #6994
Currently, any `Timeline::schedule_uploads` will generate a fresh
`TimelineMetadata` instead of updating the values, which it means to
update. This makes it impossible for #6994 to work while `Timeline`
receives layer flushes by overwriting any configured new
`ancestor_timeline_id` and possible `ancestor_lsn`.
The solution is to only make full `TimelineMetadata` "updates" from one
place: branching. At runtime, update only the three fields, same as
before in `Timeline::schedule_updates`.
There were two issues with the test at page boundaries:
1. If the first logical message with 10 bytes payload crossed a page
boundary, the calculated 'base_size' was too large because it included
the page header.
2. If it was inserted near the end of a page so that there was not
enough room for another one, we did "remaining_lsn += XLOG_BLCKSZ" but
that didn't take into account the page headers either.
As a result, the test would fail if the WAL insert position at the
beginning of the test was too close to the end of a WAL page. Fix the
calculations by repeating the 10-byte logical message if the starting
position is not suitable.
I bumped into this with PR #7377; it changed the arguments of a few SQL
functions in neon_test_utils extension, which changed the WAL positions
slightly, and caused a test failure.
This is similar to https://github.com/neondatabase/neon/pull/7436, but
for different test.
As noted in the comment, the craft_internal() function fails if the
inserted WAL happens to land at page boundary. I bumped into that with
PR #7377; it changed the arguments of a few SQL functions in
neon_test_utils extension, which changed the WAL positions slightly, and
caused a test failure.
## Problem
`cargo deny check` is complaining about our rustls versions, causing
CI to fail:
```
error[vulnerability]: `rustls::ConnectionCommon::complete_io` could fall into an infinite loop based on network input
┌─ /__w/neon/neon/Cargo.lock:395:1
│
395 │ rustls 0.21.9 registry+https://github.com/rust-lang/crates.io-index
│ ------------------------------------------------------------------- security vulnerability detected
│
= ID: RUSTSEC-2024-0336
= Advisory: https://rustsec.org/advisories/RUSTSEC-2024-0336
= If a `close_notify` alert is received during a handshake, `complete_io`
does not terminate.
Callers which do not call `complete_io` are not affected.
`rustls-tokio` and `rustls-ffi` do not call `complete_io`
and are not affected.
`rustls::Stream` and `rustls::StreamOwned` types use
`complete_io` and are affected.
= Announcement: https://github.com/rustls/rustls/security/advisories/GHSA-6g7w-8wpp-frhj
= Solution: Upgrade to >=0.23.5 OR >=0.22.4, <0.23.0 OR >=0.21.11, <0.22.0 (try `cargo update -p rustls`)
error[vulnerability]: `rustls::ConnectionCommon::complete_io` could fall into an infinite loop based on network input
┌─ /__w/neon/neon/Cargo.lock:396:1
│
396 │ rustls 0.22.2 registry+https://github.com/rust-lang/crates.io-index
│ ------------------------------------------------------------------- security vulnerability detected
│
= ID: RUSTSEC-2024-0336
= Advisory: https://rustsec.org/advisories/RUSTSEC-2024-0336
= If a `close_notify` alert is received during a handshake, `complete_io`
does not terminate.
Callers which do not call `complete_io` are not affected.
`rustls-tokio` and `rustls-ffi` do not call `complete_io`
and are not affected.
`rustls::Stream` and `rustls::StreamOwned` types use
`complete_io` and are affected.
= Announcement: https://github.com/rustls/rustls/security/advisories/GHSA-6g7w-8wpp-frhj
= Solution: Upgrade to >=0.23.5 OR >=0.22.4, <0.23.0 OR >=0.21.11, <0.22.0 (try `cargo update -p rustls`)
```
## Summary of changes
`cargo update -p rustls@0.21.9 -p rustls@0.22.2`
- Cleanup part for `docker/setup-buildx-action` started to fail with the following error (for no obvious reason):
```
/nvme/actions-runner/_work/_actions/docker/setup-buildx-action/v3/webpack:/docker-setup-buildx/node_modules/@actions/cache/lib/cache.js:175
throw new Error(`Path Validation Error: Path(s) specified in the action for caching do(es) not exist, hence no cache is being saved.`);
^
Error: Path Validation Error: Path(s) specified in the action for caching do(es) not exist, hence no cache is being saved.
at Object.rejected (/nvme/actions-runner/_work/_actions/docker/setup-buildx-action/v3/webpack:/docker-setup-buildx/node_modules/@actions/cache/lib/cache.js:175:1)
at Generator.next (<anonymous>)
at fulfilled (/nvme/actions-runner/_work/_actions/docker/setup-buildx-action/v3/webpack:/docker-setup-buildx/node_modules/@actions/cache/lib/cache.js:29:1)
```
- Downgrade `docker/setup-buildx-action` from v3 to v2
## Problem
When we migrate a large existing tenant, we would like to be able to
ensure it has pre-loaded layers onto a pageserver managed by the storage
controller.
## Summary of changes
- Add `storcon_cli tenant-warmup`, which configures the tenant into
PlacementPolicy::Secondary (unless it's already attached), and then
polls the secondary download API reporting progress.
- Extend a test case to check that when onboarding with a secondary
location pre-created, we properly use that location for our first
attachment.
## Problem
Vectored read path may return an image that's newer than the request lsn
under certain circumstances.
```
LSN
^
|
|
500 | ------------------------- -> branch point
400 | X
300 | X
200 | ------------------------------------> requested lsn
100 | X
|---------------------------------> Key
Legend:
* X - page images
```
The vectored read path inspects each ancestor timeline one by one
starting from the current one.
When moving into the ancestor timeline, the current code resets the
current search lsn (called `cont_lsn` in code)
to the lsn of the ancestor timeline
([here](d5708e7435/pageserver/src/tenant/timeline.rs (L2971))).
For instance, if the request lsn was 200, we would:
1. Look into the current timeline and find nothing for the key
2. Descend into the ancestor timeline and set `cont_lsn=500`
3. Return the page image at LSN 400
Myself and Christian find it very unlikely for this to have happened in
prod since the vectored read path
is always used at the last record lsn.
This issue was found by a regress test during the work to migrate get
page handling to use the vectored
implementation. I've applied my fix to that wip branch and it fixed the
issue.
## Summary of changes
The fix is to set the current search lsn to the min between the
requested LSN and the ancestor lsn.
Hence, at step 2 above we would set the current search lsn to 200 and
ignore the images above that.
A test illustrating the bug is also included. Fails without the patch
and passes with it.
#7030 introduced an annoying papercut, deeming a failure to acquire a
strong reference to `LayerInner` from `DownloadedLayer::drop` as a
canceled eviction. Most of the time, it wasn't that, but just timeline
deletion or tenant detach with the layer not wanting to be deleted or
evicted.
When a Layer is dropped as part of a normal shutdown, the `Layer` is
dropped first, and the `DownloadedLayer` the second. Because of this, we
cannot detect eviction being canceled from the `DownloadedLayer::drop`.
We can detect it from `LayerInner::drop`, which this PR adds.
Test case is added which before had 1 started eviction, 2 canceled. Now
it accurately finds 1 started, 1 canceled.
## Problem
External contributors need information on how to use the storage
controller.
## Summary of changes
- Background content on what the storage controller is.
- Deployment information on how to use it.
This is not super-detailed, but should be enough for a well motivated
third party to get started, with an occasional peek at the code.
## Problem
Some tenants were observed to stop doing downloads after some time
## Summary of changes
- Fix a rogue `<` that was incorrectly scheduling work when `now` was
_before_ the scheduling target, rather than after. This usually resulted
in too-frequent execution, but could also result in never executing, if
the current time has advanced ahead of `next_download` at the time we
call `schedule()`.
- Fix in-memory list of timelines not being amended after timeline
deletion: the resulted in repeated harmless logs about the timeline
being removed, and redundant calls to remove_dir_all for the timeline
path.
- Add a log at startup to make it easier to see a particular tenant
starting in secondary mode (this is for parity with the logging that
exists when spawning an attached tenant). Previously searching on tenant
ID didn't provide a clear signal as to how the tenant was started during
pageserver start.
- Add a test that exercises secondary downloads using the background
scheduling, whereas existing tests were using the API hook to invoke
download directly.
For "timeline ancestor merge" or "timeline detach," we need to "cut"
delta layers at particular LSN. The name "truncate" is not used as it
would imply that a layer file changes, instead of what happens: we copy
keys with Lsn less than a "cut point".
Cc: #6994
Add the "copy delta layer prefix" operation to DeltaLayerInner, re-using
some of the vectored read internals. The code is `cfg(test)` until it
will be used later with a more complete integration test.
## Problem
Sometimes rejected metric might record invalid events.
## Summary of changes
* Only record it `rejected` was explicitly set.
* Change order in logs.
* Report metrics if not under high-load.
## Problem
This trace is emitted whenever a vectored read touches the end of a
delta layer file. It's a perfectly normal case, but I expected it to be
more rare when implementing the code.
## Summary of changes
Demote log to debug.
## Problem
When calling `./neon_local timeline` a confusing error message pops up:
`command failed: no tenant subcommand provided`
## Summary of changes
Add `add-help-for-timeline-arg` for timeline commands so when no
argument for the timeline is provided help is printed.
## Problem
There is an unused dead code.
## Summary of changes
Let's remove it. In case we would need it in the future, we can always
return it back.
Also removed cli arguments. They shouldn't be used by anyone but us.
## Problem
For PRs, by default, we check out a phantom merge commit (merge a branch
into the main), but using a real branches head when finding `build-tools`
image tag.
## Summary of changes
- Change `COMMIT_SHA` to use `${{ github.sha }}` instead of `${{
github.event.pull_request.head.sha }}` for PRs
## Checklist before requesting a review
- [x] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
Leftover from my LFC benchmarks. Safekeepers only listen on `127.0.0.1`
for `neon_local`. This pull request adds support for listening on other
address. To specify a custom address, modify `.neon/config`.
```
[[safekeepers]]
listen_addr = "192.168.?.?"
```
Endpoints created by neon_local still use 127.0.0.1 and I will fix them
later. I didn't fix it in the same pull request because my benchmark
setting does not use neon_local to create compute nodes so I don't know
how to fix it yet -- maybe replacing a few `127.0.0.1`s.
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
We specify a bunch of possible error codes in the pageserver api swagger
spec. This is error prone and annoying to work with.
https://github.com/neondatabase/cloud/pull/11907 introduced generic
error handling on the control plane side, so we can now clean up the
spec.
## Summary of changes
* Remove generic error codes from swagger spec
* Update a couple route handlers which would previously return an error
without a `msg` field in the response body.
Tested via https://github.com/neondatabase/cloud/pull/12340
Related https://github.com/neondatabase/cloud/issues/7238
## Problem
Many users have access to ipv6 subnets (eg a /64). That gives them 2^64
addresses to play with
## Summary of changes
Truncate the address to /64 to reduce the attack surface.
Todo:
~~Will NAT64 be an issue here? AFAIU they put the IPv4 address at the
end of the IPv6 address. By truncating we will lose all that detail.~~
It's the same problem as a host sharing IPv6 addresses between clients.
I don't think it's up to us to solve. If a customer is getting DDoSed,
then they likely need to arrange a dedicated IP with us.
## Problem
- https://github.com/neondatabase/neon/issues/7355
The optimize_secondary function calls schedule_shard to check for
improvements, but if there are exactly the same number of nodes as there
are replicas of the shard, it emits some scary looking logs about no
nodes being elegible.
Closes https://github.com/neondatabase/neon/issues/7355
## Summary of changes
- Add a mode to SchedulingContext that controls logging: this should be
useful in future any time we add a log to the scheduling path, to avoid
it becoming a source of spam when the scheduler is called during
optimization.
## Problem
test_sharding_smoke recently got an added section that checks deletion
of a sharded tenant. The storage controller does a retry loop for
deletion, waiting for a 404 response. When deletion is a bit slow (debug
builds), the retry of deletion was getting a 500 response -- this caused
the test to become flaky (example failure:
https://neon-github-public-dev.s3.amazonaws.com/reports/release-proxy/8659801445/index.html#testresult/b4cbf5b58190f60e/retries)
There was a false comment in the code:
```
match tenant.current_state() {
TenantState::Broken { .. } | TenantState::Stopping { .. } => {
- // If a tenant is broken or stopping, DeleteTenantFlow can
- // handle it: broken tenants proceed to delete, stopping tenants
- // are checked for deletion already in progress.
```
If the tenant is stopping, DeleteTenantFlow does not in fact handle it,
but returns a 500-yielding errror.
## Summary of changes
Before calling into DeleteTenantFlow, if the tenant is in
stopping|broken state then return 202 if a deletion is in progress. This
makes the API friendlier for retries.
The historic AlreadyInProgress (409) response still exists for if we
enter DeleteTenantFlow and unexpectedly see the tenant stopping. That
should go away when we implement #5080 . For the moment, callers that
handle 409s should continue to do so.
Before this PR, the `nix::poll::poll` call would stall the executor.
This PR refactors the `walredo::process` module to allow for different
implementations, and adds a new `async` implementation which uses
`tokio::process::ChildStd{in,out}` for IPC.
The `sync` variant remains the default for now; we'll do more testing in
staging and gradual rollout to prod using the config variable.
Performance
-----------
I updated `bench_walredo.rs`, demonstrating that a single `async`-based
walredo manager used by N=1...128 tokio tasks has lower latency and
higher throughput.
I further did manual less-micro-benchmarking in the real pageserver
binary.
Methodology & results are published here:
https://neondatabase.notion.site/2024-04-08-async-walredo-benchmarking-8c0ed3cc8d364a44937c4cb50b6d7019?pvs=4
tl;dr:
- use pagebench against a pageserver patched to answer getpage request &
small-enough working set to fit into PS PageCache / kernel page cache.
- compare knee in the latency/throughput curve
- N tenants, each 1 pagebench clients
- sync better throughput at N < 30, async better at higher N
- async generally noticable but not much worse p99.X tail latencies
- eyeballing CPU efficiency in htop, `async` seems significantly more
CPU efficient at ca N=[0.5*ncpus, 1.5*ncpus], worse than `sync` outside
of that band
Mental Model For Walredo & Scheduler Interactions
-------------------------------------------------
Walredo is CPU-/DRAM-only work.
This means that as soon as the Pageserver writes to the pipe, the
walredo process becomes runnable.
To the Linux kernel scheduler, the `$ncpus` executor threads and the
walredo process thread are just `struct task_struct`, and it will divide
CPU time fairly among them.
In `sync` mode, there are always `$ncpus` runnable `struct task_struct`
because the executor thread blocks while `walredo` runs, and the
executor thread becomes runnable when the `walredo` process is done
handling the request.
In `async` mode, the executor threads remain runnable unless there are
no more runnable tokio tasks, which is unlikely in a production
pageserver.
The above means that in `sync` mode, there is an implicit concurrency
limit on concurrent walredo requests (`$num_runtimes *
$num_executor_threads_per_runtime`).
And executor threads do not compete in the Linux kernel scheduler for
CPU time, due to the blocked-runnable-ping-pong.
In `async` mode, there is no concurrency limit, and the walredo tasks
compete with the executor threads for CPU time in the kernel scheduler.
If we're not CPU-bound, `async` has a pipelining and hence throughput
advantage over `sync` because one executor thread can continue
processing requests while a walredo request is in flight.
If we're CPU-bound, under a fair CPU scheduler, the *fixed* number of
executor threads has to share CPU time with the aggregate of walredo
processes.
It's trivial to reason about this in `sync` mode due to the
blocked-runnable-ping-pong.
In `async` mode, at 100% CPU, the system arrives at some (potentially
sub-optiomal) equilibrium where the executor threads get just enough CPU
time to fill up the remaining CPU time with runnable walredo process.
Why `async` mode Doesn't Limit Walredo Concurrency
--------------------------------------------------
To control that equilibrium in `async` mode, one may add a tokio
semaphore to limit the number of in-flight walredo requests.
However, the placement of such a semaphore is non-trivial because it
means that tasks queuing up behind it hold on to their request-scoped
allocations.
In the case of walredo, that might be the entire reconstruct data.
We don't limit the number of total inflight Timeline::get (we only
throttle admission).
So, that queue might lead to an OOM.
The alternative is to acquire the semaphore permit *before* collecting
reconstruct data.
However, what if we need to on-demand download?
A combination of semaphores might help: one for reconstruct data, one
for walredo.
The reconstruct data semaphore permit is dropped after acquiring the
walredo semaphore permit.
This scheme effectively enables both a limit on in-flight reconstruct
data and walredo concurrency.
However, sizing the amount of permits for the semaphores is tricky:
- Reconstruct data retrieval is a mix of disk IO and CPU work.
- If we need to do on-demand downloads, it's network IO + disk IO + CPU
work.
- At this time, we have no good data on how the wall clock time is
distributed.
It turns out that, in my benchmarking, the system worked fine without a
semaphore. So, we're shipping async walredo without one for now.
Future Work
-----------
We will do more testing of `async` mode and gradual rollout to prod
using the config flag.
Once that is done, we'll remove `sync` mode to avoid the temporary code
duplication introduced by this PR.
The flag will be removed.
The `wait()` for the child process to exit is still synchronous; the
comment [here](
655d3b6468/pageserver/src/walredo.rs (L294-L306))
is still a valid argument in favor of that.
The `sync` mode had another implicit advantage: from tokio's
perspective, the calling task was using up coop budget.
But with `async` mode, that's no longer the case -- to tokio, the writes
to the child process pipe look like IO.
We could/should inform tokio about the CPU time budget consumed by the
task to achieve fairness similar to `sync`.
However, the [runtime function for this is
`tokio_unstable`](`https://docs.rs/tokio/latest/tokio/task/fn.consume_budget.html).
Refs
----
refs #6628
refs https://github.com/neondatabase/neon/issues/2975
No functional changes, this is a comments/naming PR.
While merging sharding changes, some cleanup of the shard.rs types was
deferred.
In this PR:
- Rename `is_zero` to `is_shard_zero` to make clear that this method
doesn't literally mean that the entire object is zeros, just that it
refers to the 0th shard in a tenant.
- Pull definitions of types to the top of shard.rs and add a big comment
giving an overview of which type is for what.
Closes: https://github.com/neondatabase/neon/issues/6072
## Problem
`create-test-report` job takes more than 8 minutes, the longest step is
uploading Allure report to S3:
Before:
```
+ aws s3 cp --recursive --only-show-errors /tmp/pr-7362-1712847045/report s3://neon-github-public-dev/reports/pr-7362/8647730612
real 6m10.572s
user 6m37.717s
sys 1m9.429s
```
After:
```
+ s5cmd --log error cp '/tmp/pr-7362-1712858221/report/*' s3://neon-github-public-dev/reports/pr-7362/8650636861/
real 0m9.698s
user 1m9.438s
sys 0m6.419s
```
## Summary of changes
- Add `s5cmd`(https://github.com/peak/s5cmd) to build-tools image
- Use `s5cmd` instead of `aws s3` for uploading Allure reports
## Problem
possible for the database connections to not close in time.
## Summary of changes
force the closing of connections if the client has hung up
## Problem
Actually read redis events.
## Summary of changes
This is revert of https://github.com/neondatabase/neon/pull/7350 +
fixes.
* Fixed events parsing
* Added timeout after connection failure
* Separated regional and global redis clients.
## Problem
My benchmarks show that prometheus is not very good.
https://github.com/conradludgate/measured
We're already using it in storage_controller and it seems to be working
well.
## Summary of changes
Replace prometheus with my new measured crate in proxy only.
Apologies for the large diff. I tried to keep it as minimal as I could.
The label types add a bit of boiler plate (but reduce the chance we
mistype the labels), and some of our custom metrics like CounterPair and
HLL needed to be rewritten.
## Problem
`build-build-tools-image` workflow is designed to be run only in one
example per the whole repository. Currently, the job gets cancelled if a
newer one is scheduled, here's an example:
https://github.com/neondatabase/neon/actions/runs/8419610607
## Summary of changes
- Explicitly set `cancel-in-progress: false` for all jobs that aren't
supposed to be cancelled
## Problem
We are seeing some mysterious long waits when sending requests.
## Summary of changes
- To eliminate risk that we are incurring some unreasonable overheads
from setup, e.g. DNS, use a single Client (internally a pool) instead of
repeatedly constructing a fresh one.
- To make it clearer where a timeout is occurring, apply a 10 second
timeout to requests as we send them.
## Problem
See https://neondb.slack.com/archives/C03QLRH7PPD/p1712529369520409
In case of statements CREATE TABLE AS SELECT... or INSERT FROM SELECT...
we are fetching data from source table and storing it in destination
table. It cause problems with prefetch last-written-lsn is known for the
pages of source table
(which for example happens after compute restart). In this case we get
get global value of last-written-lsn which is changed frequently as far
as we are writing pages of destination table. As a result request-isn
for the prefetch and request-let when this page is actually needed are
different and we got exported prefetch request. So it actually disarms
prefetch.
## Summary of changes
Proposed simple patch stores last-written LSN for the page when it is
not found. So next time we will request last-written LSN for this page,
we will get the same value (certainly if the page was not changed).
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
Part of neondatabase/cloud#12047.
The basic idea is that for our VMs, we want to enable swap and disable
Linux memory overcommit. Alongside these, we should set postgres'
dynamic_shared_memory_type to mmap, but we want to avoid setting it to
mmap if swap is not enabled.
Implementing this in the control plane would be fiddly, but it's
relatively straightforward to add to compute_ctl.
## Problem
hyper1 offers control over the HTTP connection that hyper0_14 does not.
We're blocked on switching all services to hyper1 because of how we use
tonic, but no reason we can't switch proxy over.
## Summary of changes
1. hyper0.14 -> hyper1
1. self managed server
2. Remove the `WithConnectionGuard` wrapper from `protocol2`
2. Remove TLS listener as it's no longer necessary
3. include first session ID in connection startup logs
Adds another tool to the DR toolbox: ability in pagectl to
recover arbitrary prefixes in remote storage. Requires remote storage config,
the prefix, and the travel-to timestamp parameter
to be specified as cli args.
The done-if-after parameter is also supported.
Example invocation (after `aws login --profile dev`):
```
RUST_LOG=remote_storage=debug AWS_PROFILE=dev cargo run -p pagectl time-travel-remote-prefix 'remote_storage = { bucket_name = "neon-test-bucket-name", bucket_region = "us-east-2" }' wal/3aa8fcc61f6d357410b7de754b1d9001/641e5342083b2235ee3deb8066819683/ 2024-04-05T17:00:00Z
```
This has been written to resolve a customer recovery case:
https://neondb.slack.com/archives/C033RQ5SPDH/p1712256888468009
There is validation of the prefix to prevent accidentially specifying
too generic prefixes, which can cause corruption and data
loss if used wrongly. Still, the validation is not perfect and it is
important that the command is used with caution.
If possible, `time_travel_remote_storage` should
be used instead which has additional checks in place.
Problem
Currently, we base our time based layer rolling decision on the last
time we froze a layer. This means that if we roll a layer and then go
idle for longer than the checkpoint timeout the next layer will be
rolled after the first write. This is of course not desirable.
Summary of changes
Record the timepoint of the first write to an open layer and use that
for time based layer rolling decisions. Note that I had to keep
`Timeline::last_freeze_ts` for the sharded tenant disk consistent lsn
skip hack.
Fixes#7241
## Problem
Proxy doesn't know about existing endpoints.
## Summary of changes
* Added caching of all available endpoints.
* On the high load, use it before going to cplane.
* Report metrics for the outcome.
* For rate limiter and credentials caching don't distinguish between
`-pooled` and not
TODOs:
* Make metrics more meaningful
* Consider integrating it with the endpoint rate limiter
* Test it together with cplane in preview
## Problem
```
Could not resolve host: console.stage.neon.tech
```
## Summary of changes
- replace `console.stage.neon.tech` with `console-stage.neon.build`
## Problem
After switching the default pageserver io-engine to `tokio-epoll-uring`
on CI, we tuned a query that finds flaky tests (in
https://github.com/neondatabase/neon/pull/7077).
It has been almost a month since then, additional query tuning is not
required anymore.
## Summary of changes
- Remove extra condition from flaky tests query
- Also return back parameterisation to the query
## Problem
Some awkwardness in the measured API.
Missing process metrics.
## Summary of changes
Update measured to use the new convenience setup features.
Added measured-process lib.
Added measured support for libmetrics
## Problem
We have two places that use a helper (`ser_rfc3339_millis`) to get serde
to stringify SystemTimes into the desired format.
## Summary of changes
Created a new module `utils::serde_system_time` and inside it a wrapper
type `SystemTime` for `std::time::SystemTime` that
serializes/deserializes to the RFC3339 format.
This new type is then used in the two places that were previously using
the helper for serialization, thereby eliminating the need to decorate
structs.
Closes#7151.
This PR is an off-by-default revision v2 of the (since-reverted) PR
#6555 / commit `3220f830b7fbb785d6db8a93775f46314f10a99b`.
See that PR for details on why running with a single runtime is
desirable and why we should be ready.
We reverted #6555 because it showed regressions in prodlike cloudbench,
see the revert commit message `ad072de4209193fd21314cf7f03f14df4fa55eb1`
for more context.
This PR makes it an opt-in choice via an env var.
The default is to use the 4 separate runtimes that we have today, there
shouldn't be any performance change.
I tested manually that the env var & added metric works.
```
# undefined env var => no change to before this PR, uses 4 runtimes
./target/debug/neon_local start
# defining the env var enables one-runtime mode, value defines that one runtime's configuration
NEON_PAGESERVER_USE_ONE_RUNTIME=current_thread ./target/debug/neon_local start
NEON_PAGESERVER_USE_ONE_RUNTIME=multi_thread:1 ./target/debug/neon_local start
NEON_PAGESERVER_USE_ONE_RUNTIME=multi_thread:2 ./target/debug/neon_local start
NEON_PAGESERVER_USE_ONE_RUNTIME=multi_thread:default ./target/debug/neon_local start
```
I want to use this change to do more manualy testing and potentially
testing in staging.
Future Work
-----------
Testing / deployment ergonomics would be better if this were a variable
in `pageserver.toml`.
It can be done, but, I don't need it right now, so let's stick with the
env var.
It's just unnecessary to use spawn_blocking there, and with
https://github.com/neondatabase/neon/pull/7331 , it will result in
really just one executor thread when enabling one-runtime with
current_thread executor.
We can currently underflow `pageserver_resident_physical_size_global`,
so the used disk bytes would show `u63::MAX` by mistake. The assumption
of the API (and the documented behavior) was to give the layer files
disk usage.
Switch to reporting numbers that match `df` output.
Fixes: #7336
This is the other main failure mode called out in #6092 , that the test
can shut down the pageserver while it has "future layers" in the index,
and that this results in unexpected stats after restart.
We can avoid this nondeterminism by shutting down the endpoint, flushing
everything from SK to PS, checkpointing, and then waiting for that final
LSN to be uploaded. This is more heavyweight than most of our tests
require, but useful in the case of tests that expect a particular
behavior after restart wrt layer deletions.
## Problem
- Previously, an async mutex was held for the duration of
`ComputeHook::notify`. This served multiple purposes:
- Ensure updates to a given tenant are sent in the proper order
- Prevent concurrent calls into neon_local endpoint updates in test
environments (neon_local is not safe to call concurrently)
- Protect the inner ComputeHook::state hashmap that is used to calculate
when to send notifications.
This worked, but had the major downside that while we're waiting for a
compute hook request to the control plane to succeed, we can't notify
about any other tenants. Notifications block progress of live
migrations, so this is a problem.
## Summary of changes
- Protect `ComputeHook::state` with a sync lock instead of an async lock
- Use a separate async lock ( `ComputeHook::neon_local_lock` ) for
preventing concurrent calls into neon_local, and only take this in the
neon_local code path.
- Add per-tenant async locks in ShardedComputeHookTenant, and use these
to ensure that only one remote notification can be sent at once per
tenant. If several shards update concurrently, their updates will be
coalesced.
- Add an explicit semaphore that limits concurrency of calls into the
cloud control plane.
## Problem
Ingest filtering wasn't being applied to timeline creations, so a
timeline created on a sharded tenant would use 20MB+ on each shard (each
shard got a full copy). This didn't break anything, but is inefficient
and leaves the system in a harder-to-validate state where shards
initially have some data that they will eventually drop during
compaction.
Closes: https://github.com/neondatabase/neon/issues/6649
## Summary of changes
- in `import_rel`, filter block-by-block with is_key_local
- During test_sharding_smoke, check that per-shard physical sizes are as
expected
- Also extend the test to check deletion works as expected (this was an
outstanding tech debt task)
## Problem
As https://github.com/neondatabase/neon/issues/6092 points out, this
test was (ab)using a failpoint!() with 'pause', which was occasionally
causing index uploads to get hung on a stuck executor thread, resulting
in timeouts waiting for remote_consistent_lsn.
That is one of several failure modes, but by far the most frequent.
## Summary of changes
- Replace the failpoint! with a `sleep_millis_async`, which is not only
async but also supports clean shutdown.
- Improve debugging: log the consistent LSN when scheduling an index
upload
- Tidy: remove an unnecessary checkpoint in the test code, where
last_flush_lsn_upload had just been called (this does a checkpoint
internally)
The binary etc were renamed some time ago, but the path in the source
tree remained "attachment_service" to avoid disruption to ongoing PRs.
There aren't any big PRs out right now, so it's a good time to cut over.
- Rename `attachment_service` to `storage_controller`
- Move it to the top level for symmetry with `storage_broker` & to avoid
mixing the non-prod neon_local stuff (`control_plane/`) with the storage
controller which is a production component.
## Problem
Would be nice to have a bit more info on cold start metrics.
## Summary of changes
* Change connect compute latency to include `cold_start_info`.
* Update `ColdStartInfo` to include HttpPoolHit and WarmCached.
* Several changes to make more use of interned strings
Updates the `test-context` dev-dependency of the `remote_storage` crate
to 0.3. This removes a lot of `async_trait` instances.
Related earlier work: #6305, #6464
Found these logs on staging safekeepers:
```
INFO Partial backup{ttid=X/Y}: failed to upload 000000010000000000000000_173_0000000000000000_0000000000000000_sk56.partial: Failed to open file "/storage/safekeeper/data/X/Y/000000010000000000000000.partial" for wal backup: No such file or directory (os error 2)
INFO Partial backup{ttid=X/Y}:upload{name=000000010000000000000000_173_0000000000000000_0000000000000000_sk56.partial}: starting upload PartialRemoteSegment { status: InProgress, name: "000000010000000000000000_173_0000000000000000_0000000000000000_sk56.partial", commit_lsn: 0/0, flush_lsn: 0/0, term: 173 }
```
This is because partial backup tries to upload zero segment when there
is no data in timeline. This PR fixes this bug introduced in #6530.
This test was occasionally flaky: it already allowed the log for the
scheduler complaining about Stop state, but not the log for
maybe_reconcile complaining.
## Problem
When a location_conf request was repeated with no changes, we failed to
build the list of shards in the result.
## Summary of changes
Remove conditional that only generated a list of updates if something
had really changed. This does some redundant database updates, but it is
preferable to having a whole separate code path for no-op changes.
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
The API client was written around the same time as some of the server
APIs changed from TenantId to TenantShardId
Closes: https://github.com/neondatabase/neon/issues/6154
## Summary of changes
- Refactor mgmt_api timeline_info and keyspace methods to use
TenantShardId to match the server
This doesn't make pagebench sharding aware, but it paves the way to do
so later.
## Problem
In the test for https://github.com/neondatabase/neon/pull/6776, a test
cases uses tiny layer sizes and tiny stripe sizes. This hits a scenario
where a shard's checkpoint interval spans a region where none of the
content in the WAL is ingested by this shard. Since there is no layer to
flush, we do not advance disk_consistent_lsn, and this causes the test
to fail while waiting for LSN to advance.
## Summary of changes
- Pass an LSN through `layer_flush_start_tx`. This is the LSN to which
we have frozen at the time we ask the flush to flush layers frozen up to
this point.
- In the layer flush task, if the layers we flush do not reach
`frozen_to_lsn`, then advance disk_consistent_lsn up to this point.
- In `maybe_freeze_ephemeral_layer`, handle the case where
last_record_lsn has advanced without writing a layer file: this ensures
that disk_consistent_lsn and remote_consistent_lsn advance anyway.
The net effect is that the disk_consistent_lsn is allowed to advance
past regions in the WAL where a shard ingests no data, and that we
uphold our guarantee that remote_consistent_lsn always eventually
reaches the tip of the WAL.
The case of no layer at all is hard to test at present due to >0 shards
being polluted with SLRU writes, but I have tested it locally with a
branch that disables SLRU writes on shards >0. We can tighten up the
testing on this in future as/when we refine shard filtering (currently
shards >0 need the SLRU because they use it to figure out cutoff in GC
using timestamp-to-lsn).
Some time ago, we had an issue where a deletion queue hang was also
causing timeline deletions to hang.
This was unnecessary because the timeline deletion doesn't _need_ to
flush the deletion queue, it just does it as a pleasantry to make the
behavior easier to understand and test.
In this PR, we wrap the flush calls in a 10 second timeout (typically
the flush takes milliseconds) so that in the event of issues with the
deletion queue, timeline deletions are slower but not entirely blocked.
Closes: https://github.com/neondatabase/neon/issues/6440
part of #6628
Before this PR, we used a std::sync::RwLock to coalesce multiple
callers on one walredo spawning. One thread would win the write lock
and others would queue up either at the read() or write() lock call.
In a scenario where a compute initiates multiple getpage requests
from different Postgres backends (= different page_service conns),
and we don't have a walredo process around, this means all these
page_service handler tasks will enter the spawning code path,
one of them will do the spawning, and the others will stall their
respective executor thread because they do a blocking
read()/write() lock call.
I don't know exactly how bad the impact is in reality because
posix_spawn uses CLONE_VFORK under the hood, which means that the
entire parent process stalls anyway until the child does `exec`,
which in turn resumes the parent.
But, anyway, we won't know until we fix this issue.
And, there's definitely a future way out of stalling the
pageserver on posix_spawn, namely, forking template walredo processes
that fork again when they need to be per-tenant.
This idea is tracked in
https://github.com/neondatabase/neon/issues/7320.
Changes
-------
This PR fixes that scenario by switching to use `heavier_once_cell`
for coalescing. There is a comment on the struct field that explains
it in a bit more nuance.
### Alternative Design
An alternative would be to use tokio::sync::RwLock.
I did this in the first commit in this PR branch,
before switching to `heavier_once_cell`.
Performance
-----------
I re-ran the `bench_walredo` and updated the results, showing that
the changes are neglible.
For the record, the earlier commit in this PR branch that uses
`tokio::sync::RwLock` also has updated benchmark numbers, and the
results / kinds of tiny regression were equivalent to
`heavier_once_cell`.
Note that the above doesn't measure performance on the cold path, i.e.,
when we need to launch the process and coalesce. We don't have a
benchmark
for that, and I don't expect any significant changes. We have metrics
and we log spawn latency, so, we can monitor it in staging & prod.
Risks
-----
As "usual", replacing a std::sync primitive with something that yields
to
the executor risks exposing concurrency that was previously implicitly
limited to the number of executor threads.
This would be the first one for walredo.
The risk is that we get descheduled while the reconstruct data is
already there.
That could pile up reconstruct data.
In practice, I think the risk is low because once we get scheduled
again, we'll
likely have a walredo process ready, and there is no further await point
until walredo is complete and the reconstruct data has been dropped.
This will change with async walredo PR #6548, and I'm well aware of it
in that PR.
## Problem
Currently, using `storcon-cli` requires user to select a token with
either `pageserverapi` or `admin` scope depending on which endpoint
they're using.
## Summary of changes
- In check_permissions, permit access with the admin scope even if the
required scope is missing. The effect is that an endpoint that required
`pageserverapi` now accepts either `pageserverapi` or `admin`, and for
the CLI one can simply use an `admin` scope token for everything.
## Problem
Running test_pageserver_restarts_under_workload in POR #7275 I get the
following assertion failure in prefetch:
```
#5 0x00005587220d4bf0 in ExceptionalCondition (
conditionName=0x7fbf24d003c8 "(ring_index) < MyPState->ring_unused && (ring_index) >= MyPState->ring_last",
fileName=0x7fbf24d00240 "/home/knizhnik/neon.main//pgxn/neon/pagestore_smgr.c", lineNumber=644)
at /home/knizhnik/neon.main//vendor/postgres-v16/src/backend/utils/error/assert.c:66
#6 0x00007fbf24cebc9b in prefetch_set_unused (ring_index=1509) at /home/knizhnik/neon.main//pgxn/neon/pagestore_smgr.c:644
#7 0x00007fbf24cec613 in prefetch_register_buffer (tag=..., force_latest=0x0, force_lsn=0x0)
at /home/knizhnik/neon.main//pgxn/neon/pagestore_smgr.c:891
#8 0x00007fbf24cef21e in neon_prefetch (reln=0x5587233b7388, forknum=MAIN_FORKNUM, blocknum=14110)
at /home/knizhnik/neon.main//pgxn/neon/pagestore_smgr.c:2055
(gdb) p ring_index
$1 = 1509
(gdb) p MyPState->ring_unused
$2 = 1636
(gdb) p MyPState->ring_last
$3 = 1636
```
## Summary of changes
Check status of `prefetch_wait_for`
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
`test_empty_tenant_size` was marked `xfail` and a few other tests were
skipped.
## Summary of changes
Stabilise `test_empty_tenant_size`. This test attempted to disable
checkpointing for the postgres instance
and expected that the synthetic size remains stable for an empty tenant.
When debugging I noticed that
postgres *was* issuing a checkpoint after the transaction in the test
(perhaps something changed since the
test was introduced). Hence, I relaxed the size check to allow for the
checkpoint key written on the pageserver.
Also removed the checks for synthetic size inputs since the expected
values differ between postgres versions.
Closes https://github.com/neondatabase/neon/issues/7138
This PR is a fallout from work on #7062.
# Changes
- Unify the freeze-and-flush and hard shutdown code paths into a single
method `Timeline::shutdown` that takes the shutdown mode as an argument.
- Replace `freeze_and_flush` bool arg in callers with that mode
argument, makes them more expressive.
- Switch timeline deletion to use `Timeline::shutdown` instead of its
own slightly-out-of-sync copy.
- Remove usage of `task_mgr::shutdown_watcher` /
`task_mgr::shutdown_token` where possible
# Future Work
Do we really need the freeze_and_flush?
If we could get rid of it, then there'd be no need for a specific
shutdown order.
Also, if you undo this patch's changes to the `eviction_task.rs` and
enable RUST_LOG=debug, it's easy to see that we do leave some task
hanging that logs under span `Connection{...}` at debug level. I think
it's a pre-existing issue; it's probably a broker client task.
## Problem
For reasons unrelated to this PR, I would like to make use of the tenant
conf in the `InMemoryLayer`. Previously, this was not possible without
copying and manually updating the copy to keep it in sync with updates.
## Summary of Changes:
Replace the `Arc<RwLock<AttachedTenantConf>>` with
`Arc<ArcSwap<AttachedTenantConf>>` (how many `Arc(s)` can one fit in a
type?). The most interesting part of this change is the updating of the
tenant config (`set_new_tenant_config` and
`set_new_location_config`). In theory, these two may race, although the
storage controller should prevent this via the tenant exclusive op lock.
Particular care has been taken to not "lose" a location config update by
using the read-copy-update approach when updating only the config.
Add support for backing up partial segments to remote storage. Disabled
by default, can be enabled with `--partial-backup-enabled`.
Safekeeper timeline has a background task which is subscribed to
`commit_lsn` and `flush_lsn` updates. After the partial segment was
updated (`flush_lsn` was changed), the segment will be uploaded to S3 in
about 15 minutes.
The filename format for partial segments is
`Segment_Term_Flush_Commit_skNN.partial`, where:
- `Segment` – the segment name, like `000000010000000000000001`
- `Term` – current term
- `Flush` – flush_lsn in hex format `{:016X}`, e.g. `00000000346BC568`
- `Commit` – commit_lsn in the same hex format
- `NN` – safekeeper_id, like `1`
The full object name example:
`000000010000000000000002_2_0000000002534868_0000000002534410_sk1.partial`
Each safekeeper will keep info about remote partial segments in its
control file. Code updates state in the control file before doing any S3
operations. This way control file stores information about all
potentially existing remote partial segments and can clean them up after
uploading a newer version.
Closes#6336
## Problem
(Follows https://github.com/neondatabase/neon/pull/7237)
Some API users will query a tenant to wait for it to activate.
Currently, we return the current status of the tenant, whatever that may
be. Under heavy load, a pageserver starting up might take a long time to
activate such a tenant.
## Summary of changes
- In `tenant_status` handler, call wait_to_become_active on the tenant.
If the tenant is currently waiting for activation, this causes it to
skip the queue, similiar to other API handlers that require an active
tenant, like timeline creation. This avoids external services waiting a
long time for activation when polling GET /v1/tenant/<id>.
Tiered compaction hasn't scheduled the upload of image layers. In the
`test_gc_feedback.py` test this has caused warnings like with tiered
compaction:
```
INFO request[...] Deleting layer [...] not found in latest_files list, never uploaded?
```
Which caused errors like:
```
ERROR layer_delete[...] was unlinked but was not dangling
```
Fixes#7244
We want to move the code base away from task_mgr.
This PR refactors the walreceiver code such that it doesn't use
`task_mgr` anymore.
# Background
As a reminder, there are three tasks in a Timeline that's ingesting WAL.
`WalReceiverManager`, `WalReceiverConnectionHandler`, and
`WalReceiverConnectionPoller`.
See the documentation in `task_mgr.rs` for how they interact.
Before this PR, cancellation was requested through
task_mgr::shutdown_token() and `TaskHandle::shutdown`.
Wait-for-task-finish was implemented using a mixture of
`task_mgr::shutdown_tasks` and `TaskHandle::shutdown`.
This drawing might help:
<img width="300" alt="image"
src="https://github.com/neondatabase/neon/assets/956573/b6be7ad6-ecb3-41d0-b410-ec85cb8d6d20">
# Changes
For cancellation, the entire WalReceiver task tree now has a
`child_token()` of `Timeline::cancel`. The `TaskHandle` no longer is a
cancellation root.
This means that `Timeline::cancel.cancel()` is propagated.
For wait-for-task-finish, all three tasks in the task tree hold the
`Timeline::gate` open until they exit.
The downside of using the `Timeline::gate` is that we can no longer wait
for just the walreceiver to shut down, which is particularly relevant
for `Timeline::flush_and_shutdown`.
Effectively, it means that we might ingest more WAL while the
`freeze_and_flush()` call is ongoing.
Also, drive-by-fix the assertiosn around task kinds in `wait_lsn`. The
check for `WalReceiverConnectionHandler` was ineffective because that
never was a task_mgr task, but a TaskHandle task. Refine the assertion
to check whether we would wait, and only fail in that case.
# Alternatives
I contemplated (ab-)using the `Gate` by having a separate `Gate` for
`struct WalReceiver`.
All the child tasks would use _that_ gate instead of `Timeline::gate`.
And `struct WalReceiver` itself would hold an `Option<GateGuard>` of the
`Timeline::gate`.
Then we could have a `WalReceiver::stop` function that closes the
WalReceiver's gate, then drops the `WalReceiver::Option<GateGuard>`.
However, such design would mean sharing the WalReceiver's `Gate` in an
`Arc`, which seems awkward.
A proper abstraction would be to make gates hierarchical, analogous to
CancellationToken.
In the end, @jcsp and I talked it over and we determined that it's not
worth the effort at this time.
# Refs
part of #7062
The latest failures of test_secondary_downloads are spooky: layers are
missing on disk according to the test, but present according to the
pageserver logs:
- Make the pageserver assert that layers are really present on disk and
log the full path (debug mode only)
- Make the test dump a full listing on failure of the assert that failed
the last two times
Related: #6966
## Problem
https://github.com/neondatabase/cloud/issues/11051
additionally, I felt like the http logic was a bit complex.
## Summary of changes
1. Removes timeout for HTTP requests.
2. Split out header parsing to a `HttpHeaders` type.
3. Moved db client handling to `QueryData::process` and
`BatchQueryData::process` to simplify the logic of `handle_inner` a bit.
## Problem
During incidents, we may need to quickly access the storage controller's
API without trying API client code or crafting `curl` CLIs on the fly. A
basic CLI client is needed for this.
## Summary of changes
- Update storage controller node listing API to only use public types in
controller_api.rs
- Add a storage controller API for listing tenants
- Add a basic test that the CLI can list and modify nodes and tenants.
## Problem
The vectored read path holds the layer map lock while visiting a
timeline.
## Summary of changes
* Rework the fringe order to hold `Layer` on `Arc<InMemoryLayer>`
handles instead of descriptions that are resolved by the layer map at
the time of read. Note that previously `get_values_reconstruct_data` was
implemented for the layer description which already knew the lsn range
for the read. Now it is implemented on the new `ReadableLayer` handle
and needs to get the lsn range as an argument.
* Drop the layer map lock after updating the fringe.
Related https://github.com/neondatabase/neon/issues/6833
- Cleanup from
https://github.com/neondatabase/neon/pull/7040#discussion_r1521120263 --
in that PR, we needed to let compat tests manually register a node,
because it would run an old binary that doesn't self-register.
- Cleanup vectored get config workaround
- Cleanup a log allow list for which the underlying log noise has been
fixed.
## Problem
During Nightly Benchmarks, we want to collect pgbench results for
sharded tenants as well.
## Summary of changes
- Add pre-created sharded project for pgbench
## Problem
## Summary of changes
`cargo update -p tokio`.
The only risky change I could see is the `tokio::io::split` moving from
a spin-lock to a mutex but I think that's ok.
Fix#7278
## Summary of changes
* Explicitly create the extension download directory and assign correct
permissoins.
* Fix the problem that the extension download failure will cause all
future downloads to fail.
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
- When we scheduled locations, we were doing it without any context
about other shards in the same tenant
- After a shard split, there wasn't an automatic mechanism to migrate
the attachments away from the split location
- After a shard split and the migration away from the split location,
there wasn't an automatic mechanism to pick new secondary locations so
that the end state has no concentration of locations on the nodes where
the split happened.
Partially completes: https://github.com/neondatabase/neon/issues/7139
## Summary of changes
- Scheduler now takes a `ScheduleContext` object that can be populated
with information about other shards
- During tenant creation and shard split, we incrementally build up the
ScheduleContext, updating it for each shard as we proceed.
- When scheduling new locations, the ScheduleContext is used to apply a
soft anti-affinity to nodes where a tenant already has shards.
- The background reconciler task now has an extra phase `optimize_all`,
which runs only if the primary `reconcile_all` phase didn't generate any
work. The separation is that `reconcile_all` is needed for availability,
but optimize_all is purely "nice to have" work to balance work across
the nodes better.
- optimize_all calls into two new TenantState methods called
optimize_attachment and optimize_secondary, which seek out opportunities
to improve placment:
- optimize_attachment: if the node where we're currently attached has an
excess of attached shard locations for this tenant compared with the
node where we have a secondary location, then cut over to the secondary
location.
- optimize_secondary: if the node holding our secondary location has an
excessive number of locations for this tenant compared with some other
node where we don't currently have a location, then create a new
secondary location on that other node.
- a new debug API endpoint is provided to run background tasks
on-demand. This returns a number of reconciliations in progress, so
callers can keep calling until they get a `0` to advance the system to
its final state without waiting for many iterations of the background
task.
Optimization is run at an implicitly low priority by:
- Omitting the phase entirely if reconcile_all has work to do
- Skipping optimization of any tenant that has reconciles in flight
- Limiting the total number of optimizations that will be run from one
call to optimize_all to a constant (currently 2).
The idea of that low priority execution is to minimize the operational
risk that optimization work overloads any part of the system. It happens
to also make the system easier to observe and debug, as we avoid running
large numbers of concurrent changes. Eventually we may relax these
limitations: there is no correctness problem with optimizing lots of
tenants concurrently, and optimizing multiple shards in one tenant just
requires housekeeping changes to update ShardContext with the result of
one optimization before proceeding to the next shard.
## Problem
Part of the legacy (but current) compaction algorithm is to find a stack
of overlapping delta layers which will be turned
into an image layer. This operation is exponential in terms of the
number of matching layers and we do it roughly every 20 seconds.
## Summary of changes
Only check if a new image layer is required if we've ingested a certain
amount of WAL since the last check.
The amount of wal is expressed in terms of multiples of checkpoint
distance, with the intuition being that
that there's little point doing the check if we only have two new L1
layers (not enough to create a new image).
## Problem
- Control plane can deadlock if it calls into a function that requires
reconciliation to complete, while refusing compute notification hooks
API calls.
## Summary of changes
- Fail faster in the notify path in 438 errors: these were originally
expected to be transient, but in practice it's more common that a 438
results from an operation blocking on the currently API call, rather
than something happening in the background.
- In ensure_attached, relax the condition for spawning a reconciler:
instead of just the general maybe_reconcile path, do a pre-check that
skips trying to reconcile if the shard appears to be attached. This
avoids doing work in cases where the tenant is attached, but is dirty
from a reconciliation point of view, e.g. due to a failed compute
notification.
## Problem
Proxy release to a preprod automatically triggers a deployment of storage
controller (`deployStorageController=true` by default)
## Summary of changes
- Set `deployStorageController=false` for proxy releases to preprod
- Set explicitly `deployStorageController=true` for storage releases to
preprod and prod
## Problem
During this week's deployment we observed panics due to the blobs
for certain keys not fitting in the vectored read buffers. The likely
cause of this is a bloated AUX_FILE_KEY caused by logical replication.
## Summary of changes
This pr fixes the issue by allocating a buffer big enough to fit
the widest read. It also has the benefit of saving space if all keys
in the read have blobs smaller than the max vectored read size.
If the soft limit for the max size of a vectored read is violated,
we print a warning which includes the offending key and lsn.
A randomised (but deterministic) end to end test is also added for
vectored reads on the delta layer.
## Problem
In the event of bugs with scheduling or reconciliation, we need to be
able to switch this off at a per-tenant granularity.
This is intended to mitigate risk of issues with
https://github.com/neondatabase/neon/pull/7181, which makes scheduling
more involved.
Closes: #7103
## Summary of changes
- Introduce a scheduling policy per tenant, with API to set it
- Refactor persistent.rs helpers for updating tenants to be more general
- Add tests
Many tests like `test_live_migration` or
`test_timeline_deletion_with_files_stuck_in_upload_queue` set
`compaction_threshold` to 1, to create a lot of changes/updates. The
compaction threshold was passed as `fanout` parameter to the
tiered_compaction function, which didn't support values of 1 however.
Now we change the assert to support it, while still retaining the
exponential nature of the increase in range in terms of lsn that a layer
is responsible for.
A large chunk of the failures in #6964 was due to hitting this issue
that we now resolved.
Part of #6768.
## Problem
See https://github.com/neondatabase/cloud/issues/11559
If we have multiple shards, we need to reset connections to all shards
involved in prefetch (having active prefetch requests) if connection
with any of them is lost.
## Summary of changes
In `prefetch_on_ps_disconnect` drop connection to all shards with active
page requests.
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
We don't want to run an excessive e2e test suite on neonvm if there are
no relevant changes.
## Summary of changes
- Check PR diff and if there are no relevant compute changes (in
`vendor/`, `pgxn/`, `libs/vm_monitor` or `Dockerfile.compute-node`
- Switch job from `small` to `ubuntu-latest` runner to make it possible
to use GitHub CLI
# Problem
As pointed out through doc-comments in this PR, `drop_old_connection` is
not cancellation-safe.
This means we can leave a `handle_walreceiver_connection` tokio task
dangling during Timeline shutdown.
More details described in the corresponding issue #7062.
# Solution
Don't cancel-by-drop the `connection_manager_loop_step` from the
`tokio::select!()` in the task_mgr task.
Instead, transform the code to use a `CancellationToken` ---
specifically, `task_mgr::shutdown_token()` --- and make code responsive
to it.
The `drop_old_connection()` is still not cancellation-safe and also
doesn't get a cancellation token, because there's no point inside the
function where we could return early if cancellation were requested
using a token.
We rely on the `handle_walreceiver_connection` to be sensitive to the
`TaskHandle`s cancellation token (argument name: `cancellation`).
Currently it checks for `cancellation` on each WAL message. It is
probably also sensitive to `Timeline::cancel` because ultimately all
that `handle_walreceiver_connection` does is interact with the
`Timeline`.
In summary, the above means that the following code (which is found in
`Timeline::shutdown`) now might **take longer**, but actually ensures
that all `handle_walreceiver_connection` tasks are finished:
```rust
task_mgr::shutdown_tasks(
Some(TaskKind::WalReceiverManager),
Some(self.tenant_shard_id),
Some(self.timeline_id)
)
```
# Refs
refs #7062
## Problem
https://github.com/neondatabase/cloud/issues/9642
## Summary of changes
1. Make `EndpointRateLimiter` generic, renamed as `BucketRateLimiter`
2. Add support for claiming multiple tokens at once
3. Add `AuthRateLimiter` alias.
4. Check `(Endpoint, IP)` pair during authentication, weighted by how
many hashes proxy would be doing.
TODO: handle ipv6 subnets. will do this in a separate PR.
## Problem
This is a refactor.
This PR was a precursor to a much smaller change
e5bd602dc1,
where as I was writing it I found that we were not far from getting rid
of the last non-deprecated code paths that use `mgr::` scoped functions
to get at the TenantManager state.
We're almost done cleaning this up as per
https://github.com/neondatabase/neon/issues/5796. The only significant
remaining mgr:: item is `get_active_tenant_with_timeout`, which is
page_service's path for fetching tenants.
## Summary of changes
- Remove the bool argument to get_attached_tenant_shard: this was almost
always false from API use cases, and in cases when it was true, it was
readily replacable with an explicit check of the returned tenant's
status.
- Rather than letting the timeline eviction task query any tenant it
likes via `mgr::`, pass an `Arc<Tenant>` into the task. This is still an
ugly circular reference, but should eventually go away: either when we
switch to exclusively using disk usage eviction, or when we change
metadata storage to avoid the need to imitate layer accesses.
- Convert all the mgr::get_tenant call sites to use
TenantManager::get_attached_tenant_shard
- Move list_tenants into TenantManager.
## Problem
https://github.com/neondatabase/neon/pull/7227 destabilized various
tests in the performance suite, with log errors during shutdown. It's
because we switched shutdown order to stop the storage controller before
the pageservers.
## Summary of changes
- Tolerate "connection failed" errors from pageservers trying to
validation their deletion queue.
## Problem
- Creations were not idempotent (unique key violation)
- Creations waited for reconciliation, which control plane blocks while
an operation is in flight
## Summary of changes
- Handle unique key constraint violation as an OK situation: if we're
creating the same tenant ID and shard count, it's reasonable to assume
this is a duplicate creation.
- Make the wait for reconcile during creation tolerate failures: this is
similar to location_conf, where the cloud control plane blocks our
notification calls until it is done with calling into our API (in future
this constraint is expected to relax as the cloud control plane learns
to run multiple operations concurrently for a tenant)
## Problem
Follows: https://github.com/neondatabase/neon/pull/7182
- Sufficient concurrent writes could OOM a pageserver from the size of
indices on all the InMemoryLayer instances.
- Enforcement of checkpoint_period only happened if there were some
writes.
Closes: https://github.com/neondatabase/neon/issues/6916
## Summary of changes
- Add `ephemeral_bytes_per_memory_kb` config property. This controls the
ratio of ephemeral layer capacity to memory capacity. The weird unit is
to enable making the ratio less than 1:1 (set this property to 1024 to
use 1MB of ephemeral layers for every 1MB of RAM, set it smaller to get
a fraction).
- Implement background layer rolling checks in
Timeline::compaction_iteration -- this ensures we apply layer rolling
policy in the absence of writes.
- During background checks, if the total ephemeral layer size has
exceeded the limit, then roll layers whose size is greater than the mean
size of all ephemeral layers.
- Remove the tick() path from walreceiver: it isn't needed any more now
that we do equivalent checks from compaction_iteration.
- Add tests for the above.
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
Currently, we return 409 (Conflict) in two cases:
- Temporary: Timeline creation cannot proceed because another timeline
with the same ID is being created
- Permanent: Timeline creation cannot proceed because another timeline
exists with different parameters but the same ID.
Callers which time out a request and retry should be able to distinguish
these cases.
Closes: #7208
## Summary of changes
- Expose `AlreadyCreating` errors as 429 instead of 409
## Problem
We currently hold the layer map read lock while doing IO on the read
path. This is not required for correctness.
## Summary of changes
Drop the layer map lock after figuring out which layer we wish to read
from.
Why is this correct:
* `Layer` models the lifecycle of an on disk layer. In the event the
layer is removed from local disk, it will be on demand downloaded
* `InMemoryLayer` holds the `EphemeralFile` which wraps the on disk
file. As long as the `InMemoryLayer` is in scope, it's safe to read from it.
Related https://github.com/neondatabase/neon/issues/6833
## Problem
- https://github.com/neondatabase/neon/issues/6966
This test occasionally failed with some layers unexpectedly not present
on the secondary pageserver. The issue in that failure is the attached
pageserver uploading heatmaps that refer to not-yet-uploaded layers.
## Summary of changes
After uploading heatmap, drain upload queue on attached pageserver, to
guarantee that all the layers referenced in the haetmap are uploaded.
## Problem
While most forms of split rollback don't interrupt clients, there are a
couple of cases that do -- this interruption is brief, driven by the
time it takes the controller to kick off Reconcilers during the async
abort of the split, so it's operationally fine, but can trip up a test.
- #7148
## Summary of changes
- Relax test check to require that the tenant is eventually available
after split failure, rather than immediately. In the vast majority of
cases this will pass on the first iteration.
This test had two flaky failure modes:
- pageserver log error for timeline not found: this resulted from
changes for DR when timeline destroy/create was added, but endpoint was
left running during that operation.
- storage controller log error because the test was running for long
enough that a background reconcile happened at almost the exact moment
of test teardown, and our test fixtures tear down the pageservers before
the controller.
Closes: #7224
Postgres can always write some more WAL, so previous checks that WAL doesn't
change after something had been crafted were wrong; remove them. Add comments
here and there.
should fix https://github.com/neondatabase/neon/issues/4691
## Problem
Large quantities of ephemeral layer data can lead to excessive memory
consumption (https://github.com/neondatabase/neon/issues/6939). We
currently don't have a way to know how much ephemeral layer data is
present on a pageserver.
Before we can add new behaviors to proactively roll layers in response
to too much ephemeral data, we must calculate that total.
Related: https://github.com/neondatabase/neon/issues/6916
## Summary of changes
- Create GlobalResources and GlobalResourceUnits types, where timelines
carry a GlobalResourceUnits in their TimelineWriterState.
- Periodically update the size in GlobalResourceUnits:
- During tick()
- During layer roll
- During put() if the latest value has drifted more than 10MB since our
last update
- Expose the value of the global ephemeral layer bytes counter as a
prometheus metric.
- Extend the lifetime of TimelineWriterState:
- Instead of dropping it in TimelineWriter::drop, let it remain.
- Drop TimelineWriterState in roll_layer: this drops our guard on the
global byte count to reflect the fact that we're freezing the layer.
- Ensure the validity of the later in the writer state by clearing the
state in the same place we freeze layers, and asserting on the
write-ability of the layer in `writer()`
- Add a 'context' parameter to `get_open_layer_action` so that it can
skip the prev_lsn==lsn check when called in tick() -- this is needed
because now tick is called with a populated state, where
prev_lsn==Some(lsn) is true for an idle timeline.
- Extend layer rolling test to use this metric
- Remove code for using AWS secrets manager, as we're deploying with
k8s->env vars instead
- Load each secret independently, so that one can mix CLI args with
environment variables, rather than requiring that all secrets are loaded
with the same mechanism.
- Add a 'strict mode', enabled by default, which will refuse to start if
secrets are not loaded. This avoids the risk of accidentially disabling
auth by omitting the public key, for example
## Problem
We recently introduced log file validation for the storage controller.
The heartbeater will WARN when it fails
for a node, hence the test fails.
Closes https://github.com/neondatabase/neon/issues/7159
## Summary of changes
* Warn only once for each set of heartbeat retries
* Allow list heartbeat warns
Before this PR, each core had 3 executor threads from 3 different
runtimes. With this PR, we just have one runtime, with one thread per
core. Switching to a single tokio runtime should reduce that effective
over-commit of CPU and in theory help with tail latencies -- iff all
tokio tasks are well-behaved and yield to the runtime regularly.
Are All Tasks Well-Behaved? Are We Ready?
-----------------------------------------
Sadly there doesn't seem to be good out-of-the box tokio tooling to
answer this question.
We *believe* all tasks are well behaved in today's code base, as of the
switch to `virtual_file_io_engine = "tokio-epoll-uring"` in production
(https://github.com/neondatabase/aws/pull/1121).
The only remaining executor-thread-blocking code is walredo and some
filesystem namespace operations.
Filesystem namespace operations work is being tracked in #6663 and not
considered likely to actually block at this time.
Regarding walredo, it currently does a blocking `poll` for read/write to
the pipe file descriptors we use for IPC with the walredo process.
There is an ongoing experiment to make walredo async (#6628), but it
needs more time because there are surprisingly tricky trade-offs that
are articulated in that PR's description (which itself is still WIP).
What's relevant for *this* PR is that
1. walredo is always CPU-bound
2. production tail latencies for walredo request-response
(`pageserver_wal_redo_seconds_bucket`) are
- p90: with few exceptions, low hundreds of micro-seconds
- p95: except on very packed pageservers, below 1ms
- p99: all below 50ms, vast majority below 1ms
- p99.9: almost all around 50ms, rarely at >= 70ms
- [Dashboard
Link](https://neonprod.grafana.net/d/edgggcrmki3uof/2024-03-walredo-latency?orgId=1&var-ds=ZNX49CDVz&var-pXX_by_instance=0.9&var-pXX_by_instance=0.99&var-pXX_by_instance=0.95&var-adhoc=instance%7C%21%3D%7Cpageserver-30.us-west-2.aws.neon.tech&var-per_instance_pXX_max_seconds=0.0005&from=1711049688777&to=1711136088777)
The ones below 1ms are below our current threshold for when we start
thinking about yielding to the executor.
The tens of milliseconds stalls aren't great, but, not least because of
the implicit overcommit of CPU by the three runtimes, we can't be sure
whether these tens of milliseconds are inherently necessary to do the
walredo work or whether we could be faster if there was less contention
for CPU.
On the first item (walredo being always CPU-bound work): it means that
walredo processes will always compete with the executor threads.
We could yield, using async walredo, but then we hit the trade-offs
explained in that PR.
tl;dr: the risk of stalling executor threads through blocking walredo
seems low, and switching to one runtime cleans up one potential source
for higher-than-necessary stall times (explained in the previous
paragraphs).
Code Changes
------------
- Remove the 3 different runtime definitions.
- Add a new definition called `THE_RUNTIME`.
- Use it in all places that previously used one of the 3 removed
runtimes.
- Remove the argument from `task_mgr`.
- Fix failpoint usage where `pausable_failpoint!` should have been used.
We encountered some actual failures because of this, e.g., hung
`get_metric()` calls during test teardown that would client-timeout
after 300s.
As indicated by the comment above `THE_RUNTIME`, we could take this
clean-up further.
But before we create so much churn, let's first validate that there's no
perf regression.
Performance
-----------
We will test this in staging using the various nightly benchmark runs.
However, the worst-case impact of this change is likely compaction
(=>image layer creation) competing with compute requests.
Image layer creation work can't be easily generated & repeated quickly
by pagebench.
So, we'll simply watch getpage & basebackup tail latencies in staging.
Additionally, I have done manual benchmarking using pagebench.
Report:
https://neondatabase.notion.site/2024-03-23-oneruntime-change-benchmarking-22a399c411e24399a73311115fb703ec?pvs=4
Tail latencies and throughput are marginally better (no regression =
good).
Except in a workload with 128 clients against one tenant.
There, the p99.9 and p99.99 getpage latency is about 2x worse (at
slightly lower throughput).
A dip in throughput every 20s (compaction_period_ is clearly visible,
and probably responsible for that worse tail latency.
This has potential to improve with async walredo, and is an edge case
workload anyway.
Future Work
-----------
1. Once this change has shown satisfying results in production, change
the codebase to use the ambient runtime instead of explicitly
referencing `THE_RUNTIME`.
2. Have a mode where we run with a single-threaded runtime, so we
uncover executor stalls more quickly.
3. Switch or write our own failpoints library that is async-native:
https://github.com/neondatabase/neon/issues/7216
## Problem
stack overflow in blanket impl for `CancellationPublisher`
## Summary of changes
Removes `async_trait` and fixes the impl order to make it non-recursive.
## Problem
The service that receives consumption metrics has lower availability
than S3. Writing metrics to S3 improves their availability.
Closes: https://github.com/neondatabase/cloud/issues/9824
## Summary of changes
- The same data as consumption metrics POST bodies is also compressed
and written to an S3 object with a timestamp-formatted path.
- Set `metric_collection_bucket` (same format as `remote_storage`
config) to configure the location to write to
## Problem
We want to deploy releases to a preprod region first to perform required
checks
## Summary of changes
- Deploy `release-XXX` / `release-proxy-YYY` docker tags to a preprod region
## Problem
I noticed code coverage for auth_quirks was pretty bare
## Summary of changes
Adds 3 happy path unit tests for auth_quirks
* scram
* cleartext (websockets)
* cleartext (password hack)
A test was added which exercises secondary locations more, and there was
a location in the secondary downloader that warned on ephemeral files.
This was intended to be fixed in this faulty commit:
8cea866adf
## Problem
Support of IAM Roles for Service Accounts for authentication.
## Summary of changes
* Obtain aws 15m-long credentials
* Retrieve redis password from credentials
* Update every 1h to keep connection for more than 12h
* For now allow to have different endpoints for pubsub/stream redis.
TODOs:
* PubSub doesn't support credentials refresh, consider using stream
instead.
* We need an AWS role for proxy to be able to connect to both: S3 and
elasticache.
Credentials obtaining and connection refresh was tested on xenon
preview.
https://github.com/neondatabase/cloud/issues/10365
Release notes: https://blog.rust-lang.org/2024/03/21/Rust-1.77.0.html
Thanks to #6886 the diff is reasonable, only for one new lint
`clippy::suspicious_open_options`. I added `truncate()` calls to the
places where it is obviously the right choice to me, and added allows
everywhere else, leaving it for followups.
I had to specify cargo install --locked because the build would fail otherwise.
This was also recommended by upstream.
See the updated `bench_walredo.rs` module comment.
tl;dr: we measure avg latency of single redo operations issues against a
single redo manager from N tokio tasks.
part of https://github.com/neondatabase/neon/issues/6628
## Problem
for HTTP/WS/password hack flows we imitate SCRAM to validate passwords.
This code was unnecessarily complicated.
## Summary of changes
Copy in the `pbkdf2` and 'derive keys' steps from the
`postgres_protocol` crate in our `rust-postgres` fork. Derive the
`client_key`, `server_key` and `stored_key` from the password directly.
Use constant time equality to compare the `stored_key` and `server_key`
with the ones we are sent from cplane.
This change improves the resilience of the system to unclean restarts.
Previously, re-attach responses only included attached tenants
- If the pageserver had local state for a secondary location, it would
remain, but with no guarantee that it was still _meant_ to be there.
After this change, the pageserver will only retain secondary locations
if the /re-attach response indicates that they should still be there.
- If the pageserver had local state for an attached location that was
omitted from a re-attach response, it would be entirely detached. This
is wasteful in a typical HA setup, where an offline node's tenants might
have been re-attached elsewhere before it restarts, but the offline
node's location should revert to a secondary location rather than being
wiped. Including secondary tenants in the re-attach response enables the
pageserver to avoid throwing away local state unnecessarily.
In this PR:
- The re-attach items are extended with a 'mode' field.
- Storage controller populates 'mode'
- Pageserver interprets it (default is attached if missing) to construct
either a SecondaryTenant or a Tenant.
- A new test exercises both cases.
## Problem
If a shutdown happens when a tenant is attaching, we were logging at
ERROR severity and with a backtrace. Yuck.
## Summary of changes
- Pass a flag into `make_broken` to enable quietening this non-scary
case.
Stacks on:
- https://github.com/neondatabase/neon/pull/7165
Fixes while working on background optimization of scheduling after a
split:
- When a tenant has secondary locations, we weren't detaching the parent
shards' secondary locations when doing a split
- When a reconciler detaches a location, it was feeding back a
locationconf with `Detached` mode in its `observed` object, whereas it
should omit that location. This could cause the background reconcile
task to keep kicking off no-op reconcilers forever (harmless but
annoying).
- During shard split, we were scheduling secondary locations for the
child shards, but no reconcile was run for these until the next time the
background reconcile task ran. Creating these ASAP is useful, because
they'll be used shortly after a shard split as the destination locations
for migrating the new shards to different nodes.
## Problem
Storage controller had basically no metrics.
## Summary of changes
1. Migrate the existing metrics to use Conrad's
[`measured`](https://docs.rs/measured/0.0.14/measured/) crate.
2. Add metrics for incoming http requests
3. Add metrics for outgoing http requests to the pageserver
4. Add metrics for outgoing pass through requests to the pageserver
5. Add metrics for database queries
Note that the metrics response for the attachment service does not use
chunked encoding like the rest of the metrics endpoints. Conrad has
kindly extended the crate such that it can now be done. Let's leave it
for a follow-up since the payload shouldn't be that big at this point.
Fixes https://github.com/neondatabase/neon/issues/6875
## Problem
The current implementation of struct Layer supports canceled read
requests, but those will leave the internal state such that a following
`Layer::keep_resident` call will need to repair the state. In
pathological cases seen during generation numbers resetting in staging
or with too many in-progress on-demand downloads, this repair activity
will need to wait for the download to complete, which stalls disk
usage-based eviction. Similar stalls have been observed in staging near
disk-full situations, where downloads failed because the disk was full.
Fixes#6028 or the "layer is present on filesystem but not evictable"
problems by:
1. not canceling pending evictions by a canceled
`LayerInner::get_or_maybe_download`
2. completing post-download initialization of the `LayerInner::inner`
from the download task
Not canceling evictions above case (1) and always initializing (2) lead
to plain `LayerInner::inner` always having the up-to-date information,
which leads to the old `Layer::keep_resident` never having to wait for
downloads to complete. Finally, the `Layer::keep_resident` is replaced
with `Layer::is_likely_resident`. These fix#7145.
## Summary of changes
- add a new test showing that a canceled get_or_maybe_download should
not cancel the eviction
- switch to using a `watch` internally rather than a `broadcast` to
avoid hanging eviction while a download is ongoing
- doc changes for new semantics and cleanup
- fix `Layer::keep_resident` to use just `self.0.inner.get()` as truth
as `Layer::is_likely_resident`
- remove `LayerInner::wanted_evicted` boolean as no longer needed
Builds upon: #7185. Cc: #5331.
`pgxn/` also contains WAL proposer code, so modifications to this
directory should be able to be approved by the safekeeper team.
Signed-off-by: Alex Chi Z <chi@neon.tech>
Before this PR, cancellation for `LayerInner::get_or_maybe_download`
could occur so that we have downloaded the layer file in the filesystem,
but because of the cancellation chance, we have not set the internal
`LayerInner::inner` or initialized the state. With the detached init
support introduced in #7135 and in place in #7152, we can now initialize
the internal state after successfully downloading in the spawned task.
The next PR will fix the remaining problems that this PR leaves:
- `Layer::keep_resident` is still used because
- `Layer::get_or_maybe_download` always cancels an eviction, even when
canceled
Split off from #7030. Stacked on top of #7152. Cc: #5331.
- Enable debug logs for this test
- Add some debug logging detail in downloader.rs
- Add an info-level message in scheduler.rs that makes it obvious if a
command is waiting for an existing task rather than spawning a new one.
The second part of work towards fixing `Layer::keep_resident` so that it
does not need to repair the internal state. #7135 added a nicer API for
initialization. This PR uses it to remove a few indentation levels and
the loop construction. The next PR #7175 will use the refactorings done
in this PR, and always initialize the internal state after a download.
Cc: #5331
Since #6115 with more often used get_value_reconstruct_data and friends,
we should not have needless INFO level span creation near hot paths. In
our prod configuration, INFO spans are always created, but in practice,
very rarely anything at INFO level is logged underneath.
`ResidentLayer::load_keys` is only used during compaction so it is not
that hot, but this aligns the access paths and their span usage.
PR changes the span level to debug to align with others, and adds the
layer name to the error which was missing.
Split off from #7030.
## Problem
faster sha2 hashing.
## Summary of changes
enable asm feature for sha2. this feature will be default in sha2 0.11,
so we might as well lean into it now. It provides a noticeable speed
boost on macos aarch64. Haven't tested on x86 though
Warm-up (and the "tenant startup complete" metric update) happens in
a background tokio task. The tenant map is eagerly updated (can happen
before the task finishes).
The test assumed that if the tenant map was updated, then the metric
should reflect that. That's not the case, so we tweak the test to wait
for the metric.
Fixes https://github.com/neondatabase/neon/issues/7158
This is a mixed bag of changes split out for separate review while
working on other things, and batched together to reduce load on CI
runners. Each commits stands alone for review purposes:
- do_tenant_shard_split was a long function and had a synchronous
validation phase at the start that could readily be pulled out into a
separate function. This also avoids the special casing of
ApiError::BadRequest when deciding whether an abort is needed on errors
- Add a 'describe' API (GET on tenant ID) that will enable storcon-cli
to see what's going on with a tenant
- the 'locate' API wasn't really meant for use in the field. It's for
tests: demote it to the /debug/ prefix
- The `Single` placement policy was a redundant duplicate of Double(0),
and Double was a bad name. Rename it Attached.
(https://github.com/neondatabase/neon/issues/7107)
- Some neon_local commands were added for debug/demos, which are now
replaced by commands in storcon-cli (#7114 ). Even though that's not
merged yet, we don't need the neon_local ones any more.
Closes https://github.com/neondatabase/neon/issues/7107
## Backward compat of Single/Double -> `Attached(n)` change
A database migration is used to convert any existing values.
e2e tests cannot run on macOS unless the file engine env var is
supplied.
```
./scripts/pytest test_runner/regress/test_neon_superuser.py -s
```
will fail with tokio-epoll-uring not supported.
This is because we persist the file engine config by default. In this
pull request, we only persist when someone specifies it, so that it can
use the default platform-variant config in the page server.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
As with the pageserver, we should fail tests that emit unexpected log
errors/warnings.
## Summary of changes
- Refactor existing log checks to be reusable
- Run log checks for attachment_service
- Add allow lists as needed.
Add shard_number to PageserverFeedback and parse it on the compute side.
When compute receives a new ps_feedback, it calculates min LSNs among
feedbacks from all shards, and uses those LSNs for backpressure.
Add `test_sharding_backpressure` to verify that backpressure slows down
compute to wait for the slowest shard.
Manual testing of the changes in #7160 revealed that, if the
thread-local destructor ever runs (it apparently doesn't in our test
suite runs, otherwise #7160 would not have auto-merged), we can
encounter an `abort()` due to a double-panic in the tracing code.
This github comment here contains the stack trace:
https://github.com/neondatabase/neon/pull/7160#issuecomment-2003778176
This PR reverts #7160 and uses a atomic counter to identify the
thread-local in log messages, instead of the memory address of the
thread local, which may be re-used.
with `immediate_gc` the span only covered the `gc_iteration`, make it
cover the whole needless spawned task, which also does waiting for layer
drops and stray logging in tests.
also clarify some comments while we are here.
Fixes: #6910
The PR #7141 added log message
```
ThreadLocalState is being dropped and id might be re-used in the future
```
which was supposed to be emitted when the thread-local is destroyed.
Instead, it was emitted on _each_ call to `thread_local_system()`,
ie.., on each tokio-epoll-uring operation.
Testing
-------
Reproduced the issue locally and verified that this PR fixes the issue.
## Problem
Followup to https://github.com/neondatabase/neon/pull/6725
In that PR, code for purging local files from a tenant shard was
duplicated.
## Summary of changes
- Refactor detach code into TenantManager
- `spawn_background_purge` method can now be common between detach and
split operations
refs https://github.com/neondatabase/neon/issues/7136
Problem
-------
Before this PR, we were using
`tokio_epoll_uring::thread_local_system()`,
which panics on tokio_epoll_uring::System::launch() failure
As we've learned in [the
past](https://github.com/neondatabase/neon/issues/6373#issuecomment-1905814391),
some older Linux kernels account io_uring instances as locked memory.
And while we've raised the limit in prod considerably, we did hit it
once on 2024-03-11 16:30 UTC.
That was after we enabled tokio-epoll-uring fleet-wide, but before
we had shipped release-5090 (c6ed86d3d0)
which did away with the last mass-creation of tokio-epoll-uring
instances as per
commit 3da410c8fe
Author: Christian Schwarz <christian@neon.tech>
Date: Tue Mar 5 10:03:54 2024 +0100
tokio-epoll-uring: use it on the layer-creating code paths (#6378)
Nonetheless, it highlighted that panicking in this situation is probably
not ideal, as it can leave the pageserver process in a semi-broken
state.
Further, due to low sampling rate of Prometheus metrics, we don't know
much about the circumstances of this failure instance.
Solution
--------
This PR implements a custom thread_local_system() that is
pageserver-aware
and will do the following on failure:
- dump relevant stats to `tracing!`, hopefully they will be useful to
understand the circumstances better
- if it's the locked memory failure (or any other ENOMEM): abort() the
process
- if it's ENOMEM, retry with exponential back-off, capped at 3s.
- add metric counters so we can create an alert
This makes sense in the production environment where we know that
_usually_, there's ample locked memory allowance available, and we know
the failure rate is rare.
## Problem
The existing secondary download API relied on the caller to wait as long
as it took to complete -- for large shards that could be a long time, so
typical clients that might have a baked-in ~30s timeout would have a
problem.
## Summary of changes
- Take a `wait_ms` query parameter to instruct the pageserver how long
to wait: if the download isn't complete in this duration, then 201 is
returned instead of 200.
- For both 200 and 201 responses, include response body describing
download progress, in terms of layers and bytes. This is sufficient for
the caller to track how much data is being transferred and log/present
that status.
- In storage controller live migrations, use this API to apply a much
longer outer timeout, with smaller individual per-request timeouts, and
log the progress of the downloads.
- Add a test that injects layer download delays to exercise the new
behavior
# Problem
On-demand downloads are still using `tokio::fs`, which we know is
inefficient.
# Changes
- Add `pagebench ondemand-download-churn` to quantify on-demand download
throughput
- Requires dumping layer map, which required making `history_buffer`
impl `Deserialize`
- Implement an equivalent of `tokio::io::copy_buf` for owned buffers =>
`owned_buffers_io` module and children.
- Make layer file download sensitive to `io_engine::get()`, using
VirtualFile + above copy loop
- For this, I had to move some code into the `retry_download`, e.g.,
`sync_all()` call.
Drive-by:
- fix missing escaping in `scripts/ps_ec2_setup_instance_store`
- if we failed in retry_download to create a file, we'd try to remove
it, encounter `NotFound`, and `abort()` the process using
`on_fatal_io_error`. This PR adds treats `NotFound` as a success.
# Testing
Functional
- The copy loop is generic & unit tested.
Performance
- Used the `ondemand-download-churn` benchmark to manually test against
real S3.
- Results (public Notion page):
https://neondatabase.notion.site/Benchmarking-tokio-epoll-uring-on-demand-downloads-2024-04-15-newer-code-03c0fdc475c54492b44d9627b6e4e710?pvs=4
- Performance is equivalent at low concurrency. Jumpier situation at
high concurrency, but, still less CPU / throughput with
tokio-epoll-uring.
- It’s a win.
# Future Work
Turn the manual performance testing described in the above results
document into a performance regression test:
https://github.com/neondatabase/neon/issues/7146
## Problem
Tenant deletion had a couple of TODOs where we weren't using proper
cancellation tokens that would have aborted the deletions during process
shutdown.
## Summary of changes
- Refactor enough that deletion/shutdown code has access to the
TenantManager's cancellation toke
- Use that cancellation token in tenant deletion instead of dummy
tokens.
fixes https://github.com/neondatabase/neon/issues/7116
Changes:
- refactor PageServerConfigBuilder: support not-set values
- implement runtime feature test
- use runtime feature test to determine `virtual_file_io_engine` if not
explicitly configured in the config
- log the effective engine at startup
- drive-by: improve assertion messages in `test_pageserver_init_node_id`
This needed a tiny bit of tokio-epoll-uring work, hence bumping it.
Changelog:
```
git log --no-decorate --oneline --reverse 868d2c42b5d54ca82fead6e8f2f233b69a540d3e..342ddd197a060a8354e8f11f4d12994419fff939
c7a74c6 Bump mio from 0.8.8 to 0.8.11
4df3466 Bump mio from 0.8.8 to 0.8.11 (#47)
342ddd1 lifecycle: expose `LaunchResult` enum (#49)
```
## Problem
See:
- https://github.com/neondatabase/neon/issues/6374
## Summary of changes
Whereas previously we calculated synthetic size from the gc_horizon or
the pitr_interval (whichever is the lower LSN), now we ignore gc_horizon
and exclusively start from the `pitr_interval`. This is a more generous
calculation for billing, where we do not charge users for data retained
due to gc_horizon.
These test runs usually take 20-30 minutes. if something hangs, we see
actions proceeding for several hours: it's more convenient to have them
time out sooner so that we notice that something has hung faster.
Aiming for the design where `heavier_once_cell::OnceCell` is initialized
by a future factory lead to awkwardness with how
`LayerInner::get_or_maybe_download` looks right now with the `loop`. The
loop helps with two situations:
- an eviction has been scheduled but has not yet happened, and a read
access should cancel the eviction
- a previous `LayerInner::get_or_maybe_download` that canceled a pending
eviction was canceled leaving the `heavier_once_cell::OnceCell`
uninitialized but needing repair by the next
`LayerInner::get_or_maybe_download`
By instead supporting detached initialization in
`heavier_once_cell::OnceCell` via an `OnceCell::get_or_detached_init`,
we can fix what the monolithic #7030 does:
- spawned off download task initializes the
`heavier_once_cell::OnceCell` regardless of the download starter being
canceled
- a canceled `LayerInner::get_or_maybe_download` no longer stops
eviction but can win it if not canceled
Split off from #7030.
Cc: #5331
Split off from #7030:
- each early exit is counted as canceled init, even though it most
likely was just `LayerInner::keep_resident` doing the no-download repair
check
- `downloaded_after` could had been accounted for multiple times, and
also when repairing to match on-disk state
Cc: #5331
Switched the order; doing https://github.com/neondatabase/neon/pull/6139
first then can remove uninit marker after.
## Problem
Previously, existence of a timeline directory was treated as evidence of
the timeline's logical existence. That is no longer the case since we
treat remote storage as the source of truth on each startup: we can
therefore do without this mark file.
The mark file had also been used as a pseudo-lock to guard against
concurrent creations of the same TimelineId -- now that persistence is
no longer required, this is a bit unwieldy.
In #6139 the `Tenant::timelines_creating` was added to protect against
concurrent creations on the same TimelineId, making the uninit mark file
entirely redundant.
## Summary of changes
- Code that writes & reads mark file is removed
- Some nearby `pub` definitions are amended to `pub(crate)`
- `test_duplicate_creation` is added to demonstrate that mutual
exclusion of creations still works.
## Problem
These fields were only optional for the convenience of the `local_fs`
test helper -- real remote storage backends provide them. It complicated
any code that actually wanted to use them for anything.
## Summary of changes
- Make these fields non-optional
- For azure/S3 it is an error if the server doesn't provide them
- For local_fs, use random strings as etags and the file's mtime for
last_modified.
We need to shard our Tenants to support larger databases without those
large databases dominating our pageservers and/or requiring dedicated
pageservers.
This RFC aims to define an initial capability that will permit creating
large-capacity databases using a static configuration
defined at time of Tenant creation.
Online re-sharding is deferred as future work, as is offloading layers
for historical reads. However, both of these capabilities would be
implementable without further changes to the control plane or compute:
this RFC aims to define the cross-component work needed to bootstrap
sharding end-to-end.
## Problem
We have no regression tests for websocket flow
## Summary of changes
Add a hacky implementation of the postgres protocol over websockets just
to verify the protocol behaviour does not regress over time.
This pull request disables neon extension auto upgrade to help the next
compute image upgrade smooth.
## Summary of changes
We have two places to auto-upgrade neon extension: during compute spec
update, and when the compute node starts. The compute spec update logic
is always there, and the compute node start logic is added in
https://github.com/neondatabase/neon/pull/7029. In this pull request, we
disable both of them, so that we can still roll back to an older version
of compute before figuring out the best way of extension
upgrade-downgrade. https://github.com/neondatabase/neon/issues/6936
We will enable auto-upgrade in the next release following this release.
There are no other extension upgrades from release 4917 and therefore
after this pull request, it would be safe to revert to release 4917.
Impact:
* Project created after unpinning the compute image -> if we need to
roll back, **they will stuck**, because the default neon extension
version is 1.3. Need to manually pin the compute image version if such
things happen.
* Projects already stuck on staging due to not downgradeable -> I don't
know their current status, maybe they are already running the latest
compute image?
* Other projects -> can be rolled back to release 4917.
Signed-off-by: Alex Chi Z <chi@neon.tech>
We have been using #5681 for quite some time, and at least since #6931
the tests have assumed `cargo-nextest` to work around our use of global
statics. Unlike the `cargo test`, the `cargo nextest run` runs each test
as a separate process that can be timeouted.
Add a mention of using `cargo-nextest` in the top-level README.md.
Sub-crates can still declare they support `cargo test`, like
`compute_tools/README.md` does.
A node with a bad DNS configuration can register itself with the storage
controller, and the controller will try and schedule work onto the node,
but never succeed because it can't reach the node.
The DNS case is a special case of asymmetric network issues. The general
case isn't covered here -- but might make sense to tighten up after
#6844 merges -- then we can avoid assuming a node is immediately
available in re_attach.
## Problem
If a pageserver was offline when the storage controller started, there
was no mechanism to update the
storage controller state when the pageserver becomes active.
## Summary of changes
* Add a heartbeater module. The heartbeater must be driven by an
external loop.
* Integrate the heartbeater into the service.
- Extend the types used by the service and scheduler to keep track of a
nodes' utilisation score.
- Add a background loop to drive the heartbeater and update the state
based on the deltas it generated
- Do an initial round of heartbeats at start-up
# Problem
While investigating #7124, I noticed that the benchmark was always using
the `DEFAULT_*` `virtual_file_io_engine` , i.e., `tokio-epoll-uring` as
of https://github.com/neondatabase/neon/pull/7077.
The fundamental problem is that the `control_plane` code has its own
view of `PageServerConfig`, which, I believe, will always be a subset of
the real pageserver's `pageserver/src/config.rs`.
For the `virtual_file_io_engine` and `get_vectored_impl` parametrization
of the test suite, we were constructing a dict on the Python side that
contained these parameters, then handed it to
`control_plane::PageServerConfig`'s derived `serde::Deserialize`.
The default in serde is to ignore unknown fields, so, the Deserialize
impl silently ignored the fields.
In consequence, the fields weren't propagated to the `pageserver --init`
call, and the tests ended up using the
`pageserver/src/config.rs::DEFAULT_` values for the respective options
all the time.
Tests that explicitly used overrides in `env.pageserver.start()` and
similar were not affected by this.
But, it means that all the test suite runs where with parametrization
didn't properly exercise the code path.
# Changes
- use `serde(deny_unknown_fields)` to expose the problem
- With this change, the Python tests that override
`virtual_file_io_engine` and
`get_vectored_impl` fail on `pageserver --init`, exposing the problem.
- use destructuring to uncover the issue in the future
- fix the issue by adding the missing fields to the `control_plane`
crate's `PageServerConf`
- A better solution would be for control plane to re-use a struct
provided
by the pageserver crate, so that everything is in one place in
`pageserver/src/config.rs`, but, our config parsing code is (almost)
beyond repair anyways.
- fix the `pageserver_virtual_file_io_engine` to be responsive to the
env var
- => required to make parametrization work in benchmarks
# Testing
Before merging this PR, I re-ran the regression tests & CI with the full
matrix of `virtual_file_io_engine` and `tokio-epoll-uring`, see
9c7ea364e0
## Problem
Shard splits worked, but weren't safe against failures (e.g. node crash
during split) yet.
Related: #6676
## Summary of changes
- Introduce async rwlocks at the scope of Tenant and Node:
- exclusive tenant lock is used to protect splits
- exclusive node lock is used to protect new reconciliation process that
happens when setting node active
- exclusive locks used in both cases when doing persistent updates (e.g.
node scheduling conf) where the update to DB & in-memory state needs to
be atomic.
- Add failpoints to shard splitting in control plane and pageserver
code.
- Implement error handling in control plane for shard splits: this
detaches child chards and ensures parent shards are re-attached.
- Crash-safety for storage controller restarts requires little effort:
we already reconcile with nodes over a storage controller restart, so as
long as we reset any incomplete splits in the DB on restart (added in
this PR), things are implicitly cleaned up.
- Implement reconciliation with offline nodes before they transition to
active:
- (in this context reconciliation means something like
startup_reconcile, not literally the Reconciler)
- This covers cases where split abort cannot reach a node to clean it
up: the cleanup will eventually happen when the node is marked active,
as part of reconciliation.
- This also covers the case where a node was unavailable when the
storage controller started, but becomes available later: previously this
allowed it to skip the startup reconcile.
- Storage controller now terminates on panics. We only use panics for
true "should never happen" assertions, and these cases can leave us in
an un-usable state if we keep running (e.g. panicking in a shard split).
In the unlikely event that we get into a crashloop as a result, we'll
rely on kubernetes to back us off.
- Add `test_sharding_split_failures` which exercises a variety of
failure cases during shard split.
## Problem
hyper auto-cancels the request futures on connection close.
`sql_over_http::handle` is not 'drop cancel safe', so we need to do some
other work to make sure connections are queries in the right way.
## Summary of changes
1. tokio::spawn the request handler to resolve the initial cancel-safety
issue
2. share a cancellation token, and cancel it when the request `Service`
is dropped.
3. Add a new log span to be able to track the HTTP connection lifecycle.
## Problem
Before this PR, `Timeline::get_vectored` would be throttled twice if the
sequential option was enabled or if validation was enabled.
Also, `pageserver_get_vectored_seconds` included the time spent in the
throttle, which turns out to be undesirable for what we use that metric
for.
## Summary of changes
Double-throttle:
* Add `Timeline::get0` method which is unthrottled.
* Use that method from within the `Timeline::get_vectored` code path.
Metric:
* return throttled time from `throttle()` method
* deduct the value from the observed time
* globally rate-limited logging of duration subtraction errors, like in
all other places that do the throttled-time deduction from observations
The `tenant_id` in `TenantLocationConfigRequest` in the
`location_config` endpoint was only used in the storage
controller/attachment service, and there it was only used for assertions
and the creation part.
Currently, the flushing operation could flush multiple frozen layers to
the disk and store the aggregate time in the histogram. The result is a
bimodal distribution with short and over 1000-second flushes. Change it
so that we record how long one layer flush takes.
## Problem
Currently cplane communication is a part of the latency monitoring. It
doesn't allow to setup the proper alerting based on proxy latency.
## Summary of changes
Added dimension to exclude cplane latency.
## Problem
* quotes in serialized string
* no status if connection is from local cache
## Summary of changes
* remove quotes
* report warm if connection if from local cache
## Problem
Missing error classification for SQL-over-HTTP queries.
Not respecting `UserFacingError` for SQL-over-HTTP queries.
## Summary of changes
Adds error classification.
Adds user facing errors.
## Summary
- Currently we can set stripe size at tenant creation, but it doesn't
mean anything until we have multiple shards
- When onboarding an existing tenant, it will always get a default shard
stripe size, so we would like to be able to pick the actual stripe size
at the point we split.
## Why do this inline with a split?
The alternative to this change would be to have a separate endpoint on
the storage controller for setting the stripe size on a tenant, and only
permit writes to that endpoint when the tenant has only a single shard.
That would work, but be a little bit more work for a client, and not
appreciably simpler (instead of having a special argument to the split
functions, we'd have a special separate endpoint, and a requirement that
the controller must sync its config down to the pageserver before
calling the split API). Either approach would work, but this one feels a
bit more robust end-to-end: the split API is the _very last moment_ that
the stripe size is mutable, so if we aim to set it before splitting, it
makes sense to do it as part of the same operation.
## Summary of changes
The problem it fixes is when `request_lsn` is `u64::MAX-1` the
`cont_lsn` becomes `u64::MAX` which is the same as `prev_lsn` which
stops the loop.
Closes https://github.com/neondatabase/neon/issues/6812
## Problem
Currently we manually register nodes with the storage controller, and
use a script during deploy to register with the cloud control plane.
Rather than extend that script further, nodes should just register on
startup.
## Summary of changes
- Extend the re-attach request to include an optional
NodeRegisterRequest
- If the `register` field is set, handle it like a normal node
registration before executing the normal re-attach work.
- Update tests/neon_local that used to rely on doing an explicit
register step that could be enabled/disabled.
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
Now that we have tls-listener vendored, we can refactor and remove a lot
of bloated code and make the whole flow a bit simpler
## Summary of changes
1. Remove dead code
2. Move the error handling to inside the `TlsListener` accept() function
3. Extract the peer_addr from the PROXY protocol header and log it with
errors
Previously we aggregated ps_feedback on each safekeeper and sent it to
walproposer with every AppendResponse. This PR changes it to send
ps_feedback to walproposer right after receiving it from pageserver,
without aggregating it in memory. Also contains some preparations for
implementing backpressure support for sharding.
## Problem
On HTTP query timeout, we should try and cancel the current in-flight
SQL query.
## Summary of changes
Trigger a cancellation command in postgres once the timeout is reach
Not a user-facing change, but can break any existing `.neon` directories
created by neon_local, as the name of the database used by the storage
controller changes.
This PR changes all the locations apart from the path of
`control_plane/attachment_service` (waiting for an opportune moment to
do that one, because it's the most conflict-ish wrt ongoing PRs like
#6676 )
This test occasionally fails with a difference in "pg_xact/0000" file
between the local and restored datadirs. My hypothesis is that
something changed in the database between the last explicit checkpoint
and the shutdown. I suspect autovacuum, it could certainly create
transactions.
To fix, be more precise about the point in time that we compare. Shut
down the endpoint first, then read the last LSN (i.e. the shutdown
checkpoint's LSN), from the local disk with pg_controldata. And use
exactly that LSN in the basebackup.
Closes#559
The walproposer pretends to be a walsender in many ways. It has a
WalSnd slot, it claims to be a walsender by calling
MarkPostmasterChildWalSender() etc. But one different to real
walsenders was that the postmaster still treated it as a bgworker
rather than a walsender. The difference is that at shutdown,
walsenders are not killed until the very end, after the checkpointer
process has written the shutdown checkpoint and exited.
As a result, the walproposer always got killed before the shutdown
checkpoint was written, so the shutdown checkpoint never made it to
safekeepers. That's fine in principle, we don't require a clean
shutdown after all. But it also feels a bit silly not to stream the
shutdown checkpoint. It could be useful for initializing hot standby
mode in a read replica, for example.
Change postmaster to treat background workers that have called
MarkPostmasterChildWalSender() as walsenders. That unfortunately
requires another small change in postgres core.
After doing that, walproposers stay alive longer. However, it also
means that the checkpointer will wait for the walproposer to switch to
WALSNDSTATE_STOPPING state, when the checkpointer sends the
PROCSIG_WALSND_INIT_STOPPING signal. We don't have the machinery in
walproposer to receive and handle that signal reliably. Instead, we
mark walproposer as being in WALSNDSTATE_STOPPING always.
In commit 568f91420a, I assumed that shutdown will wait for all the
remaining WAL to be streamed to safekeepers, but before this commit
that was not true, and the test became flaky. This should make it
stable again.
Some tests wrongly assumed that no WAL could have been written between
pg_current_wal_flush_lsn and quick pg stop after it. Fix them by introducing
flush_ep_to_pageserver which first stops the endpoint and then waits till all
committed WAL reaches the pageserver.
In passing extract safekeeper http client to its own module.
## Problem
Returning from PG_TRY is a bug, and we currently do that
## Summary of changes
Make it break and then return false. This should also help stabilize
test_bad_connection.py
This is a follow-up to #7051 where `LayerInner::drop` and
`LayerInner::evict_blocking` were not noticed to require a gate before
the file deletion. The lack of entering a gate opens up a similar
possibility of deleting a layer file which a newer Timeline instance has
already checked out to be resident in a similar case as #7051.
Tenant::shutdown or Timeline::shutdown completes and becomes externally
observable before the corresponding Tenant/Timeline object is dropped.
For example, after observing a Tenant::shutdown to complete, we could
attach the same tenant_id again. The shut down Tenant object might still
be around at the time of the attach.
The race is then the following:
- old object's metrics are still around
- new object uses with_label_values
- old object calls remove_label_values
The outcome is that the new object will have the metric objects (they're
an Arc internall) but the metrics won't be part of the internal registry
and hence they'll be missing in `/metrics`.
Later, when the new object gets shut down and tries to
remove_label_value, it will observe an error because
the metric was already removed by the old object.
Changes
-------
This PR moves metric removal to `shutdown()`.
An alternative design would be to multi-version the metrics using a
distinguishing label, or, to use a better metrics crate that allows
removing metrics from the registry through the locally held metric
handle instead of interacting with the (globally shared) registry.
refs https://github.com/neondatabase/neon/pull/7051
All of production is using it now as of
https://github.com/neondatabase/aws/pull/1121
The change in `flaky_tests.py` resets the flakiness detection logic.
The alternative would have been to repeat the choice of io engine in
each test name, which would junk up the various test reports too much.
---------
Co-authored-by: Alexander Bayandin <alexander@neon.tech>
result_tx and compute_hook were in ServiceState (i.e. behind a sync
mutex), but didn't need to be.
Moving them up into Service removes a bunch of boilerplate clones.
While we're here, create a helper `Service::maybe_reconcile_shard` which
avoids writing out all the `&self.` arguments to
`TenantState::maybe_reconcile` everywhere we call it.
Otherwise, it might happen that we never get to witness the same state
on subsequent restarts, thus the time series will show the value from a
few restarts ago.
The actual case here was that "Activating" was showing `3` while I was
doing tenant migration testing on staging. The number 3 was however from
a startup that happened some time ago which had been interrupted by
another deployment.
## Problem
`422 Unprocessable Entity: compute time quota of non-primary branches is
exceeded` being marked as a control plane error.
## Summary of changes
Add the manual checks to make this a user error that should not be
retried.
## Problem
Before this PR, it was possible that on-demand downloads were started
after `Timeline::shutdown()`.
For example, we have observed a walreceiver-connection-handler-initiated
on-demand download that was started after `Timeline::shutdown()`s final
`task_mgr::shutdown_tasks()` call.
The underlying issue is that `task_mgr::shutdown_tasks()` isn't sticky,
i.e., new tasks can be spawned during or after
`task_mgr::shutdown_tasks()`.
Cc: https://github.com/neondatabase/neon/issues/4175 in lieu of a more
specific issue for task_mgr. We already decided we want to get rid of it
anyways.
Original investigation:
https://neondb.slack.com/archives/C033RQ5SPDH/p1709824952465949
## Changes
- enter gate while downloading
- use timeline cancellation token for cancelling download
thereby, fixes#7054
Entering the gate might also remove recent "kept the gate from closing"
in staging.
## Problem
When we start compute with newer version of extension (i.e. 1.2) and
then rollback the release, downgrading the compute version, next compute
start will try to update extension to the latest version available in
neon.control (i.e. 1.1).
Thus we need to provide downgrade scripts like neon--1.2--1.1.sql
These scripts must revert the changes made by the upgrade scripts in the
reverse order. This is necessary to ensure that the next upgrade will
work correctly.
In general, we need to write upgrade and downgrade scripts to be more
robust and add IF EXISTS / CREATE OR REPLACE clauses to all statements
(where applicable).
## Summary of changes
Adds downgrade scripts.
Adds test cases for extension downgrade/upgrade.
fixes#7066
This is a follow-up for
https://app.incident.io/neondb/incidents/167?tab=follow-ups
Signed-off-by: Alex Chi Z <chi@neon.tech>
Co-authored-by: Alex Chi Z <iskyzh@gmail.com>
Co-authored-by: Anastasia Lubennikova <anastasia@neon.tech>
## Problem
Tenants created via the storage controller have a `PlacementPolicy` that
defines their HA/secondary/detach intent. For backward compat we can
just set it to Single, for onboarding tenants using /location_conf it is
automatically set to Double(1) if there are at least two pageservers,
but for freshly created tenants we didn't have a way to specify it.
This unblocks writing tests that create HA tenants on the storage
controller and do failure injection testing.
## Summary of changes
- Add optional fields to TenantCreateRequest for specifying
PlacementPolicy. This request structure is used both on pageserver API
and storage controller API, but this method is only meaningful for the
storage controller (same as existing `shard_parameters` attribute).
- Use the value from the creation request in tenant creation, if
provided.
## Problem
For the ephemeral endpoint feature, it's not really too helpful to keep
them around in the connection pool. This isn't really pressing but I
think it's still a bit better this way.
## Summary of changes
Add `is_ephemeral` function to `NeonOptions`. Allow
`serverless::ConnInfo::endpoint_cache_key()` to return an `Option`.
Handle that option appropriately
## Problem
We reverted https://github.com/neondatabase/neon/pull/6661 a few days
ago. The change led to OOMs in
benchmarks followed by large WAL reingests.
The issue was that we removed [this
code](d04af08567/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs (L409-L417)).
That call may trigger a roll of the open layer due to
the keepalive messages received from the safekeeper. Removing it meant
that enforcing
of checkpoint timeout became even more lax and led to using up large
amounts of memory
for the in memory layer indices.
## Summary of changes
Piggyback on keep alive messages to enforce checkpoint timeout. This is
a hack, but it's exactly what
the current code is doing.
## Alternatives
Christhian, Joonas and myself sketched out a timer based approach
[here](https://github.com/neondatabase/neon/pull/6940). While discussing
it further, it became obvious that's also a bit of a hack and not the
desired end state. I chose not
to take that further since it's not what we ultimately want and it'll be
harder to rip out.
Right now it's unclear what the ideal system behaviour is:
* early flushing on memory pressure, or ...
* detaching tenants on memory pressure
## Problem
It seems that even though we have a retry on basebackup, it still
sometimes fails to fetch it with the failpoint enabled, resulting in a
test error.
## Summary of changes
If we fail to get the basebackup, disable the failpoint and try again.
## Problem
Closes: https://github.com/neondatabase/neon/issues/6847
Closes: https://github.com/neondatabase/neon/issues/7006
## Summary of changes
- Pageserver API calls are wrapped in timeout/retry logic: this prevents
a reconciler getting hung on a pageserver API hang, and prevents
reconcilers having to totally retry if one API call returns a retryable
error (e.g. 503).
- Add a cancellation token to `Node`, so that when we mark a node
offline we will cancel any API calls in progress to that node, and avoid
issuing any more API calls to that offline node.
- If the dirty locations of a shard are all on offline nodes, then don't
spawn a reconciler
- In re-attach, if we have no observed state object for a tenant then
construct one with conf: None (which means "unknown"). Then in
Reconciler, implement a TODO for scanning such locations before running,
so that we will avoid spuriously incrementing a generation in the case
of a node that was offline while we started (this is the case that
tripped up #7006)
- Refactoring: make Node contents private (and thereby guarantee that
updates to availability mode reliably update the cancellation token.)
- Refactoring: don't pass the whole map of nodes into Reconciler (and
thereby remove a bunch of .expect() calls)
Some of this was discovered/tested with a new failure injection test
that will come in a separate PR, once it is stable enough for CI.
## Problem
When vectored get encountered a portion of the key range that could
not be mapped to any layer in the current timeline it would incorrectly
bail out of the current timeline. This is incorrect since we may have
had layers queued for a visit in the fringe.
## Summary of changes
* Add a repro unit test
* Remove the early bail out path
* Simplify range search return value
We have a benchmark for creating a lot of branches, but it does random
things, and the branch count is not what we is the largest maximum we
aim to support. If this PR would stabilize the benchmark total duration
it means that there are some structures which are very much slower than
others. Then we should add a seed-outputting variant to help find and
reproduce such cases.
Additionally, record for the benchmark:
- shutdown duration
- startup metrics once done (on restart)
- duration of first compaction completion via debug logging
## Problem
The storage controller binary still has its historic
`attachment_service` name -- it will be painful to change this later
because we can't atomically update this repo and the helm charts used to
deploy.
Companion helm chart change:
https://github.com/neondatabase/helm-charts/pull/70
## Summary of changes
- Change the name of the binary to `storage_controller`
- Skipping renaming things in the source right now: this is just to get
rid of the legacy name in external interfaces.
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
We attempted validation for cancelled errors under the assumption that
if vectored get fails, sequential get will too.
That's not right 100% of times though because sequential get may have
the values cached and slip them through
even when shutting down.
## Summary of changes
Don't validate if either search impl failed due to tenant shutdown.
## Problem
Fix https://github.com/neondatabase/neon/issues/7003. Fix
https://github.com/neondatabase/neon/issues/6982. Currently, neon
extension is only upgraded when new compute spec gets applied, for
example, when creating a new role or creating a new database. This also
resolves `neon.lfc_stat` not found warnings in prod.
## Summary of changes
This pull request adds the logic to spawn a background thread to upgrade
the neon extension version if the compute is a primary. If for whatever
reason the upgrade fails, it reports an error to the console and does
not impact compute node state.
This change can be further applied to 3rd-party extension upgrades. We
can silently upgrade the version of 3rd party extensions in the
background in the future.
Questions:
* Does alter extension takes some kind of lock that will block user
requests?
* Does `ALTER EXTENSION` writes to the database if nothing needs to be
upgraded? (may impact storage size).
Otherwise it's safe to land this pull request.
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
Collection of small changes, batched together to reduce CI overhead.
## Summary of changes
- Layer download messages include size -- this is useful when watching a
pageserver hydrate its on disk cache in the log.
- Controller migrate API could put an invalid NodeId into TenantState
- Scheduling errors during tenant create could result in creating some
shards and not others.
- Consistency check could give hard-to-understand failures in tests if a
reconcile was in process: explicitly fail the check if reconciles are in
progress instead.
## Problem
- The storage controller is the source of truth for a tenant's stripe
size, but doesn't currently have a way to propagate that to compute:
we're just using the default stripe size everywhere.
Closes: https://github.com/neondatabase/neon/issues/6903
## Summary of changes
- Include stripe size in `ComputeHookNotifyRequest`
- Include stripe size in `LocationConfigResponse`
The stripe size is optional: it will only be advertised for
multi-sharded tenants. This enables the controller to defer the choice
of stripe size until we split a tenant for the first time.
## Problem
If large numbers of shards are attached to a pageserver concurrently,
for example after another node fails, it can cause excessive I/O queue
depths due to all the newly attached shards trying to calculate logical
sizes concurrently.
#6907 added the `lazy` flag to handle this.
## Summary of changes
- Use `lazy=true` from all /location_config calls in the storage
controller Reconciler.
Moves some of the (legacy) compaction code to compaction.rs. No
functional changes, just moves of code.
Before, compaction.rs was only for the new tiered compaction mechanism,
now it's for both the old and new mechanisms.
Part of #6768
## Problem
Branch/project and coldStart were not populated to data events.
## Summary of changes
Populate it. Also added logging for the coldstart info.
## Problem
ref https://github.com/neondatabase/neon/issues/6188
## Summary of changes
This pull request fixes `-Wmissing-prototypes` for the neon extension.
Note that (1) the gcc version in CI and macOS is different, therefore
some of the warning does not get reported when developing the neon
extension locally. (2) the CI env variable `COPT = -Werror` does not get
passed into the docker build process, therefore warnings are not treated
as errors on CI.
e62baa9704/.github/workflows/build_and_test.yml (L22)
There will be follow-up pull requests on solving other warnings. By the
way, I did not figure out the default compile parameters in the CI env,
and therefore this pull request is tested by manually adding
`-Wmissing-prototypes` into the `COPT`.
Signed-off-by: Alex Chi Z <chi@neon.tech>
part of https://github.com/neondatabase/neon/issues/5899
Problem
-------
Before this PR, the time spent waiting on the throttle was charged
towards the higher-level page_service metrics, i.e.,
`pageserver_smgr_query_seconds`.
The metrics are the foundation of internal SLIs / SLOs.
A throttled tenant would cause the SLI to degrade / SLO alerts to fire.
Changes
-------
- don't charge time spent in throttle towards the page_service metrics
- record time spent in throttle in RequestContext and subtract it from
the elapsed time
- this works because the page_service path doesn't create child context,
so, all the throttle time is recorded in the parent
- it's quite brittle and will break if we ever decide to spawn child
tasks that need child RequestContexts, which would have separate
instances of the `micros_spent_throttled` counter.
- however, let's punt that to a more general refactoring of
RequestContext
- add a test case that ensures that
- throttling happens for getpage requests; this aspect of the test
passed before this PR
- throttling delays aren't charged towards the page_service metrics;
this aspect of the test only passes with this PR
- drive-by: make the throttle log message `info!`, it's an expected
condition
Performance
-----------
I took the same measurements as in #6706 , no meaningful change in CPU
overhead.
Future Work
-----------
This PR enables us to experiment with the throttle for select tenants
without affecting the SLI metrics / triggering SLO alerts.
Before declaring this feature done, we need more work to happen,
specifically:
- decide on whether we want to retain the flexibility of throttling any
`Timeline::get` call, filtered by TaskKind
- versus: separate throttles for each page_service endpoint, potentially
with separate config options
- the trouble here is that this decision implies changes to the
TenantConfig, so, if we start using the current config style now, then
decide to switch to a different config, it'll be a breaking change
Nice-to-haves but probably not worth the time right now:
- Equivalent tests to ensure the throttle applies to all other
page_service handlers.
## Problem
Last weeks enablement of vectored get generated a number of panics.
From them, I diagnosed two issues in the delta layer index traversal
logic
1. The `key >= range.start && lsn >= lsn_range.start`
was too aggressive. Lsns are not monotonically increasing in the delta
layer index (keys are though), so we cannot assert on them.
2. Lsns greater or equal to `lsn_range.end` were not skipped. This
caused the query to consider records newer than the request Lsn.
## Summary of changes
* Fix the issues mentioned above inline
* Refactor the layer traversal logic to make it unit testable
* Add unit test which reproduces the failure modes listed above.
## Problem
The value reconstruct of AUX_FILES_KEY from records is not deterministic
since it uses a hash map under the hood. This caused vectored get validation
failures when enabled in staging.
## Summary of changes
Deserialise AUX_FILES_KEY blobs comparing. All other keys should
reconstruct deterministically, so we simply compare the blobs.
Before this PR, the layer file download code would fsync the inode after
rename instead of the timeline directory. That is not in line with what
a comment further up says we're doing, and it's obviously not achieving
the goal of making the rename durable.
part of https://github.com/neondatabase/neon/issues/6663
`std` has had `pin!` macro for some time, there is no need for us to use
the older alternatives. Cannot disallow `tokio::pin` because tokio
macros use that.
part of #6663
See that epic for more context & related commits.
Problem
-------
Before this PR, the layer-file-creating code paths were using
VirtualFile, but under the hood these were still blocking system calls.
Generally this meant we'd stall the executor thread, unless the caller
"knew" and used the following pattern instead:
```
spawn_blocking(|| {
Handle::block_on(async {
VirtualFile::....().await;
})
}).await
```
Solution
--------
This PR adopts `tokio-epoll-uring` on the layer-file-creating code paths
in pageserver.
Note that on-demand downloads still use `tokio::fs`, these will be
converted in a future PR.
Design: Avoiding Regressions With `std-fs`
------------------------------------------
If we make the VirtualFile write path truly async using
`tokio-epoll-uring`, should we then remove the `spawn_blocking` +
`Handle::block_on` usage upstack in the same commit?
No, because if we’re still using the `std-fs` io engine, we’d then block
the executor in those places where previously we were protecting us from
that through the `spawn_blocking` .
So, if we want to see benefits from `tokio-epoll-uring` on the write
path while also preserving the ability to switch between
`tokio-epoll-uring` and `std-fs` , where `std-fs` will behave identical
to what we have now, we need to ***conditionally* use `spawn_blocking +
Handle::block_on`** .
I.e., in the places where we use that know, we’ll need to make that
conditional based on the currently configured io engine.
It boils down to investigating all the places where we do
`spawn_blocking(... block_on(... VirtualFile::...))`.
Detailed [write-up of that investigation in
Notion](https://neondatabase.notion.site/Surveying-VirtualFile-write-path-usage-wrt-tokio-epoll-uring-integration-spawn_blocking-Handle-bl-5dc2270dbb764db7b2e60803f375e015?pvs=4
), made publicly accessible.
tl;dr: Preceding PRs addressed the relevant call sites:
- `metadata` file: turns out we could simply remove it (#6777, #6769,
#6775)
- `create_delta_layer()`: made sensitive to `virtual_file_io_engine` in
#6986
NB: once we are switched over to `tokio-epoll-uring` everywhere in
production, we can deprecate `std-fs`; to keep macOS support, we can use
`tokio::fs` instead. That will remove this whole headache.
Code Changes In This PR
-----------------------
- VirtualFile API changes
- `VirtualFile::write_at`
- implement an `ioengine` operation and switch `VirtualFile::write_at`
to it
- `VirtualFile::metadata()`
- curiously, we only use it from the layer writers' `finish()` methods
- introduce a wrapper `Metadata` enum because `std::fs::Metadata` cannot
be constructed by code outside rust std
- `VirtualFile::sync_all()` and for completeness sake, add
`VirtualFile::sync_data()`
Testing & Rollout
-----------------
Before merging this PR, we ran the CI with both io engines.
Additionally, the changes will soak in staging.
We could have a feature gate / add a new io engine
`tokio-epoll-uring-write-path` to do a gradual rollout. However, that's
not part of this PR.
Future Work
-----------
There's still some use of `std::fs` and/or `tokio::fs` for directory
namespace operations, e.g. `std::fs::rename`.
We're not addressing those in this PR, as we'll need to add the support
in tokio-epoll-uring first. Note that rename itself is usually fast if
the directory is in the kernel dentry cache, and only the fsync after
rename is slow. These fsyncs are using tokio-epoll-uring, so, the impact
should be small.
## Problem
Fix https://github.com/neondatabase/neon/issues/6498
## Summary of changes
Only re-authenticate with zenith_admin if authentication fails.
Otherwise, directly return the error message.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
tokio 1.36 has been out for a month.
Release notes don't indicate major changes.
Skimming through their issue tracker, I can't find open `C-bug` issues
that would affect us.
(My personal motivation for this is `JoinSet::try_join_next`.)
As pointed out in the comments added in this PR:
the in-memory state of the filesystem already has the layer file in its
final place.
If the fsync fails, but pageserver continues to execute, it's quite easy
for subsequent pageserver code to observe the file being there and
assume it's durable, when it really isn't.
It can happen that we get ENOSPC during the fsync.
However,
1. the timeline dir is small (remember, the big layer _file_ has already
been synced).
Small data means ENOSPC due to delayed allocation races etc are less
likely.
2. what else are we going to do in that case?
If we decide to bubble up the error, the file remains on disk.
We could try to unlink it and fsync after the unlink.
If that fails, we would _definitely_ need to error out.
Is it worth the trouble though?
Side note: all this logic about not carrying on after fsync failure
implies that we `sync` the filesystem successfully before we restart
the pageserver. We don't do that right now, but should (=>
https://github.com/neondatabase/neon/issues/6989)
part of https://github.com/neondatabase/neon/issues/6663
## Problem
The current implementation of `deploy-prod` workflow doesn't allow to
run parallel deploys on Storage and Proxy.
## Summary of changes
- Call `deploy-proxy-prod` workflow that deploys only Proxy components,
and that can be run in parallel with `deploy-prod` for Storage.
Usually RFC documents are not modified, but the vast mentions of
"zenith" in early RFC documents make it desirable to update the product
name to today's name, to avoid confusion.
## Problem
Early RFC documents use the old "zenith" product name a lot, which is
not something everyone is aware of after the product was renamed.
## Summary of changes
Replace occurrences of "zenith" with "neon".
Images are excluded.
---------
Co-authored-by: Andreas Scherbaum <andreas@neon.tech>
The `writer.finish()` methods already fsync the inode, using
`VirtualFile::sync_all()`.
All that the callers need to do is fsync their directory, i.e., the
timeline directory.
Note that there's a call in the new compaction code that is apparently
dead-at-runtime, so, I couldn't fix up any fsyncs there
[Link](502b69b33b/pageserver/src/tenant/timeline/compaction.rs (L204-L211)).
Note that layer durability still matters somewhat, even after #5198
which made remote storage authoritative.
We do have the layer file length as an indicator, but no checksums on
the layer file contents.
So, a series of overwrites without fsyncs in the middle, plus a
subsequent crash, could cause us to end up in a state where the file
length matches but the contents are garbage.
part of https://github.com/neondatabase/neon/issues/6663
## Problem
Actually it's good idea to distinguish between cases when it's a cold
start, but we took the compute from the pool
## Summary of changes
Updated to enum.
## Problem
- #6966
- Existing logs aren't pointing to a cause: it looks like heatmap upload
and download are happening, but for some reason the evicted layer isn't
removed on the secondary location.
## Summary of changes
- Assert evicted layer is gone from heatmap before checking its gone
from local disk: this will give clarity on whether the issue is with the
uploads or downloads.
- On assertion failures, log the contents of heatmap.
## Problem
- Walredo errors, e.g. during image creation, mention the LSN affected
but not the key.
## Summary of changes
- Add key to "error applying ... WAL records" log message
During onboarding, the control plane may attempt ad-hoc creation of a
secondary location to facilitate live migration. This gives us two
problems to solve:
- Accept 'Secondary' mode in /location_config and use it to put the
tenant into secondary mode on some physical pageserver, then pass
through /tenant/xyz/secondary/download requests
- Create tenants with no generation initially, since the initial
`Secondary` mode call will not provide us a generation.
This PR also fixes modification of a tenant's TenantConf during
/location_conf, which was previously ignored, and refines the flow for
config modification:
- avoid bumping generations when the only reason we're reconciling an
attached location is a config change
- increment TenantState.sequence when spawning a reconciler: usually
schedule() does this, but when we do config changes that doesn't happen,
so without this change waiters would think reconciliation was done
immediately. `sequence` is a bit of a murky thing right now, as it's
dual-purposed for tracking waiters, and for checking if an existing
reconciliation is already making updates to our current sequence. I'll
follow up at some point to clarify it's purpose.
- test config modification at the end of onboarding test
## Problem
At high ingest rates, pageservers spuriously disconnect from safekeepers
because stats updates don't come in frequently enough to keep the
broker/safekeeper LSN delta under the wal lag limit.
## Summary of changes
- Increase DEFAULT_MAX_WALRECEIVER_LSN_WAL_LAG from 10MiB to 1GiB. This
should be enough for realistic per-timeline throughputs.
## Problem
PR #6837 fixed secondary locations to avoid spamming log warnings on
temp files, but we also have ".temp_download" files to consider.
## Summary of changes
- Give temp_download files the same behavior as temp files.
- Refactor the relevant helper to pub(crate) from pub
Nightly has added a bunch of compiler and linter warnings. There is also
two dependencies that fail compilation on latest nightly due to using
the old `stdsimd` feature name. This PR fixes them.
## Problem
https://github.com/neondatabase/neon/pull/6661 changed the layer
flushing logic and led to OOMs in staging.
The issue turned out to be holding on to in-memory layers for too long.
After OOMing we'd need to replay potentially
a lot of WAL.
## Summary of changes
Test that open layers get flushed after the `checkpoint_timeout` config
and do not require WAL reingest upon restart.
The workload creates a number of timelines and writes some data to each,
but not enough to trigger flushes via the `checkpoint_distance` config.
I ran this test against https://github.com/neondatabase/neon/pull/6661
and it was indeed failing.
## Problem
PR https://github.com/neondatabase/neon/pull/6851 implemented new output
in PostgreSQL explain.
this is a test case for the new function.
## Summary of changes
## Checklist before requesting a review
- [x] I have performed a self-review of my code.
- [x] If it is a core feature, I have added thorough tests.
- [no ] Do we need to implement analytics? if so did you add the
relevant metrics to the dashboard?
- [no] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
## Problem
shard_id in span is repeated:
- https://github.com/neondatabase/neon/issues/6723Closes: #6723
## Summary of changes
- Only add shard_id to the span when fetching a cached timeline, as it
is already added when loading an uncached timeline.
Extracted from https://github.com/neondatabase/neon/pull/6953
Part of https://github.com/neondatabase/neon/issues/5899
Core Change
-----------
In #6953, we need the ability to scan the log _after_ a specific line
and ignore anything before that line.
This PR changes `log_contains` to returns a tuple of `(matching line,
cursor)`.
Hand that cursor to a subsequent `log_contains` call to search the log
for the next occurrence of the pattern.
Other Changes
-------------
- Inspect all the callsites of `log_contains` to handle the new tuple
return type.
- Above inspection unveiled many callers aren't using `assert
log_contains(...) is not None` but some weaker version of the code that
breaks if `log_contains` ever returns a not-None but falsy value. Fix
that.
- Above changes unveiled that `test_remote_storage_upload_queue_retries`
was using `wait_until` incorrectly; after fixing the usage, I had to
raise the `wait_until` timeout. So, maybe this will fix its flakiness.
Because of bugs evictions could hang and pause disk usage eviction task.
One such bug is known and fixed#6928. Guard each layer eviction with a
modest timeout deeming timeouted evictions as failures, to be
conservative.
In addition, add logging and metrics recording on each eviction
iteration:
- log collection completed with duration and amount of layers
- per tenant collection time is observed in a new histogram
- per tenant layer count is observed in a new histogram
- record metric for collected, selected and evicted layer counts
- log if eviction takes more than 10s
- log eviction completion with eviction duration
Additionally remove dead code for which no dead code warnings appeared
in earlier PR.
Follow-up to: #6060.
## Summary of changes
Calculate number of unique page accesses at compute.
It can be used to estimate working set size and adjust cache size
(shared_buffers or local file cache).
Approximation is made using HyperLogLog algorithm.
It is performed by local file cache and so is available only when local
file cache is enabled.
This calculation doesn't take in account access to the pages present in
shared buffers, but includes pages available in local file cache.
This information can be retrieved using
approximate_working_set_size(reset bool) function from neon extension.
reset parameter can be used to reset statistic and so collect unique
accesses for the particular interval.
Below is an example of estimating working set size after pgbench -c 10
-S -T 100 -s 10:
```
postgres=# select approximate_working_set_size(false);
approximate_working_set_size
------------------------------
19052
(1 row)
postgres=# select pg_table_size('pgbench_accounts')/8192;
?column?
----------
16402
(1 row)
```
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
Add off-by-default support for lazy queued tenant activation on attach.
This should be useful on bulk migrations as some tenants will be
activated faster due to operations or endpoint startup. Eventually all
tenants will get activated by reusing the same mechanism we have at
startup (`PageserverConf::concurrent_tenant_warmup`).
The difference to lazy attached tenants to startup ones is that we leave
their initial logical size calculation be triggered by WalReceiver or
consumption metrics.
Fixes: #6315
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
Sometimes folks prefer not to expose secrets as CLI args.
## Summary of changes
- Add ability to load secrets from environment variables.
We can eventually remove the AWS SM code path here if nobody is using it
-- we don't need to maintain three ways to load secrets.
## Problem
We build compute-tools binary twice — in `compute-node` and in
`compute-tools` jobs, and we build them slightly differently:
- `cargo build --locked --profile release-line-debug-size-lto`
(previously in `compute-node`)
- `mold -run cargo build -p compute_tools --locked --release`
(previously in `compute-tools`)
Before:
- compute-node: **6m 34s**
- compute-tools (as a separate job): **7m 47s**
After:
- compute-node: **7m 34s**
- compute-tools (as a separate step, within compute-node job): **5s**
## Summary of changes
- Move compute-tools image creation to `Dockerfile.compute-node`
- Delete `Dockerfile.compute-tools`
## Problem
Callers of the timeline creation API may issue timeline GETs ahead of
creation to e.g. check if their intended timeline already exists, or to
learn the LSN of a parent timeline.
Although the timeline creation API already triggers activation of a
timeline if it's currently waiting to activate, the GET endpoint
doesn't, so such callers will encounter 503 responses for several
minutes after a pageserver restarts, while tenants are lazily warming
up.
The original scope of which APIs will activate a timeline was quite
small, but really it makes sense to do it for any API that needs a
particular timeline to be active.
## Summary of changes
- In the timeline detail GET handler, use wait_to_become_active, which
triggers immediate activation of a tenant if it was currently waiting
for the warmup semaphore, then waits up to 5 seconds for the activation
to complete. If it doesn't complete promptly, we return a 503 as before.
- Modify active_timeline_for_active_tenant to also use
wait_to_become_active, which indirectly makes several other
timeline-scope request handlers fast-activate a tenant when called. This
is important because a timeline creation flow could also use e.g.
get_lsn_for_timestamp as a precursor to creating a timeline.
- There is some risk to this change: an excessive number of timeline GET
requests could cause too many tenant activations to happen at the same
time, leading to excessive queue depth to the S3 client. However, this
was already the case for e.g. many concurrent timeline creations.
## Problem
`pin-build-tools-image` job doesn't have access to secrets and thus
fails. Missed in the original PR[0]
- [0] https://github.com/neondatabase/neon/pull/6795
## Summary of changes
- pass secrets to `pin-build-tools-image` job
## Problem
The "z" and "y" letters are switched on the English keyboard, and I'm
used to a German keyboard. Very embarrassing.
## Summary of changes
Fix syntax error in README
Co-authored-by: Andreas Scherbaum <andreas@neon.tech>
## Problem
Hard to find error reasons by endpoint for HTTP flow.
## Summary of changes
I want all root spans to have session id and endpoint id. I want all
root spans to be consistent.
## Problem
Currently, after updating `Dockerfile.build-tools` in a PR, it requires
a manual action to make it `pinned`, i.e., the default for everyone. It
also makes all opened PRs use such images (even created in the PR and
without such changes).
This PR overhauls the way we build and use `build-tools` image (and uses
the image from Docker Hub).
## Summary of changes
- The `neondatabase/build-tools` image gets tagged with the latest
commit sha for the `Dockerfile.build-tools` file
- Each PR calculates the tag for `neondatabase/build-tools`, tries to
pull it, and rebuilds the image with such tag if it doesn't exist.
- Use `neondatabase/build-tools` as a default image
- When running on `main` branch — create a `pinned` tag and push it to
ECR
- Use `concurrency` to ensure we don't build `build-tools` image for the
same commit in parallel from different PRs
## Problem
The vectored read path proposed in
https://github.com/neondatabase/neon/pull/6576 seems
to be functionally correct, but in my testing (see below) it is about 10-20% slower than the naive
sequential vectored implementation.
## Summary of changes
There's three parts to this PR:
1. Supporting vectored blob reads. This is actually trickier than it
sounds because on disk blobs are prefixed with a variable length size header.
Since the blobs are not necessarily fixed size, we need to juggle the offsets
such that the callers can retrieve the blobs from the resulting buffer.
2. Merge disk read requests issued by the vectored read path up to a
maximum size. Again, the merging is complicated by the fact that blobs
are not fixed size. We keep track of the begin and end offset of each blob
and pass them into the vectored blob reader. In turn, the reader will return
a buffer and the offsets at which the blobs begin and end.
3. A benchmark for basebackup requests against tenant with large SLRU
block counts is added. This required a small change to pagebench and a new config
variable for the pageserver which toggles the vectored get validation.
We can probably optimise things further by adding a little bit of
concurrency for our IO. In principle, it's as simple as spawning a task which deals with issuing
IO and doing the serialisation and handling on the parent task which receives input via a
channel.
This reverts commits 587cb705b8 (PR #6661)
and fcbe9fb184 (PR #6842).
Conflicts:
pageserver/src/tenant.rs
pageserver/src/tenant/timeline.rs
The conflicts were with
* pageserver: adjust checkpoint distance for sharded tenants (#6852)
* pageserver: add vectored get implementation (#6576)
Also we had to keep the `allowed_errors` to make `test_forward_compatibility` happy,
see the PR thread on GitHub for details.
## Problem
Starting up the pageserver before the storage controller is ready can
lead
to a round of reconciliation, which leads to the previous tenant being
shut down.
This disturbs some tests.
## Summary of changes
Wait for the storage controller to become ready on neon env start-up.
Closes https://github.com/neondatabase/neon/issues/6724
Not allowing evicting wanted deleted layers is something I've forgotten
to implement on #5645. This PR makes it possible to evict such layers,
which should reduce the amount of hanging evictions.
Fixes: #6928
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
After commit [840abe3954] (store AUX files
as deltas) we avoid quadratic growth of storage size when storing LR
snapshots but get quadratic slowdown of reconstruct time.
As a result storing 70k snapshots at my local Neon instance took more
than 3 hours and starting node (creation of basecbackup): ~10 minutes.
In prod 70k AUX files cause increase of startup time to 40 minutes:
https://neondb.slack.com/archives/C03F5SM1N02/p1708513010480179
## Summary of changes
Enforce storing full AUX directory (some analog of FPI) each 1024 files.
Time of creation 70k snapshots is reduced to 6 minutes and startup time
- to 1.5 minutes (100 seconds).
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
This is a precursor to adding a convenience CLI for the storage
controller.
## Summary of changes
- move controller api structs into pageserver_api::controller_api to
make them visible to other crates
- rename pageserver_api::control_api to pageserver_api::upcall_api to
match the /upcall/v1/ naming in the storage controller.
Why here rather than a totally separate crate? It's convenient to have
all the pageserver-related stuff in one place, and if we ever wanted to
move it to a different crate it's super easy to do that later.
Rebased version of #5234, part of #6768
This consists of three parts:
1. A refactoring and new contract for implementing and testing
compaction.
The logic is now in a separate crate, with no dependency on the
'pageserver' crate. It defines an interface that the real pageserver
must implement, in order to call the compaction algorithm. The interface
models things like delta and image layers, but just the parts that the
compaction algorithm needs to make decisions. That makes it easier unit
test the algorithm and experiment with different implementations.
I did not convert the current code to the new abstraction, however. When
compaction algorithm is set to "Legacy", we just use the old code. It
might be worthwhile to convert the old code to the new abstraction, so
that we can compare the behavior of the new algorithm against the old
one, using the same simulated cases. If we do that, have to be careful
that the converted code really is equivalent to the old.
This inclues only trivial changes to the main pageserver code. All the
new code is behind a tenant config option. So this should be pretty safe
to merge, even if the new implementation is buggy, as long as we don't
enable it.
2. A new compaction algorithm, implemented using the new abstraction.
The new algorithm is tiered compaction. It is inspired by the PoC at PR
#4539, although I did not use that code directly, as I needed the new
implementation to fit the new abstraction. The algorithm here is less
advanced, I did not implement partial image layers, for example. I
wanted to keep it simple on purpose, so that as we add bells and
whistles, we can see the effects using the included simulator.
One difference to #4539 and your typical LSM tree implementations is how
we keep track of the LSM tree levels. This PR doesn't have a permanent
concept of a level, tier or sorted run at all. There are just delta and
image layers. However, when compaction starts, we look at the layers
that exist, and arrange them into levels, depending on their shapes.
That is ephemeral: when the compaction finishes, we forget that
information. This allows the new algorithm to work without any extra
bookkeeping. That makes it easier to transition from the old algorithm
to new, and back again.
There is just a new tenant config option to choose the compaction
algorithm. The default is "Legacy", meaning the current algorithm in
'main'. If you set it to "Tiered", the new algorithm is used.
3. A simulator, which implements the new abstraction.
The simulator can be used to analyze write and storage amplification,
without running a test with the full pageserver. It can also draw an SVG
animation of the simulation, to visualize how layers are created and
deleted.
To run the simulator:
cargo run --bin compaction-simulator run-suite
---------
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
## Problem
## Summary of changes
Updates the neon.tech link to point to a /github page in order to
correctly attribute visits originating from the repo.
## Problem
Data team cannot distinguish between cold start and not cold start.
## Summary of changes
Report `is_cold_start` to analytics.
---------
Co-authored-by: Conrad Ludgate <conrad@neon.tech>
Noticed that we are failing to handle `Result::Err` when entering a gate
for logical size calculation. Audited rest of the gate enters, which
seem fine, unified two instances.
Noticed that the gate guard allows to remove a failpoint, then noticed
that adjacent failpoint was blocking the executor thread instead of
using `pausable_failpoint!`, fix both.
eviction_task.rs now maintains a gate guard as well.
Cc: #4733
## Problem
We want to show connection counts to console users.
## Summary of changes
Start exporting connection counts grouped by database name and
connection state.
## Problem
LFC has high impact on Neon application performance but there is no way
for user to check efficiency of its usage
## Summary of changes
Show LFC statistic in EXPLAIN ANALYZE
## Description
**Local file cache (LFC)**
A layer of caching that stores frequently accessed data from the storage
layer in the local memory of the Neon compute instance. This cache helps
to reduce latency and improve query performance by minimizing the need
to fetch data from the storage layer repeatedly.
**Externalization of LFC in explain output**
Then EXPLAIN ANALYZE output is extended to display important counts for
local file cache (LFC) hits and misses.
This works both, for EXPLAIN text and json output.
**File cache: hits**
Whenever the Postgres backend retrieves a page/block from SGMR, it is
not found in shared buffer but the page is already found in the LFC this
counter is incremented.
**File cache: misses**
Whenever the Postgres backend retrieves a page/block from SGMR, it is
not found in shared buffer and also not in then LFC but the page is
retrieved from Neon storage (page server) this counter is incremented.
Example (for explain text output)
```sql
explain (analyze,buffers,prefetch,filecache) select count(*) from pgbench_accounts;
QUERY PLAN
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Finalize Aggregate (cost=214486.94..214486.95 rows=1 width=8) (actual time=5195.378..5196.034 rows=1 loops=1)
Buffers: shared hit=178875 read=143691 dirtied=128597 written=127346
Prefetch: hits=0 misses=1865 expired=0 duplicates=0
File cache: hits=141826 misses=1865
-> Gather (cost=214486.73..214486.94 rows=2 width=8) (actual time=5195.366..5196.025 rows=3 loops=1)
Workers Planned: 2
Workers Launched: 2
Buffers: shared hit=178875 read=143691 dirtied=128597 written=127346
Prefetch: hits=0 misses=1865 expired=0 duplicates=0
File cache: hits=141826 misses=1865
-> Partial Aggregate (cost=213486.73..213486.74 rows=1 width=8) (actual time=5187.670..5187.670 rows=1 loops=3)
Buffers: shared hit=178875 read=143691 dirtied=128597 written=127346
Prefetch: hits=0 misses=1865 expired=0 duplicates=0
File cache: hits=141826 misses=1865
-> Parallel Index Only Scan using pgbench_accounts_pkey on pgbench_accounts (cost=0.43..203003.02 rows=4193481 width=0) (actual time=0.574..4928.995 rows=3333333 loops=3)
Heap Fetches: 3675286
Buffers: shared hit=178875 read=143691 dirtied=128597 written=127346
Prefetch: hits=0 misses=1865 expired=0 duplicates=0
File cache: hits=141826 misses=1865
```
The json output uses the following keys and provides integer values for
those keys:
```
...
"File Cache Hits": 141826,
"File Cache Misses": 1865
...
```
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
fixes https://github.com/neondatabase/neon/issues/6889
# Problem
The failure in the last 3 flaky runs on `main` is
```
test_runner/regress/test_remote_storage.py:460: in test_remote_timeline_client_calls_started_metric
churn("a", "b")
test_runner/regress/test_remote_storage.py:457: in churn
assert gc_result["layers_removed"] > 0
E assert 0 > 0
```
That's this code
cd449d66ea/test_runner/regress/test_remote_storage.py (L448-L460)
So, the test expects GC to remove some layers but the GC doesn't.
# Fix
My impression is that the VACUUM isn't re-using pages aggressively
enough, but I can't really prove that. Tried to analyze the layer map
dump but it's too complex.
So, this PR:
- Creates more churn by doing the overwrite twice.
- Forces image layer creation.
It also drive-by removes the redundant call to timeline_compact,
because, timeline_checkpoint already does that internally.
## Problem
Attachment service does not do auth based on JWT scopes.
## Summary of changes
Do JWT based permission checking for requests coming into the attachment
service.
Requests into the attachment service must use different tokens based on
the endpoint:
* `/control` and `/debug` require `admin` scope
* `/upcall` requires `generations_api` scope
* `/v1/...` requires `pageserverapi` scope
Requests into the pageserver from the attachment service must use
`pageserverapi` scope.
## Problem
README.md is missing cleanup instructions
## Summary of changes
Add cleanup instructions
Add instructions how to handle errors during initialization
---------
Co-authored-by: Andreas Scherbaum <andreas@neon.tech>
Use the remote_timeline_client metrics instead, they work for layer file
uploads and are reasonable close to what the
`pageserver_created_persistent_*` metrics were.
Should we wait for empty upload queue before calling `report_size()`?
part of https://github.com/neondatabase/neon/issues/6737
## Problem
Customers should be able to determine the size of their workload's
working set to right size their compute.
Since Neon uses Local file cache (LFC) instead of shared buffers on
bigger compute nodes to cache pages we need to externalize a means to
determine LFC hit ratio in addition to shared buffer hit ratio.
Currently the following end user documentation
fb7cd3af0e/content/docs/manage/endpoints.md (L137)
is wrong because it describes how to right size a compute node based on
shared buffer hit ratio.
Note that the existing functionality in extension "neon" is NOT
available to end users but only to superuser / cloud_admin.
## Summary of changes
- externalize functions and views in neon extension to end users
- introduce a new view `NEON_STAT_FILE_CACHE` with the following DDL
```sql
CREATE OR REPLACE VIEW NEON_STAT_FILE_CACHE AS
WITH lfc_stats AS (
SELECT
stat_name,
count
FROM neon_get_lfc_stats() AS t(stat_name text, count bigint)
),
lfc_values AS (
SELECT
MAX(CASE WHEN stat_name = 'file_cache_misses' THEN count ELSE NULL END) AS file_cache_misses,
MAX(CASE WHEN stat_name = 'file_cache_hits' THEN count ELSE NULL END) AS file_cache_hits,
MAX(CASE WHEN stat_name = 'file_cache_used' THEN count ELSE NULL END) AS file_cache_used,
MAX(CASE WHEN stat_name = 'file_cache_writes' THEN count ELSE NULL END) AS file_cache_writes,
-- Calculate the file_cache_hit_ratio within the same CTE for simplicity
CASE
WHEN MAX(CASE WHEN stat_name = 'file_cache_misses' THEN count ELSE 0 END) + MAX(CASE WHEN stat_name = 'file_cache_hits' THEN count ELSE 0 END) = 0 THEN NULL
ELSE ROUND((MAX(CASE WHEN stat_name = 'file_cache_hits' THEN count ELSE 0 END)::DECIMAL /
(MAX(CASE WHEN stat_name = 'file_cache_hits' THEN count ELSE 0 END) + MAX(CASE WHEN stat_name = 'file_cache_misses' THEN count ELSE 0 END))) * 100, 2)
END AS file_cache_hit_ratio
FROM lfc_stats
)
SELECT file_cache_misses, file_cache_hits, file_cache_used, file_cache_writes, file_cache_hit_ratio from lfc_values;
```
This view can be used by an end user as follows:
```sql
CREATE EXTENSION NEON;
SELECT * from neon. NEON_STAT_FILE_CACHE"
```
The output looks like the following:
```
select * from NEON_STAT_FILE_CACHE;
file_cache_misses | file_cache_hits | file_cache_used | file_cache_writes | file_cache_hit_ratio
-------------------+-----------------+-----------------+-------------------+----------------------
2133643 | 108999742 | 607 | 10767410 | 98.08
(1 row)
```
## Checklist before requesting a review
- [x ] I have performed a self-review of my code.
- [x ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [x ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
## Problem
We want to report how much cache was used and what the limit was.
## Summary of changes
Added one more query to sql_exporter to expose
`neon.file_cache_size_limit`.
* decreases checkpointing and compaction targets for even more layer
files
* write 10 thousand rows 2 times instead of writing 20 thousand rows 1
time so that there is more to GC. Before it was noisily jumping between
1 and 0 layer files, now it's jumping between 19 and 20 layer files. The
0 caused an assertion error that gave the test most of its flakiness.
* larger timeout for the churn while failpoints are active thread: this
is mostly so that the test is more robust on systems with more load
Fixes#3051
## Problem
Previously we always wrote out both legacy and modern tenant config
files. The legacy write enabled rollbacks, but we are long past the
point where that is needed.
We still need the legacy format for situations where someone is running
tenants without generations (that will be yanked as well eventually),
but we can avoid writing it out at all if we do have a generation number
set. We implicitly also avoid writing the legacy config if our mode is
Secondary (secondary mode is newer than generations).
## Summary of changes
- Make writing legacy tenant config conditional on there being no
generation number set.
This PR enforces aspects of `Timeline::repartition` that were already
true at runtime:
- it's not called concurrently, so, bail out if it is anyway (see
comment why it's not called concurrently)
- the `lsn` should never be moving backwards over the lifetime of a
Timeline object, because last_record_lsn() can only move forwards
over the lifetime of a Timeline object
The switch to tokio::sync::Mutex blows up the size of the `partitioning`
field from 40 bytes to 72 bytes on Linux x86_64.
That would be concerning if it was a hot field, but, `partitioning` is
only accessed every 20s by one task, so, there won't be excessive cache
pain on it.
(It still sucks that it's now >1 cache line, but I need the Send-able
MutexGuard in the next PR)
part of https://github.com/neondatabase/neon/issues/6861
It's been dead-code-at-runtime for 9 months, let's remove it.
We can always re-introduce it at a later point.
Came across this while working on #6861, which will touch
`time_for_new_image_layer`. This is an opporunity to make that function
simpler.
## Problem
Since the location config API was added, the attach and detach endpoints
are deprecated. Hiding them from consumers of the swagger definition is
a precursor to removing them
Neon's cloud no longer uses this api since
https://github.com/neondatabase/cloud/pull/10538
Fully removing the APIs will implicitly make use of generation numbers
mandatory, and should happen alongside
https://github.com/neondatabase/neon/issues/5388, which will happen once
we're happy that the storage controller is ready for prime time.
## Summary of changes
- Remove /attach and /detach from pageserver's swagger file
Reverts neondatabase/neon#6765 , bringing back #6731
We concluded that #6731 never was the root cause for the instability in
staging.
More details:
https://neondb.slack.com/archives/C033RQ5SPDH/p1708011674755319
However, the massive amount of concurrent `spawn_blocking` calls from
the `save_metadata` calls during startups might cause a performance
regression.
So, we'll merge this PR here after we've stopped writing the metadata
#6769).
## Problem
In the original PR[0], I've missed a couple of `release` occurrences
that should also be handled for `release-proxy` branch
- [0] https://github.com/neondatabase/neon/pull/6797
## Summary of changes
- Add handling for `release-proxy` branch to allure report
- Add handling for `release-proxy` branch to e2e tests malts.com
We set it for neon replica, if primary is running.
Postgres uses this GUC at the start,
to determine if replica should wait for
RUNNING_XACTS from primary or not.
Corresponding cloud PR is
https://github.com/neondatabase/cloud/pull/10183
* Add test hot-standby replica startup.
* Extract oldest_running_xid from XlRunningXits WAL records.
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
Co-authored-by: Konstantin Knizhnik <knizhnik@garret.ru>
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
## Problem
To "build" a compute image that doesn't have anything new, kaniko takes
13m[0], docker buildx does it in 5m[1].
Also, kaniko doesn't fully support bash expressions in the Dockerfile
`RUN`, so we have to use different workarounds for this (like `bash -c
...`).
- [0]
https://github.com/neondatabase/neon/actions/runs/8011512414/job/21884933687
- [1]
https://github.com/neondatabase/neon/actions/runs/8008245697/job/21874278162
## Summary of changes
- Use docker buildx to build `compute-node` images
- Use docker buildx to build `neon-image` image
- Use docker buildx to build `compute-tools` image
- Use docker hub for image cache (instead of ECR)
## Problem
Following up https://github.com/neondatabase/neon/pull/6884, hopefully,
a real final fix for https://github.com/neondatabase/neon/issues/6236.
## Summary of changes
`handle_migrations` is done over the main `postgres` db connection.
Therefore, the privileges assigned here do not work with databases
created later (i.e., `neondb`). This pull request moves the grants to
`handle_grants`, so that it runs for each DB created. The SQL is added
into the `BEGIN/END` block, so that it takes only one RTT to apply all
of them.
Signed-off-by: Alex Chi Z <chi@neon.tech>
PR #6655 turned out to be not enough to prevent .snap files bloat; some
subscribers just don't ack flushed position, thus never advancing the
slot. Probably other bloating scenarios are also possible, so add a more direct
restriction -- drop all slots if too many .snap files has been discovered.
## Problem
See https://neondb.slack.com/archives/C04DGM6SMTM/p1708363190710839
## Summary of changes
Flush logical message with snapshot and origin state
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
## Summary of changes
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
## Problem
Following up on https://github.com/neondatabase/neon/pull/6845, we did
not make the default privileges grantable before, and therefore, even if
the users have full privileges, they are not able to grant them to
others.
Should be a final fix for
https://github.com/neondatabase/neon/issues/6236.
## Summary of changes
Add `WITH GRANT` to migrations so that neon_superuser can grant the
permissions.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
We want to release Proxy at a different cadence.
## Summary of changes
- build-and-test workflow:
- Handle the `release-proxy` branch
- Tag images built on this branch with `release-proxy-XXX` tag
- Trigger deploy workflow with `deployStorage=true` &
`deployStorageBroker=true` on `release` branch
- Trigger deploy workflow with `deployPgSniRouter=true` &
`deployProxy=true` on `release-proxy` branch
- release workflow (scheduled creation of release branch):
- Schedule Proxy releases for Thursdays (a random day to make it
different from Storage releases)
- Some checks weren't properly returning an error when they failed
- TenantState::to_persistent wasn't setting generation_pageserver
properly
- Changes to node scheduling policy weren't being persisted.
Stacks on https://github.com/neondatabase/neon/pull/6823
- Pending a heartbeating mechanism (#6844 ), use /re-attach calls as a
cue to mark an offline node as active, so that a node which is
unavailable during controller startup doesn't require manual
intervention if it later starts/restarts.
- Tweak scheduling logic so that when we schedule the attached location
for a tenant, we prefer to select from secondary locations rather than
picking a fresh one.
This is an interim state until we implement #6844 and full chaos testing
for handling failures.
The `refill_interval` switched from a milliseconds usize to a Duration
during a review follow-up, hence this slipped through manual testing.
Part of https://github.com/neondatabase/neon/issues/5899
PR adds a simple at most 1Hz refreshed informational API for querying
pageserver utilization. In this first phase, no actual background
calculation is performed. Instead, the worst possible score is always
returned. The returned bytes information is however correct.
Cc: #6835
Cc: #5331
- Add some context to logs
- Add tests for pageserver restarts when managed by storage controller
- Make /location_config tolerate compute hook failures on shard
creations, not just modifications.
## Problem
When a secondary mode location starts up, it scans local layer files.
Currently it warns on any layers whose names don't parse as a
LayerFileName, generating warning spam from perfectly normal tempfiles.
## Summary of changes
- Refactor local vars to build a Utf8PathBuf for the layer file
candidate
- Use the crate::is_temporary check to identify + clean up temp files.
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
fix https://github.com/neondatabase/neon/issues/6236 again
## Summary of changes
This pull request adds a setup command in compute spec to modify default
privileges of public schema to have full permission on table/sequence
for neon_superuser. If an extension upgrades to superuser during
creation, the tables/sequences they create in the public schema will be
automatically granted to neon_superuser.
Questions:
* does it impose any security flaws? public schema should be fine...
* for all extensions that create tables in schemas other than public, we
will need to manually handle them (e.g., pg_anon).
* we can modify some extensions to remove their superuser requirement in
the future.
* we may contribute to Postgres to allow for the creation of extensions
with a specific user in the future.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
For PRs from external contributors, we're still running `actionlint` and
`neon_extra_builds` workflows (which could fail due to lack of
permissions to secrets).
## Summary of changes
- Extract `check-permissions` job to a separate reusable workflow
- Depend all jobs from `actionlint` and `neon_extra_builds` workflows on
`check-permissions`
curl_global_init() with an IPv6 enabled curl build on macOS will cause
the calling program to become multithreaded. Unfortunately for
shared_preload_libraries, that means the postmaster becomes
multithreaded, which CANNOT happen. There are checks in Postgres to make
sure that this is not the case.
Add `--walsenders-keep-horizon` argument to safekeeper cmdline. It will
prevent deleting WAL segments from disk if they are needed by the active
START_REPLICATION connection.
This is useful for sharding. Without this option, if one of the shard
falls behind, it starts to read WAL from S3, which is much slower than
disk. This can result in huge shard lagging.
## Problem
Accidentally merged #6852 without this test stability change. The test
as-written could sometimes fail on debug-pg14.
## Summary of changes
- Write more data so that the test can more reliably assert on the ratio
of total layers to small layers
- Skip the test in debug mode, since writing any more than a tiny bit of
data tends to result in a flaky test in the much slower debug
environment.
## Problem
PR #6834 introduced an assertion that the sets of metric labels on
finished operations should equal those on started operations, which is
not true if no operations have finished yet for a particular set of
labels.
## Summary of changes
- Instead of asserting out, wait and re-check in the case that finished
metrics don't match started
## Problem
- Current file has ambiguous ownership for some paths
- The /control_plane/attachment_service is storage specific & updates
there don't need to request reviews from other teams.
## Summary of changes
- Define a single owning team per path, so that we can make reviews by
that team mandatory in future.
- Remove the top-level /control_plane as no one specific team owns
neon_local, and we would rarely see a PR that exclusively touches that
path.
- Add an entry for /control_plane/attachment_service, which is newer
storage-specific code.
The sharding service didn't have support for S3 disaster recovery.
This PR adds a new endpoint to the attachment service, which is slightly
different from the endpoint on the pageserver, in that it takes the
shard count history of the tenant as json parameters: we need to do
time travel recovery for both the shard count at the target time and the
shard count at the current moment in time, as well as the past shard
counts that either still reference.
Fixes#6604, part of https://github.com/neondatabase/cloud/issues/8233
---------
Co-authored-by: John Spray <john@neon.tech>
As noticed in #6836 some occurances of error conversions were missed in
#6697:
- `std::io::Error` popped up by `tokio::io::copy_buf` containing
`DownloadError` was turned into `DownloadError::Other`
- similarly for secondary downloader errors
These changes come at the loss of pathname context.
Cc: #6096
## Problem
Where the stripe size is the same order of magnitude as the checkpoint
distance (such as with default settings), tenant shards can easily pass
through `checkpoint_distance` bytes of LSN without actually ingesting
anything. This results in emitting many tiny L0 delta layers.
## Summary of changes
- Multiply checkpoint distance by shard count before comparing with LSN
distance. This is a heuristic and does not guarantee that we won't emit
small layers, but it fixes the issue for typical cases where the writes
in a (checkpoint_distance * shard_count) range of LSN bytes are somewhat
distributed across shards.
- Add a test that checks the size of layers after ingesting to a sharded
tenant; this fails before the fix.
---------
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
## Problem
## Summary of changes
1. Classify further cplane API errors
2. add 'serviceratelimit' and make a few of the timeout errors return
that.
3. a few additional minor changes
## Problem
During startup_reconcile we do a couple of potentially-slow things:
- Calling out to all nodes to read their locations
- Calling out to the cloud control plane to notify it of all tenants'
attached nodes
The read of node locations was not being done concurrently across nodes,
and neither operation was bounded by a well defined deadline.
## Summary of changes
- Refactor the async parts of startup_reconcile into separate functions
- Add concurrency and deadline to `scan_node_locations`
- Add deadline to `compute_notify_many`
- Run `cleanup_locations` in the background: there's no need for
startup_reconcile to wait for this to complete.
This PR introduces a new vectored implementation of the read path.
The search is basically a DFS if you squint at it long enough.
LayerFringe tracks the next layers to visit and acts as our stack.
Vertices are tuples of (layer, keyspace, lsn range). Continuously
pop the top of the stack (most recent layer) and do all the reads
for one layer at once.
The search maintains a fringe (`LayerFringe`) which tracks all the
layers that intersect the current keyspace being searched. Continuously
pop the top of the fringe (layer with highest LSN) and get all the data
required from the layer in one go.
Said search is done on one timeline at a time. If data is still required for
some keys, then search the ancestor timeline.
Apart from the high level layer traversal, vectored variants have been
introduced for grabbing data from each layer type. They still suffer from
read amplification issues and that will be addressed in a different PR.
You might notice that in some places we duplicate the code for the
existing read path. All of that code will be removed when we switch
the non-vectored read path to proxy into the vectored read path.
In the meantime, we'll have to contend with the extra cruft for the sake
of testing and gentle releasing.
This pull request adds two flags: `--update-catalog true` for `endpoint
create`, and `--create-test-user true` for `endpoint start`. The former
enables catalog updates for neon_superuser permission and many other
things, while the latter adds the user `test` and the database `neondb`
when setting up the database. A combination of these two flags will
create a Postgres similar to the production environment so that it would
be easier for us to test if extensions behave correctly when added to
Neon Postgres.
Example output:
```
❯ cargo neon endpoint start main --create-test-user true
Finished dev [unoptimized + debuginfo] target(s) in 0.22s
Running `target/debug/neon_local endpoint start main --create-test-user true`
Starting existing endpoint main...
Starting postgres node at 'postgresql://cloud_admin@127.0.0.1:55432/postgres'
Also at 'postgresql://user@127.0.0.1:55432/neondb'
```
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
A set of small changes that are too small to open a separate for each.
A notable change is adding `pytest-repeat` plugin, which can help to
ensure that a flaky test is fixed by running such a test several times.
## Summary of changes
- Update Allure from 2.24.0 to 2.27.0
- Update Ruff from 0.1.11 to 0.2.2 (update `[tool.ruff]` section of
`pyproject.toml` for it)
- Install pytest-repeat plugin
587cb705b8
changed the layer rolling logic to more closely obey the
`checkpoint_distance` config. Previously, this test was getting
layers significantly larger than the 8K it was asking for. Now the
payload in the layers is closer to 8K (which means more layers in
total).
Tweak the `checkpoint_distance` to get a number of layers more
reasonable for this test. Note that we still get more layers than
before (~8K vs ~5K).
this is to speed up suspends, see
https://github.com/neondatabase/cloud/issues/10284
## Problem
## Summary of changes
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
refs #6737
# Problem
Before this PR, on-demand downloads weren't measured per tenant_id.
This makes root-cause analysis of latency spikes harder, requiring us to
resort to log scraping for
```
{neon_service="pageserver"} |= `downloading on-demand` |= `$tenant_id`
```
which can be expensive when zooming out in Grafana.
Context: https://neondb.slack.com/archives/C033RQ5SPDH/p1707809037868189
# Solution / Changes
- Remove the calls_started histogram
- I did the dilegence, there are only 2 dashboards using this histogram,
and in fact only one uses it as a histogram, the other just as a
a counter.
- [Link
1](8115b54d9f/neonprod/dashboards/hkXNF7oVz/dashboard-Z31XmM24k.yaml (L1454)):
`Pageserver Thrashing` dashboard, linked from playbook, will fix.
- [Link
2](8115b54d9f/neonprod/dashboards/CEllzAO4z/dashboard-sJqfNFL4k.yaml (L599)):
one of my personal dashboards, unused for a long time, already broken in
other ways, no need to fix.
- replace `pageserver_remote_timeline_client_calls_unfinished` gauge
with a counter pair
- Required `Clone`-able `IntCounterPair`, made the necessary changes in
the `libs/metrics` crate
- fix tests to deal with the fallout
A subsequent PR will remove a timeline-scoped metric to compensate.
Note that we don't need additional global counters for the per-timeline
counters affected by this PR; we can use the `remote_storage` histogram
for those, which, conveniently, also include the secondary-mode
downloads, which aren't covered by the remote timeline client metrics
(should they?).
## Problem
`download_retry` correctly uses a fatal check to avoid retrying forever
on cancellations and NotFound cases. However, `download_layer_file` was
casting all download errors to "Other" in order to attach an
anyhow::Context.
Noticed this issue in the context of secondary downloads, where requests
to download layers that might not exist are issued intentionally, and
this resulted in lots of error spam from retries that shouldn't have
happened.
## Summary of changes
- Remove the `.context()` so that the original DownloadError is visible
to backoff::retry
refs https://github.com/neondatabase/neon/issues/4112
amends https://github.com/neondatabase/neon/pull/6687
Since my last PR #6687 regarding this test, the type of flakiness that
has been observed has shifted to the beginning of the test, where we
create the layers:
```
timed out while waiting for remote_consistent_lsn to reach 0/411A5D8, was 0/411A5A0
```
[Example Allure
Report](https://neon-github-public-dev.s3.amazonaws.com/reports/pr-6789/7932503173/index.html#/testresult/ddb877cfa4062f7d)
Analysis
--------
I suspect there was the following race condition:
- endpoints push out some tiny piece of WAL during their
endpoints.stop_all()
- that WAL reaches the SK (it's just one SK according to logs)
- the SKs send it into the walreceiver connection
- the SK gets shut down
- the checkpoint is taken, with last_record_lsn = 0/411A5A0
- the PS's walreceiver_connection_handler processes the WAL that was
sent into the connection by the SKs; this advances
last_record_lsn to 0/411A5D8
- we get current_lsn = 0/411A5D8
- nothing flushes a layer
Changes
-------
There's no testing / debug interface to shut down / server all
walreceiver connections.
So, this PR restarts pageserver to achieve it.
Also, it lifts the "wait for image layer uploads" further up, so that
after this first
restart, the pageserver really does _nothing_ by itself, and so, the
origianl physical size mismatch issue quoted in #6687 should be fixed.
(My initial suspicion hasn't changed that it was due to the tiny chunk
of endpoint.stop_all() WAL being ingested after the second PS restart.)
Before this PR, if remote storage is configured, `load_layer_map`'s call
to `RemoteTimelineClient::schedule_layer_file_deletion` would schedule
an empty UploadOp::Delete for each timeline.
It's jsut CPU overhead, no actual interaction with deletion queue
on-disk state or S3, as far as I can tell.
However, it shows up in the "RemoteTimelineClient calls started
metrics", which I'm refining in an orthogonal PR.
## Problem
Proxy already supported HTTP2, but I expect no one is using it because
we don't advertise it in the TLS handshake.
## Summary of changes
#6335 without the websocket changes.
## Problem
Timeline creation is meant to be very fast: it should only take
approximately on S3 PUT latency. When we have many shards in a tenant,
we should preserve that responsiveness.
## Summary of changes
- Issue create/delete pageserver API calls concurrently across all >0
shards
- During tenant deletion, delete shard zero last, separately, to avoid
confusing anything using GETs on the timeline.
- Return 201 instead of 200 on creations to make cloud control plane
happy
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
We use a bunch of deprecated actions.
See https://github.com/neondatabase/neon/actions/runs/7958569728
(Annotations section)
```
Node.js 16 actions are deprecated. Please update the following actions to use Node.js 20: actions/checkout@v3, actions/setup-java@v3, actions/cache@v3, actions/github-script@v6. For more information see: https://github.blog/changelog/2023-09-22-github-actions-transitioning-from-node-16-to-node-20/.
```
## Summary of changes
- `actions/cache@v3` -> `actions/cache@v4`
- `actions/checkout@v3` -> `actions/checkout@v4`
- `actions/github-script@v6` -> `actions/github-script@v7`
- `actions/setup-java@v3` -> `actions/setup-java@v4`
- `actions/upload-artifact@v3` -> `actions/upload-artifact@v4`
This PR stacks on https://github.com/neondatabase/neon/pull/6814
Observability:
- Because we only persist a subset of our state, and our external API is
pretty high level, it can be hard to get at the detail of what's going
on internally (e.g. the IntentState of a shard).
- Add debug endpoints for getting a full dump of all TenantState and
SchedulerNode objects
- Enrich the /control/v1/node listing endpoint to include full in-memory
detail of `Node` rather than just the `NodePersistence` subset
Consistency checks:
- The storage controller maintains separate in-memory and on-disk
states, by design. To catch subtle bugs, it is useful to occasionally
cross-check these.
- The Scheduler maintains reference counts for shard->node
relationships, which could drift if there was a bug in IntentState:
exhausively cross check them in tests.
## Problem
When investigating test failures
(https://github.com/neondatabase/neon/issues/6813) I noticed we were
doing a bunch of Reconciler runs right after splitting a tenant.
It's because the splitting test does a pageserver restart, and there was
a bug in /re-attach handling, where we would update the generation
correctly in the database and intent state, but not observed state,
thereby triggering a reconciliation on the next call to maybe_reconcile.
This didn't break anything profound (underlying rules about generations
were respected), but caused the storage controller to do an un-needed
extra round of bumping the generation and reconciling.
## Summary of changes
- Start adding metrics to the storage controller
- Assert on the number of reconciles done in test_sharding_split_smoke
- Fix /re-attach to update `observed` such that we don't spuriously
re-reconcile tenants.
Often times the tenants we want to (WAL) DR are the ones which the
pageserver marks as broken. Therefore, we should allow initdb
preservation also for broken tenants.
Fixes#6781.
## Problem
Sharded tenants could panic during compaction when they try to generate
an L1 delta layer for a region that contains no keys on a particular
shard.
This is a variant of https://github.com/neondatabase/neon/issues/6755,
where we attempt to save a delta layer with no keys. It is harder to
reproduce than the case of image layers fixed in
https://github.com/neondatabase/neon/pull/6776.
It will become even less likely once
https://github.com/neondatabase/neon/pull/6778 tweaks keyspace
generation, but even then, we should not rely on keyspace partitioning
to guarantee at least one stored key in each partition.
## Summary of changes
- Move construction of `writer` in `compact_level0_phase1`, so that we
never leave a writer constructed but without any keys.
## Problem
One of the major shortcuts in the initial version of this code was to
construct a fresh `Scheduler` each time we need it, which is an O(N^2)
cost as the tenant count increases.
## Summary of changes
- Keep `Scheduler` alive through the lifetime of ServiceState
- Use `IntentState` as a reference tracking helper, updating Scheduler
refcounts as nodes are added/removed from the intent.
There is an automated test that checks things don't get pathologically
slow with thousands of shards, but it's not included in this PR because
tests that implicitly test the runner node performance take some thought
to stabilize/land in CI.
## Problem
Secondary mode locations keep a local copy of the heatmap, which needs
cleaning up during deletion.
Closes: https://github.com/neondatabase/neon/issues/6802
## Summary of changes
- Extend test_live_migration to reproduce the issue
- Remove heatmap-v1.json during tenant deletion
1) `scram::password` was used in tests only. can be replaced with
`postgres_protocol::password`.
2) `postgres_protocol::authentication::sasl` provides a client impl of
SASL which improves our ability to test
## Problem
One WAL record can actually produce an arbitrary amount of key value pairs.
This is problematic since it might cause our frozen layers to bloat past the
max allowed size of S3 single shot uploads.
[#6639](https://github.com/neondatabase/neon/pull/6639) introduced a "should roll"
check after every batch of `ingest_batch_size` (100 WAL records by default). This helps,
but the original problem still exists.
## Summary of changes
This patch moves the responsibility of rolling the currently open layer
to the `TimelineWriter`. Previously, this was done ad-hoc via calls
to `check_checkpoint_distance`. The advantages of this approach are:
* ability to split one batch over multiple open layers
* less layer map locking
* remove ad-hoc check_checkpoint_distance calls
More specifically, we track the current size of the open layer in the
writer. On each `put` check whether the current layer should be closed
and a new one opened. Keeping track of the currently open layer results
in less contention on the layer map lock. It only needs to be acquired
on the first write and on writes that require a roll afterwards.
Rolling the open layer can be triggered by:
1. The distance from the last LSN we rolled at. This bounds the amount
of WAL that the safekeepers need to store.
2. The size of the currently open layer.
3. The time since the last roll. It helps safekeepers to regard
pageserver as caught up and suspend activity.
Closes#6624
## Problem
test_sharding_split_unsharded was flaky with log errors from tenants not
being active. This was happening when the split function enters
wait_lsn() while the child shard might still be activating. It's flaky
rather than an outright failure because activation is usually very fast.
This is also a real bug fix, because in realistic scenarios we could
proceed to detach the parent shard before the children are ready,
leading to an availability gap for clients.
## Summary of changes
- Do a short wait_to_become_active on the child shards before proceeding
to wait for their LSNs to advance
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
`test_create_snapshot` is flaky[0] on CI and fails constantly on macOS,
but with a slightly different error:
```
shutil.Error: [('/Users/bayandin/work/neon/test_output/test_create_snapshot[release-pg15-1-100]/repo/endpoints/ep-1/pgdata/pg_dynshmem', '/Users/bayandin/work/neon/test_output/compatibility_snapshot_pgv15/repo/endpoints/ep-1/pgdata/pg_dynshmem', "[Errno 2] No such file or directory: '/Users/bayandin/work/neon/test_output/test_create_snapshot[release-pg15-1-100]/repo/endpoints/ep-1/pgdata/pg_dynshmem'")]
```
Also (on macOS) `repo/endpoints/ep-1/pgdata/pg_dynshmem` is a symlink
to `/dev/shm/`.
- [0] https://github.com/neondatabase/neon/issues/6784
## Summary of changes
Ignore `pg_dynshmem` directory while copying a snapshot
## Problem
Sharded tenants would sometimes try to write empty image layers during
compaction: this was more noticeable on larger databases.
- https://github.com/neondatabase/neon/issues/6755
**Note to reviewers: the last commit is a refactor that de-intents a
whole block, I recommend reviewing the earlier commits one by one to see
the real changes**
## Summary of changes
- Fix a case where when we drop a key during compaction, we might fail
to write out keys (this was broken when vectored get was added)
- If an image layer is empty, then do not try and write it out, but
leave `start` where it is so that if the subsequent key range meets
criteria for writing an image layer, we will extend its key range to
cover the empty area.
- Add a compaction test that configures small layers and compaction
thresholds, and asserts that we really successfully did image layer
generation. This fails before the fix.
## Problem
See https://github.com/neondatabase/cloud/issues/10268
## Summary of changes
Add pg_ivm extension
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
Co-authored-by: Alexander Bayandin <alexander@neon.tech>
This reverts commit 9ad940086c.
This pull request reverts #6733 to avoid incompatibility with pgvector
and I will push further fixes later. Note that after reverting this pull
request, the postgres submodule will point to some detached branches.
## Problem
Even if you're not enforcing auth, the JwtAuth middleware barfs on
scopes it doesn't know about.
Add `generations_api` scope, which was invented in the cloud control
plane for the pageserver's /re-attach and /validate upcalls: this will
be enforced in storage controller's implementation of these in a later
PR.
Unfortunately the scope's naming doesn't match the other scope's naming
styles, so needs a manual serde decorator to give it an underscore.
## Summary of changes
- Add `Scope::GenerationsApi` variant
- Update pageserver + safekeeper auth code to print appropriate message
if they see it.
## Problem
`test_pageserver_max_throughput_getpage_at_latest_lsn` is flaky which
makes CI status red pretty frequently. `benchmarks` is not a blocking
job (doesn't block `deploy`), so having it red might hide failures in
other jobs
Ref: https://github.com/neondatabase/neon/issues/6724
## Summary of changes
- Disable `test_pageserver_max_throughput_getpage_at_latest_lsn` on CI
until it fixed
## Problem
Now that the storage controller is working end to end, we start burning
down the robustness aspects.
## Summary of changes
- Add a background task that periodically calls `reconcile_all`. This
ensures that if earlier operations couldn't succeed (e.g. because a node
was unavailable), we will eventually retry. This is a naive initial
implementation can start an unlimited number of reconcile tasks:
limiting reconcile concurrency is a later item in #6342
- Add a number of tracing spans in key locations: each background task,
each reconciler task.
- Add a top level CancellationToken and Gate, and use these to implement
a graceful shutdown that waits for tasks to shut down. This is not
bulletproof yet, because within these tasks we have remote HTTP calls
that aren't wrapped in cancellation/timeouts, but it creates the
structure, and if we don't shutdown promptly then k8s will kill us.
- To protect shard splits from background reconciliation, expose the `SplitState`
in memory and use it to guard any APIs that require an attached tenant.
## Problem
The ShardCount type has a magic '0' value that represents a legacy
single-sharded tenant, whose TenantShardId is formatted without a
`-0001` suffix (i.e. formatted as a traditional TenantId).
This was error-prone in code locations that wanted the actual number of
shards: they had to handle the 0 case specially.
## Summary of changes
- Make the internal value of ShardCount private, and expose `count()`
and `literal()` getters so that callers have to explicitly say whether
they want the literal value (e.g. for storing in a TenantShardId), or
the actual number of shards in the tenant.
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
This fixes issues where `neon-pg-ext-clean-vYY` is used as target and
resolves using the `neon-pg-ext-%` template with `$*` resolving as `clean-vYY`, for
older versions of GNU Make, rather than `neon-pg-ext-clean-%` using `$*` = `vYY`
## Problem
```
$ make clean
...
rm -f pg_config_paths.h
Compiling neon clean-v14
mkdir -p /Users/<user>/neon-build//pg_install//build/neon-clean-v14
/Applications/Xcode.app/Contents/Developer/usr/bin/make PG_CONFIG=/Users/<user>/neon-build//pg_install//clean-v14/bin/pg_config CFLAGS='-O0 -g3 ' \
-C /Users/<user>/neon-build//pg_install//build/neon-clean-v14 \
-f /Users/<user>/neon-build//pgxn/neon/Makefile install
make[1]: /Users/<user>/neon-build//pg_install//clean-v14/bin/pg_config: Command not found
make[1]: *** No rule to make target `install'. Stop.
make: *** [neon-pg-ext-clean-v14] Error 2
```
## Problem
Currently, we don't store `PLATFORM` for Nightly Benchmarks. It
causes them to be merged as reruns in Allure report (because they have
the same test name).
## Summary of changes
- Parametrize benchmarks by
- Postgres Version (14/15/16)
- Build Type (debug/release/remote)
- PLATFORM (neon-staging/github-actions-selfhosted/...)
---------
Co-authored-by: Bodobolero <peterbendel@neon.tech>
close https://github.com/neondatabase/neon/issues/6236
This pull request bumps neon postgres dependencies. The corresponding
postgres commits fix the checks for superuser permission when creating
an extension. Also, for creating native functinos, it now allows
neon_superuser only in the extension creation process.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
## Problem
- We weren't deleting parent shard contents once the split was done
- Re-downloading layers into child shards is wasteful
## Summary of changes
- Hard-link layers into child chart local storage during split
- Delete parent shards content at the end
---------
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
Cancellation and timeouts are handled at remote_storage callsites, if
they are. However they should always be handled, because we've had
transient problems with remote storage connections.
- Add cancellation token to the `trait RemoteStorage` methods
- For `download*`, `list*` methods there is
`DownloadError::{Cancelled,Timeout}`
- For the rest now using `anyhow::Error`, it will have root cause
`remote_storage::TimeoutOrCancel::{Cancel,Timeout}`
- Both types have `::is_permanent` equivalent which should be passed to
`backoff::retry`
- New generic RemoteStorageConfig option `timeout`, defaults to 120s
- Start counting timeouts only after acquiring concurrency limiter
permit
- Cancellable permit acquiring
- Download stream timeout or cancellation is communicated via an
`std::io::Error`
- Exit backoff::retry by marking cancellation errors permanent
Fixes: #6096Closes: #4781
Co-authored-by: arpad-m <arpad-m@users.noreply.github.com>
## Problem
Building on #5875 to add handy test functions for autoscaling.
Resolves#5609
## Summary of changes
This PR makes the following changes to #5875:
- Enable `neon_test_utils` extension in the compute node docker image,
so we could use it in the e2e tests (as discussed with @kelvich).
- Removed test functions related to disk as we don't use them for
autoscaling.
- Fix the warning with printf-ing unsigned long variables.
---------
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
The canonical release artifact of neon.git is the Docker image with all
the binaries in them:
```
docker pull neondatabase/neon:release-4854
docker create --name extract neondatabase/neon:release-4854
docker cp extract:/usr/local/bin/pageserver ./pageserver.release-4854
chmod +x pageserver.release-4854
cp -a pageserver.release-4854 ./target/release/pageserver
```
Before this PR, these artifacts didn't expose the `keyspace` API,
thereby preventing `pagebench get-page-latest-lsn` from working.
Having working pagebench is useful, e.g., for experiments in staging.
So, expose the API, but don't document it, as it's not part of the
interface with control plane.
## Problem
Aux files were stored with an O(N^2) cost, since on each modification
the entire map is re-written as a page image.
This addresses one axis of the inefficiency in logical replication's use
of storage (https://github.com/neondatabase/neon/issues/6626). It will
still be writing a large amount of duplicative data if writing the same
slot's state every 15 seconds, but the impact will be O(N) instead of
O(N^2).
## Summary of changes
- Introduce `NeonWalRecord::AuxFile`
- In `DatadirModification`, if the AUX_FILES_KEY has already been set,
then write a delta instead of an image
context: https://github.com/neondatabase/neon/issues/6663
Building atop #6664, this PR switches `write_all_at` to take owned
buffers.
The main challenge here is the `EphemeralFile::mutable_tail`, for which
I'm picking the ugly solution of an `Option` that is `None` while the IO
is in flight.
After this, we will be able to switch `write_at` to take owned buffers
and call tokio-epoll-uring's `write` function with that owned buffer.
That'll be done in #6378.
Some callers of `VirtualFile::crashsafe_overwrite` call it on the
executor thread, thereby potentially stalling it.
Others are more diligent and wrap it in `spawn_blocking(...,
Handle::block_on, ... )` to avoid stalling the executor thread.
However, because `crashsafe_overwrite` uses
VirtualFile::open_with_options internally, we spawn a new thread-local
`tokio-epoll-uring::System` in the blocking pool thread that's used for
the `spawn_blocking` call.
This PR refactors the situation such that we do the `spawn_blocking`
inside `VirtualFile::crashsafe_overwrite`. This unifies the situation
for the better:
1. Callers who didn't wrap in `spawn_blocking(..., Handle::block_on,
...)` before no longer stall the executor.
2. Callers who did it before now can avoid the `block_on`, resolving the
problem with the short-lived `tokio-epoll-uring::System`s in the
blocking pool threads.
A future PR will build on top of this and divert to tokio-epoll-uring if
it's configures as the IO engine.
Changes
-------
- Convert implementation to std::fs and move it into `crashsafe.rs`
- Yes, I know, Safekeepers (cc @arssher ) added `durable_rename` and
`fsync_async_opt` recently. However, `crashsafe_overwrite` is different
in the sense that it's higher level, i.e., it's more like
`std::fs::write` and the Safekeeper team's code is more building block
style.
- The consequence is that we don't use the VirtualFile file descriptor
cache anymore.
- I don't think it's a big deal because we have plenty of slack wrt
production file descriptor limit rlimit (see [this
dashboard](https://neonprod.grafana.net/d/e4a40325-9acf-4aa0-8fd9-f6322b3f30bd/pageserver-open-file-descriptors?orgId=1))
- Use `tokio::task::spawn_blocking` in
`VirtualFile::crashsafe_overwrite` to call the new
`crashsafe::overwrite` API.
- Inspect all callers to remove any double-`spawn_blocking`
- spawn_blocking requires the captures data to be 'static + Send. So,
refactor the callers. We'll need this for future tokio-epoll-uring
support anyway, because tokio-epoll-uring requires owned buffers.
Related Issues
--------------
- overall epic to enable write path to tokio-epoll-uring: #6663
- this is also kind of relevant to the tokio-epoll-uring System creation
failures that we encountered in staging, investigation being tracked in
#6667
- why is it relevant? Because this PR removes two uses of
`spawn_blocking+Handle::block_on`
## Problem
In a recent refactor, we accidentally dropped the cancel session early
## Summary of changes
Hold the cancel session during proxy passthrough
Cherry-pick Upstream commit fbf9a7ac4d to neon stable branches. We'll
get it in the next PostgreSQL minor release anyway, but we need it
now, if we want to start using the 'mmap' implementation.
See https://github.com/neondatabase/autoscaling/issues/800 for the
plans on doing that.
There is O(n^2) issues due to how we store these directories (#6626), so
it's good to keep an eye on them and ensure the numbers stay low.
The new per-timeline metric `pageserver_directory_entries_count`
isn't perfect, namely we don't calculate it every time we attach
the timeline, but only if there is an actual change.
Also, it is a collective metric over multiple scalars. Lastly,
we only emit the metric if it is above a certain threshold.
However, the metric still give a feel for the general size of the timeline.
We care less for small values as the metric is mainly there to
detect and track tenants with large directory counts.
We also expose the directory counts in `TimelineInfo` so that one can
get the detailed size distribution directly via the pageserver's API.
Related: #6642 , https://github.com/neondatabase/cloud/issues/10273
## Problem
See https://github.com/neondatabase/neon/issues/6674
Current implementation of `neon_redo_read_buffer_filter` performs fast
exist for catalog pages:
```
/*
* Out of an abundance of caution, we always run redo on shared catalogs,
* regardless of whether the block is stored in shared buffers. See also
* this function's top comment.
*/
if (!OidIsValid(NInfoGetDbOid(rinfo)))
return false;
*/
as a result last written lsn and relation size for FSM fork are not correctly updated for catalog relations.
## Summary of changes
Do not perform fast path return for catalog relations.
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with /release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above checklist
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
Test sometimes fails with `used_blocks > total_blocks`, because when
using mocked statvfs with the total blocks set to the size of data on
disk before starting, we are implicitly asserting that nothing at all
can be written to disk between startup and calling statvfs.
Related: https://github.com/neondatabase/neon/issues/6511
## Summary of changes
- Use HTTP API to invoke disk usage eviction instead of mocked statvfs
This PR contains the first version of a
[FoundationDB-like](https://www.youtube.com/watch?v=4fFDFbi3toc)
simulation testing for safekeeper and walproposer.
### desim
This is a core "framework" for running determenistic simulation. It
operates on threads, allowing to test syncronous code (like walproposer).
`libs/desim/src/executor.rs` contains implementation of a determenistic
thread execution. This is achieved by blocking all threads, and each
time allowing only a single thread to make an execution step. All
executor's threads are blocked using `yield_me(after_ms)` function. This
function is called when a thread wants to sleep or wait for an external
notification (like blocking on a channel until it has a ready message).
`libs/desim/src/chan.rs` contains implementation of a channel (basic
sync primitive). It has unlimited capacity and any thread can push or
read messages to/from it.
`libs/desim/src/network.rs` has a very naive implementation of a network
(only reliable TCP-like connections are supported for now), that can
have arbitrary delays for each package and failure injections for
breaking connections with some probability.
`libs/desim/src/world.rs` ties everything together, to have a concept of
virtual nodes that can have network connections between them.
### walproposer_sim
Has everything to run walproposer and safekeepers in a simulation.
`safekeeper.rs` reimplements all necesary stuff from `receive_wal.rs`,
`send_wal.rs` and `timelines_global_map.rs`.
`walproposer_api.rs` implements all walproposer callback to use
simulation library.
`simulation.rs` defines a schedule – a set of events like `restart <sk>`
or `write_wal` that should happen at time `<ts>`. It also has code to
spawn walproposer/safekeeper threads and provide config to them.
### tests
`simple_test.rs` has tests that just start walproposer and 3 safekeepers
together in a simulation, and tests that they are not crashing right
away.
`misc_test.rs` has tests checking more advanced simulation cases, like
crashing or restarting threads, testing memory deallocation, etc.
`random_test.rs` is the main test, it checks thousands of random seeds
(schedules) for correctness. It roughly corresponds to running a real
python integration test in an environment with very unstable network and
cpu, but in a determenistic way (each seed results in the same execution
log) and much much faster.
Closes#547
---------
Co-authored-by: Arseny Sher <sher-ars@yandex.ru>
## Problem
Not really a problem, just refactoring.
## Summary of changes
Separate authenticate from wake compute.
Do not call wake compute second time if we managed to connect to
postgres or if we got it not from cache.
In #6079 it was found that there is no test that executes the scrubber.
We now add such a test, which does the following things:
* create a tenant, write some data
* run the scrubber
* remove the tenant
* run the scrubber again
Each time, the scrubber runs the scan-metadata command. Before #6079 we
would have errored, now we don't.
Fixes#6080
Refactor out layer accesses so that we can have easy access to resident
layers, which are needed for number of cases instead of layers for
eviction. Simplifies the heatmap building by only using Layers, not
RemoteTimelineClient.
Cc: #5331
## Problem
hard to see where time is taken during HTTP flow.
## Summary of changes
add a lot more for query state. add a conn_id field to the sql-over-http
span
This PR refactors the `blob_io` code away from using slices towards
taking owned buffers and return them after use.
Using owned buffers will eventually allow us to use io_uring for writes.
part of https://github.com/neondatabase/neon/issues/6663
Depends on https://github.com/neondatabase/tokio-epoll-uring/pull/43
The high level scheme is as follows:
- call writing functions with the `BoundedBuf`
- return the underlying `BoundedBuf::Buf` for potential reuse in the
caller
NB: Invoking `BoundedBuf::slice(..)` will return a slice that _includes
the uninitialized portion of `BoundedBuf`_.
I.e., the portion between `bytes_init()` and `bytes_total()`.
It's a safe API that actually permits access to uninitialized memory.
Not great.
Another wrinkle is that it panics if the range has length 0.
However, I don't want to switch away from the `BoundedBuf` API, since
it's what tokio-uring uses.
We can always weed this out later by replacing `BoundedBuf` with our own
type.
Created an issue so we don't forget:
https://github.com/neondatabase/tokio-epoll-uring/issues/46
## Problem
`tokio::io::copy_bidirectional` doesn't close the connection once one of
the sides closes it. It's not really suitable for the postgres protocol.
## Summary of changes
Fork `copy_bidirectional` and initiate a shutdown for both connections.
---------
Co-authored-by: Conrad Ludgate <conradludgate@gmail.com>
The smaller changes I found while looking around #6584.
- rustfmt was not able to format handle_timeline_create
- fix Generation::get_suffix always allocating
- Generation was missing a `#[track_caller]` for panicky method
- attach has a lot of issues, but even with this PR it cannot be
formatted by rustfmt
- moved the `preload` span to be on top of `attach` -- it is awaited
inline
- make disconnected panic! or unreachable! into expect, expect_err
@problame noticed that the `tokio::sync::AcquireError` branch assertion
can be hit like in the added test. We haven't seen this yet in
production, but I'd prefer not to see it there. There `take_and_deinit`
is being used, but this race must be quite timing sensitive.
Rework of earlier: #6652.
VanillaPostgres constructor prints the "port={port}" line to the
config file, no need to do it in the callers.
The TODO comment that it would be nice if VanillaPostgres could pick
the port by itself is still valid though.
Commit 9a6c0be823 removed the code that printed these warnings:
marking {} as locally complete, while it doesnt exist in remote index
No timelines to attach received
Remove those warnings from all the allowlists in tests.
## Problem
When debugging/supporting this service, we sometimes need it to just
forget about a tenant or node, e.g. because of an issue cleanly tearing
them down. For example, if I create a tenant with a PlacementPolicy that
can't be scheduled on the nodes we have, we would never be able to
schedule it for a DELETE to work.
## Summary of changes
- Add APIs for dropping nodes and tenants that do no teardown other than
removing the entity from the DB and removing any references to it.
Turn the warning into an error, if there is garbage after the end of
imported tar file. However, it's normal for 'tar' to append extra
empty blocks to the end, so tolerate those without warnings or errors.
This PR reverts
- https://github.com/neondatabase/neon/pull/6589
- https://github.com/neondatabase/neon/pull/6652
because there's a performance regression that's particularly visible at
high layer counts.
Most likely it's because the switch to RwLock inflates the
```
inner: heavier_once_cell::OnceCell<ResidentOrWantedEvicted>,
```
size from 48 to 88 bytes, which, by itself is almost a doubling of the
cache footprint, and probably the fact that it's now larger than a cache
line also doesn't help.
See this chat on the Neon discord for more context:
https://discord.com/channels/1176467419317940276/1204714372295958548/1205541184634617906
I'm reverting 6652 as well because it might also have perf implications,
and we're getting close to the next release. We should re-do its changes
after the next release, though.
cc @koivunej
cc @ivaxer
## Problem
usernames and passwords can be URL 'percent' encoded in the connection
string URL provided by serverless driver.
## Summary of changes
Decode the parameters when getting conn info
Do list-delete operations in batches instead of doing full list first, to ensure
deletion makes progress even if there are a lot of files to remove.
To this end, add max_keys limit to remote storage list_files.
## Problem
Taking my ideas from https://github.com/neondatabase/neon/pull/6283 and
doing a bit less radical changes. smaller commits.
We currently don't report error classifications in proxy as the current
error handling made it hard to do so.
## Summary of changes
1. Add a `ReportableError` trait that all errors will implement. This
provides the error classification functionality.
2. Handle Client requests a strongly typed error
* this error is a `ReportableError` and is logged appropriately
3. The handle client error only has a few possible error types, to
account for the fact that at this point errors should be returned to the
user.
## Problem
In https://github.com/neondatabase/neon/pull/6637, we remove the need to
run migrations externally, but for compat tests to work we can't remove
those invocations from the neon_local binary.
Once that previous PR merges, we can make the followup changes without
upsetting compat tests.
The test was supposed to reproduce the bug fixed in commit 66fa176cc8,
i.e. that the clearing of the VM bit was not replayed in the
pageserver on HEAP_LOCK records. But it was broken in many ways and
failed to reproduce the original problem if you reverted the fix:
- The comparison of XIDs was broken. The test read the XID in to a
variable in python, but it was treated as a string rather than an
integer. As a result, e.g. "999" > "1000".
- The test accessed the locked tuple too early, in the loop. Accessing
it early, before the pg_xact page had been removed, set the hint bits.
That masked the problem on subsequent accesses.
- The on-demand SLRU download that was introduced in commit 9a9d9beaee
hid the issue. Even though an SLRU segment was removed by Postgres,
when it later tried to access it, it could still download it from
the pageserver. To ensure that doesn't happen, shorten the GC period
and compact and GC aggressively in the test.
I also added a more direct check that the VM page is updated, using
the get_page_at_lsn() debugging function. Right after locking the row,
we now fetch the VM page from pageserver and directly compare it with
the VM page in the page cache. They should match. That assertion is
more robust to things like on-demand SLRU download that could mask the
bug.
In neon_local, the default mode is now always 'fast', regardless of
'destroy'. You can override it with the "neon_local endpoint stop
--mode=immediate" flag.
In python tests, we still default to 'immediate' mode when using the
stop_and_destroy() function, and 'fast' with plain stop(). I kept that
to avoid changing behavior in existing tests. I don't think existing
tests depend on it, but I wasn't 100% certain.
## Problem
This test was a subset of the larger sharding test, and it missed the
validate() call on workload that was implicitly waiting for a tenant to
become active before trying to split it. It could therefore fail to
split due to tenant not yet being active.
## Summary of changes
- Insert .validate() call, and move the Workload setup to after the
check of shard ID (as the shard ID check should pass immediately)
This PR is preliminary cleanups and refactoring around `remote_storage`
for next PR which will move the timeouts and cancellation into
`remote_storage`.
Summary:
- smaller drive-by fixes
- code simplification
- refactor common parts like `DownloadError::is_permanent`
- align error types with `RemoteStorage::list_*` to use more
`download_retry` helper
Cc: #6096
if anon is in shared_preload_libraries.
Users cannot install it themselves, because superuser is required.
GRANT all priveleged needed to use it to db_owner
We use the neon fork of the extension, because small change to sql file
is needed to allow db_owner to use it.
This feature is behind a feature flag AnonExtension,
so it is not enabled by default.
## Problem
Drizzle needs to be able to configure the array_mode flag per query.
## Summary of changes
Adds an array_mode flag to the query data json that will otherwise
default to the header flag.
## Problem
Previous test started with a new-style TenantShardId with a non-zero
ShardCount. We also need to handle the case of a ShardCount() (aka
`unsharded`) parent shard.
**A followup PR will refactor ShardCount to make its inner value private
and thereby make this kind of mistake harder**
## Summary of changes
- Fix a place we were incorrectly treating a ShardCount as a number of
shards rather than as thing that can be zero or the number of shards.
- Add a test for this case.
This test occasionally fails with a difference in "pg_xact/0000" file
between the local and restored datadirs. My hypothesis is that something
changed in the database between the last explicit checkpoint and the
shutdown. I suspect autovacuum, it could certainly create transactions.
To fix, be more precise about the point in time that we compare. Shut
down the endpoint first, then read the last LSN (i.e. the shutdown
checkpoint's LSN), from the local disk with pg_controldata. And use
exactly that LSN in the basebackup.
Closes#559.
I'm proposing this as an alternative to
https://github.com/neondatabase/neon/pull/6662.
## Problem
test_lfc_resize sometimes filed with assertion failure when require lock
in write operation:
```
if (lfc_ctl->generation == generation)
{
Assert(LFC_ENABLED());
```
## Summary of changes
Increment generation when 0 is assigned to neon.file_cache_size_limit
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
@problame noticed that the `tokio::sync::AcquireError` branch assertion
can be hit like in the first commit. We haven't seen this yet in
production, but I'd prefer not to see it there. There `take_and_deinit`
is being used, but this race must be quite timing sensitive.
- Automatically set a node's availability to Active if it is responsive
in startup_reconcile
- Impose a 5s timeout of HTTP request to list location conf, so that an
unresponsive node can't hang it for minutes
- Do several retries if the request fails with a retryable error, to be
tolerant of concurrent pageserver & storage controller restarts
- Add a readiness hook for use with k8s so that we can tell when the
startup reconciliaton is done and the service is fully ready to do work.
- Add /metrics to the list of un-authenticated endpoints (this is
unrelated but we're touching the line in this PR already, and it fixes
auth error spam in deployed container.)
- A test for the above.
Closes: #6670
## Problem
One doesn't know at tenant creation time how large the tenant will grow.
We need to be able to dynamically adjust the shard count at runtime.
This is implemented as "splitting" of shards into smaller child shards,
which cover a subset of the keyspace that the parent covered.
Refer to RFC: https://github.com/neondatabase/neon/pull/6358
Part of epic: #6278
## Summary of changes
This PR implements the happy path (does not cleanly recover from a crash
mid-split, although won't lose any data), without any optimizations
(e.g. child shards re-download their own copies of layers that the
parent shard already had on local disk)
- Add `/v1/tenant/:tenant_shard_id/shard_split` API to pageserver: this
copies the shard's index to the child shards' paths, instantiates child
`Tenant` object, and tears down parent `Tenant` object.
- Add `splitting` column to `tenant_shards` table. This is written into
an existing migration because we haven't deployed yet, so don't need to
cleanly upgrade.
- Add `/control/v1/tenant/:tenant_id/shard_split` API to
attachment_service,
- Add `test_sharding_split_smoke` test. This covers the happy path:
future PRs will add tests that exercise failure cases.
## Problem
See #6626
If there is inactive replication slot then Postgres will not bw able to
shrink WAL and delete unused snapshots.
If she other active subscription is present, then snapshots created each
15 seconds will overflow AUX_DIR.
Setting `max_slot_wal_keep_size` doesn't solve the problem, because even
small WAL segment will be enough to overflow AUX_DIR if there is no
other activity on the system.
## Summary of changes
If there are active subscriptions and some logical replication slots are
not used during `neon.logical_replication_max_time_lag` interval, then
unused slot is dropped.
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
The password check logic for the sql-over-http is a bit non-intuitive.
## Summary of changes
1. Perform scram auth using the same logic as for websocket cleartext
password.
2. Split establish connection logic and connection pool.
3. Parallelize param parsing logic with authentication + wake compute.
4. Limit the total number of clients
## Problem
Copyright notice is outdated
## Summary of changes
Replace the initial year `2022` with `2022 - 2024`, after brief
discussion with Stas about the format
Co-authored-by: Andreas Scherbaum <andreas@neon.tech>
## Problem
The existing behavior isn't exactly incorrect, but is operationally
risky: if the control plane compute hook breaks, then all the control
plane operations trying to call /location_config will end up retrying
forever, which could put more load on the system.
## Summary of changes
- Treat 404s as fatal errors to do fewer retries: a 404 either indicates
we have the wrong URL, or some control plane bug is failing to recognize
our tenant ID as existing.
- Do not return an error on reconcilation errors in a non-creating
/location_config response: this allows the control plane to finish its
Operation (and we will eventually retry the compute notification later)
This PR adds an API to live-reconfigure the VirtualFile io engine.
It also adds a flag to `pagebench get-page-latest-lsn`, which is where I
found this functionality to be useful: it helps compare the io engines
in a benchmark without re-compiling a release build, which took ~50s on
the i3en.3xlarge where I was doing the benchmark.
Switching the IO engine is completely safe at runtime.
## Problem
We have finite amount of runners and intermediate results are often
wanted before a PR is ready for merging. Currently all PRs get e2e tests
run and this creates a lot of throwaway e2e results which may or may not
get to start or complete before a new push.
## Summary of changes
1. Skip e2e test when PR is in draft mode
2. Run e2e when PR status changes from draft to ready for review (change
this to having its trigger in below PR and update results of build and
test)
3. Abstract e2e test in a Separate workflow and call it from the main
workflow for the e2e test
5. Add a label, if that label is present run e2e test in draft
(run-e2e-test-in-draft)
6. Auto add a label(approve to ci) so that all the external contributors
PR , e2e run in draft
7. Document the new label changes and the above behaviour
Draft PR : https://github.com/neondatabase/neon/actions/runs/7729128470
Ready To Review :
https://github.com/neondatabase/neon/actions/runs/7733779916
Draft PR with label :
https://github.com/neondatabase/neon/actions/runs/7725691012/job/21062432342
and https://github.com/neondatabase/neon/actions/runs/7733854028
## Checklist before requesting a review
- [x] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Alexander Bayandin <alexander@neon.tech>
2024-02-07 16:14:10 +00:00
618 changed files with 74272 additions and 24633 deletions
gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=false
# TODO: move deployPreprodRegion to release (`"$GITHUB_REF_NAME" == "release"` block), once Staging support different compute tag prefixes for different regions
gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}} -f deployPreprodRegion=true
elif [[ "$GITHUB_REF_NAME" == "release" ]]; then
gh workflow --repo neondatabase/aws run deploy-prod.yml --ref main -f branch=main -f dockerTag=${{needs.tag.outputs.build-tag}}
gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main \
-f deployPgSniRouter=false \
-f deployProxy=false \
-f deployStorage=true \
-f deployStorageBroker=true \
-f deployStorageController=true \
-f branch=main \
-f dockerTag=${{needs.tag.outputs.build-tag}} \
-f deployPreprodRegion=true
gh workflow --repo neondatabase/aws run deploy-prod.yml --ref main \
-f deployStorage=true \
-f deployStorageBroker=true \
-f deployStorageController=true \
-f branch=main \
-f dockerTag=${{needs.tag.outputs.build-tag}}
elif [[ "$GITHUB_REF_NAME" == "release-proxy" ]]; then
gh workflow --repo neondatabase/aws run deploy-dev.yml --ref main \
-f deployPgSniRouter=true \
-f deployProxy=true \
-f deployStorage=false \
-f deployStorageBroker=false \
-f deployStorageController=false \
-f branch=main \
-f dockerTag=${{needs.tag.outputs.build-tag}} \
-f deployPreprodRegion=true
gh workflow --repo neondatabase/aws run deploy-proxy-prod.yml --ref main \
-f deployPgSniRouter=true \
-f deployProxy=true \
-f branch=main \
-f dockerTag=${{needs.tag.outputs.build-tag}}
else
echo "GITHUB_REF_NAME (value '$GITHUB_REF_NAME') is not set to either 'main' or 'release'"
storage_broker={version="0.1",path="./storage_broker/"}# Note: main broker code is inside the binary crate, so linking with the library shouldn't be heavy.
RUN curl -sL "https://github.com/peak/s5cmd/releases/download/v${S5CMD_VERSION}/s5cmd_${S5CMD_VERSION}_Linux-$(uname -m | sed 's/x86_64/64bit/g'| sed 's/aarch64/arm64/g').tar.gz"| tar zxvf - s5cmd \
&& chmod +x s5cmd \
&& mv s5cmd /usr/local/bin/s5cmd
# LLVM
ENVLLVM_VERSION=17
ENVLLVM_VERSION=18
RUN curl -fsSL 'https://apt.llvm.org/llvm-snapshot.gpg.key'| apt-key add - \
Neon is a serverless open-source alternative to AWS Aurora Postgres. It separates storage and compute and substitutes the PostgreSQL storage layer by redistributing data across a cluster of nodes.
## Quick start
Try the [Neon Free Tier](https://neon.tech/docs/introduction/technical-preview-free-tier/) to create a serverless Postgres instance. Then connect to it with your preferred Postgres client (psql, dbeaver, etc) or use the online [SQL Editor](https://neon.tech/docs/get-started-with-neon/query-with-neon-sql-editor/). See [Connect from any application](https://neon.tech/docs/connect/connect-from-any-app/) for connection instructions.
Try the [Neon Free Tier](https://neon.tech/github) to create a serverless Postgres instance. Then connect to it with your preferred Postgres client (psql, dbeaver, etc) or use the online [SQL Editor](https://neon.tech/docs/get-started-with-neon/query-with-neon-sql-editor/). See [Connect from any application](https://neon.tech/docs/connect/connect-from-any-app/) for connection instructions.
Alternatively, compile and run the project [locally](#running-local-installation).
@@ -230,8 +230,22 @@ postgres=# select * from t;
> cargo neon stop
```
More advanced usages can be found at [Control Plane and Neon Local](./control_plane/README.md).
#### Handling build failures
If you encounter errors during setting up the initial tenant, it's best to stop everything (`cargo neon stop`) and remove the `.neon` directory. Then fix the problems, and start the setup again.
## Running tests
### Rust unit tests
We are using [`cargo-nextest`](https://nexte.st/) to run the tests in Github Workflows.
Some crates do not support running plain `cargo test` anymore, prefer `cargo nextest run` instead.
You can install `cargo-nextest` with `cargo install cargo-nextest`.
### Integration tests
Ensure your dependencies are installed as described [here](https://github.com/neondatabase/neon#dependency-installation-notes).
```sh
@@ -249,6 +263,22 @@ testing locally, it is convenient to run just one set of permutations, like this
You may find yourself in need of flamegraphs for software in this repository.
You can use [`flamegraph-rs`](https://github.com/flamegraph-rs/flamegraph) or the original [`flamegraph.pl`](https://github.com/brendangregg/FlameGraph). Your choice!
>[!IMPORTANT]
> If you're using `lld` or `mold`, you need the `--no-rosegment` linker argument.
> It's a [general thing with Rust / lld / mold](https://crbug.com/919499#c16), not specific to this repository.
> See [this PR for further instructions](https://github.com/neondatabase/neon/pull/6764).
## Cleanup
For cleaning up the source tree from build artifacts, run `make clean` in the source directory.
For removing every artifact from build and configure steps, run `make distclean`, and also consider removing the cargo binaries in the `target` directory, as well as the database in the `.neon` directory. Note that removing the `.neon` directory will remove your database, with all data in it. You have been warned!
## Documentation
[docs](/docs) Contains a top-level overview of all available markdown documentation.
By default, `cargo neon` starts an endpoint with `cloud_admin` and `postgres` database. If you want to have a role and a database similar to what we have on the cloud service, you can do it with the following commands when starting an endpoint.
```shell
cargo neon endpoint create main --pg-version 16 --update-catalog true
cargo neon endpoint start main --create-test-user true
```
The first command creates `neon_superuser` and necessary roles. The second command creates `test` user and `neondb` database. You will see a connection string that connects you to the test user after running the second command.
.help("How many pageservers to create (default 1)")
.required(false)
.default_value("1");
.help("How many pageservers to create (default 1)");
letupdate_catalog=Arg::new("update-catalog")
.value_parser(value_parser!(bool))
.long("update-catalog")
.help("If set, will set up the catalog for neon_superuser")
.required(false);
letcreate_test_user=Arg::new("create-test-user")
.value_parser(value_parser!(bool))
.long("create-test-user")
.help("If set, will create test user `user` and `neondb` database. Requires `update-catalog = true`")
.required(false);
letallow_multiple=Arg::new("allow-multiple")
.help("Allow multiple primary endpoints running on the same branch. Shouldn't be used normally, but useful for tests.")
.long("allow-multiple")
.action(ArgAction::SetTrue)
.required(false);
Command::new("Neon CLI")
.arg_required_else_help(true)
@@ -1442,14 +1419,13 @@ fn cli() -> Command {
.subcommand(
Command::new("init")
.about("Initialize a new Neon repository, preparing configs for services to start with")
.arg(pageserver_config_args.clone())
.arg(num_pageservers_arg.clone())
.arg(
Arg::new("config")
.long("config")
.required(false)
.value_parser(value_parser!(PathBuf))
.value_name("config"),
.value_name("config")
)
.arg(pg_version_arg.clone())
.arg(force_arg)
@@ -1457,6 +1433,7 @@ fn cli() -> Command {
.subcommand(
Command::new("timeline")
.about("Manage timelines")
.arg_required_else_help(true)
.subcommand(Command::new("list")
.about("List all timelines, available to this pageserver")
.arg(tenant_id_arg.clone()))
@@ -1496,6 +1473,7 @@ fn cli() -> Command {
.arg(Arg::new("end-lsn").long("end-lsn")
.help("Lsn the basebackup ends at"))
.arg(pg_version_arg.clone())
.arg(update_catalog.clone())
)
).subcommand(
Command::new("tenant")
@@ -1511,19 +1489,15 @@ fn cli() -> Command {
.help("Use this tenant in future CLI commands where tenant_id is needed, but not specified"))
.arg(Arg::new("shard-count").value_parser(value_parser!(u8)).long("shard-count").action(ArgAction::Set).help("Number of shards in the new tenant (default 1)"))
.arg(Arg::new("shard-stripe-size").value_parser(value_parser!(u32)).long("shard-stripe-size").action(ArgAction::Set).help("Sharding stripe size in pages"))
.arg(Arg::new("placement-policy").value_parser(value_parser!(String)).long("placement-policy").action(ArgAction::Set).help("Placement policy shards in this tenant"))
assert_eq!(&PageServerConf::from(&conf),&self.conf,"during neon_local init, we derive the runtime state of ps conf (self.conf) from the --config flag fully");
// TODO(christian): instead of what we do here, create a pageserver_api::config::ConfigToml (PR #7656)
// FIXME: the paths should be shell-escaped to handle paths with spaces, quotas etc.
Zenith CLI as it is described here mostly resides on the same conceptual level as pg_ctl/initdb/pg_recvxlog/etc and replaces some of them in an opinionated way. I would also suggest bundling our patched postgres inside zenith distribution at least at the start.
Neon CLI as it is described here mostly resides on the same conceptual level as pg_ctl/initdb/pg_recvxlog/etc and replaces some of them in an opinionated way. I would also suggest bundling our patched postgres inside neon distribution at least at the start.
This proposal is focused on managing local installations. For cluster operations, different tooling would be needed. The point of integration between the two is storage URL: no matter how complex cluster setup is it may provide an endpoint where the user may push snapshots.
@@ -8,40 +8,40 @@ The most important concept here is a snapshot, which can be created/pushed/pulle
Since we may export the whole snapshot as one big file (tar of basebackup, maybe with some manifest) it may be shared over conventional means: http, ssh, [git+lfs](https://docs.github.com/en/github/managing-large-files/about-git-large-file-storage).
```
> zenith pg create --snapshot http://learn-postgres.com/movies_db.zenith movies
> neon pg create --snapshot http://learn-postgres.com/movies_db.neon movies
One way to rollback the database is just to init a new database from the snapshot and destroy the old one. But creating a new database from a snapshot would require a copy of that snapshot which is time consuming operation. Another option that would be cool to support is the ability to create the copy-on-write database from the snapshot without copying data, and store updated pages in a separate location, however that way would have performance implications. So to properly rollback the database to the older state we have `zenith pg checkout`.
One way to rollback the database is just to init a new database from the snapshot and destroy the old one. But creating a new database from a snapshot would require a copy of that snapshot which is time consuming operation. Another option that would be cool to support is the ability to create the copy-on-write database from the snapshot without copying data, and store updated pages in a separate location, however that way would have performance implications. So to properly rollback the database to the older state we have `neon pg checkout`.
```
> zenith pg list
> neon pg list
ID PGDATA USED STORAGE ENDPOINT
primary1 pgdata1 5G zenith-local localhost:5432
primary1 pgdata1 5G neon-local localhost:5432
> zenith snapshot create pgdata1@snap1
> neon snapshot create pgdata1@snap1
> zenith snapshot list
> neon snapshot list
ID SIZE PARENT
oldpg 5G -
pgdata1@snap1 6G -
pgdata1@CURRENT 6G -
> zenith pg checkout pgdata1@snap1
> neon pg checkout pgdata1@snap1
Stopping postgres on pgdata1.
Rolling back pgdata1@CURRENT to pgdata1@snap1.
Starting postgres on pgdata1.
> zenith snapshot list
> neon snapshot list
ID SIZE PARENT
oldpg 5G -
pgdata1@snap1 6G -
@@ -99,7 +99,7 @@ Some notes: pgdata1@CURRENT -- implicit snapshot representing the current state
PITR area acts like a continuous snapshot where you can reset the database to any point in time within this area (by area I mean some TTL period or some size limit, both possibly infinite).
Resetting the database to some state in past would require creating a snapshot on some lsn / time in this pirt area.
@@ -108,29 +108,29 @@ Resetting the database to some state in past would require creating a snapshot o
## storage
Storage is either zenith pagestore or s3. Users may create a database in a pagestore and create/move *snapshots* and *pitr regions* in both pagestore and s3. Storage is a concept similar to `git remote`. After installation, I imagine one local storage is available by default.
Storage is either neon pagestore or s3. Users may create a database in a pagestore and create/move *snapshots* and *pitr regions* in both pagestore and s3. Storage is a concept similar to `git remote`. After installation, I imagine one local storage is available by default.
**zenith storage attach** -t [native|s3] -c key=value -n name
**neon storage attach** -t [native|s3] -c key=value -n name
Attaches/initializes storage. For --type=s3, user credentials and path should be provided. For --type=native we may support --path=/local/path and --url=zenith.tech/stas/mystore. Other possible term for native is 'zstore'.
Attaches/initializes storage. For --type=s3, user credentials and path should be provided. For --type=native we may support --path=/local/path and --url=neon.tech/stas/mystore. Other possible term for native is 'zstore'.
@@ -140,29 +140,29 @@ Manages postgres data directories and can start postgres instances with proper c
Pg is a term for a single postgres running on some data. I'm trying to avoid separation of datadir management and postgres instance management -- both that concepts bundled here together.
Creates (initializes) new data directory in given storage and starts postgres. I imagine that storage for this operation may be only local and data movement to remote location happens through snapshots/pitr.
--no-start: just init datadir without creating
--snapshot snap: init from the snapshot. Snap is a name or URL (zenith.tech/stas/mystore/snap1)
--snapshot snap: init from the snapshot. Snap is a name or URL (neon.tech/stas/mystore/snap1)
--cow: initialize Copy-on-Write data directory on top of some snapshot (makes sense if it is a snapshot of currently running a database)
**zenith pg destroy**
**neon pg destroy**
**zenith pg start** [--replica] pgdata
**neon pg start** [--replica] pgdata
Start postgres with proper extensions preloaded/installed.
**zenith pg checkout**
**neon pg checkout**
Rollback data directory to some previous snapshot.
Starts REST/GraphQL proxy on top of postgres master. Not sure we should do that, just an idea.
@@ -203,35 +203,35 @@ Starts REST/GraphQL proxy on top of postgres master. Not sure we should do that,
Snapshot creation is cheap -- no actual data is copied, we just start retaining old pages. Snapshot size means the amount of retained data, not all data. Snapshot name looks like pgdata_name@tag_name. tag_name is set by the user during snapshot creation. There are some reserved tag names: CURRENT represents the current state of the data directory; HEAD{i} represents the data directory state that resided in the database before i-th checkout.
**zenith snapshot create** pgdata_name@snap_name
**neon snapshot create** pgdata_name@snap_name
Creates a new snapshot in the same storage where pgdata_name exists.
Produces binary stream of a given snapshot. Under the hood starts temp read-only postgres over this snapshot and sends basebackup stream. Receiving side should start `zenith snapshot recv` before push happens. If url has some special schema like zenith:// receiving side may require auth start `zenith snapshot recv` on the go.
Produces binary stream of a given snapshot. Under the hood starts temp read-only postgres over this snapshot and sends basebackup stream. Receiving side should start `neon snapshot recv` before push happens. If url has some special schema like neon:// receiving side may require auth start `neon snapshot recv` on the go.
**zenith snapshot recv**
**neon snapshot recv**
Starts a port listening for a basebackup stream, prints connection info to stdout (so that user may use that in push command), and expects data on that socket.
**zenith snapshot pull** --from url or path
**neon snapshot pull** --from url or path
Connects to a remote zenith/s3/file and pulls snapshot. The remote site should be zenith service or files in our format.
Connects to a remote neon/s3/file and pulls snapshot. The remote site should be neon service or files in our format.
**zenith snapshot import** --from basebackup://<...> or path
**neon snapshot import** --from basebackup://<...> or path
Creates a new snapshot out of running postgres via basebackup protocol or basebackup files.
**zenith snapshot export**
**neon snapshot export**
Starts read-only postgres over this snapshot and exports data in some format (pg_dump, or COPY TO on some/all tables). One of the options may be zenith own format which is handy for us (but I think just tar of basebackup would be okay).
Starts read-only postgres over this snapshot and exports data in some format (pg_dump, or COPY TO on some/all tables). One of the options may be neon own format which is handy for us (but I think just tar of basebackup would be okay).
**zenith snapshot diff** snap1 snap2
**neon snapshot diff** snap1 snap2
Shows size of data changed between two snapshots. We also may provide options to diff schema/data in tables. To do that start temp read-only postgreses.
**zenith snapshot destroy**
**neon snapshot destroy**
## pitr
@@ -239,7 +239,7 @@ Pitr represents wal stream and ttl policy for that stream
XXX: any suggestions on a better name?
**zenith pitr create** name
**neon pitr create** name
--ttl = inf | period
@@ -247,21 +247,21 @@ XXX: any suggestions on a better name?
Here I list some objectives to keep in mind when discussing zenith-local design and a proposal that brings all components together. Your comments on both parts are very welcome.
Here I list some objectives to keep in mind when discussing neon-local design and a proposal that brings all components together. Your comments on both parts are very welcome.
#### Why do we need it?
- For distribution - this easy to use binary will help us to build adoption among developers.
- For internal use - to test all components together.
In my understanding, we consider it to be just a mock-up version of zenith-cloud.
In my understanding, we consider it to be just a mock-up version of neon-cloud.
> Question: How much should we care about durability and security issues for a local setup?
#### Why is it better than a simple local postgres?
- Easy one-line setup. As simple as `cargo install zenith && zenith start`
- Easy one-line setup. As simple as `cargo install neon && neon start`
- Quick and cheap creation of compute nodes over the same storage.
> Question: How can we describe a use-case for this feature?
-Zenith-local can work with S3 directly.
-Neon-local can work with S3 directly.
- Push and pull images (snapshots) to remote S3 to exchange data with other users.
@@ -31,50 +31,50 @@ Ideally, just one binary that incorporates all elements we need.
#### Components:
- **zenith-CLI** - interface for end-users. Turns commands to REST requests and handles responses to show them in a user-friendly way.
CLI proposal is here https://github.com/libzenith/rfcs/blob/003-laptop-cli.md/003-laptop-cli.md
WIP code is here: https://github.com/libzenith/postgres/tree/main/pageserver/src/bin/cli
- **neon-CLI** - interface for end-users. Turns commands to REST requests and handles responses to show them in a user-friendly way.
CLI proposal is here https://github.com/neondatabase/rfcs/blob/003-laptop-cli.md/003-laptop-cli.md
WIP code is here: https://github.com/neondatabase/postgres/tree/main/pageserver/src/bin/cli
- **zenith-console** - WEB UI with same functionality as CLI.
- **neon-console** - WEB UI with same functionality as CLI.
>Note: not for the first release.
- **zenith-local** - entrypoint. Service that starts all other components and handles REST API requests. See REST API proposal below.
> Idea: spawn all other components as child processes, so that we could shutdown everything by stopping zenith-local.
- **neon-local** - entrypoint. Service that starts all other components and handles REST API requests. See REST API proposal below.
> Idea: spawn all other components as child processes, so that we could shutdown everything by stopping neon-local.
- **zenith-pageserver** - consists of a storage and WAL-replaying service (modified PG in current implementation).
- **neon-pageserver** - consists of a storage and WAL-replaying service (modified PG in current implementation).
> Question: Probably, for local setup we should be able to bypass page-storage and interact directly with S3 to avoid double caching in shared buffers and page-server?
WIP code is here: https://github.com/libzenith/postgres/tree/main/pageserver/src
WIP code is here: https://github.com/neondatabase/postgres/tree/main/pageserver/src
- **zenith-S3** - stores base images of the database and WAL in S3 object storage. Import and export images from/to zenith.
- **neon-S3** - stores base images of the database and WAL in S3 object storage. Import and export images from/to neon.
> Question: How should it operate in a local setup? Will we manage it ourselves or ask user to provide credentials for existing S3 object storage (i.e. minio)?
> Question: Do we use it together with local page store or they are interchangeable?
WIP code is ???
- **zenith-safekeeper** - receives WAL from postgres, stores it durably, answers to Postgres that "sync" is succeed.
- **neon-safekeeper** - receives WAL from postgres, stores it durably, answers to Postgres that "sync" is succeed.
> Question: How should it operate in a local setup? In my understanding it should push WAL directly to S3 (if we use it) or store all data locally (if we use local page storage). The latter option seems meaningless (extra overhead and no gain), but it is still good to test the system.
WIP code is here: https://github.com/libzenith/postgres/tree/main/src/bin/safekeeper
WIP code is here: https://github.com/neondatabase/postgres/tree/main/src/bin/safekeeper
- **zenith-computenode** - bottomless PostgreSQL, ideally upstream, but for a start - our modified version. User can quickly create and destroy them and work with it as a regular postgres database.
- **neon-computenode** - bottomless PostgreSQL, ideally upstream, but for a start - our modified version. User can quickly create and destroy them and work with it as a regular postgres database.
WIP code is in main branch and here: https://github.com/libzenith/postgres/commits/compute_node
WIP code is in main branch and here: https://github.com/neondatabase/postgres/commits/compute_node
#### REST API:
Service endpoint: `http://localhost:3000`
Resources:
- /storages - Where data lives: zenith-pageserver or zenith-s3
- /pgs - Postgres - zenith-computenode
- /storages - Where data lives: neon-pageserver or neon-s3
- /pgs - Postgres - neon-computenode
- /snapshots - snapshots **TODO**
>Question: Do we want to extend this API to manage zenith components? I.e. start page-server, manage safekeepers and so on? Or they will be hardcoded to just start once and for all?
>Question: Do we want to extend this API to manage neon components? I.e. start page-server, manage safekeepers and so on? Or they will be hardcoded to just start once and for all?
Methods and their mapping to CLI:
- /storages - zenith-pageserver or zenith-s3
- /storages - neon-pageserver or neon-s3
CLI | REST API
------------- | -------------
@@ -84,7 +84,7 @@ storage list | GET /storages
storage show -n name | GET /storages/:storage_name
Zenith CLI allows you to operate database clusters (catalog clusters) and their commit history locally and in the cloud. Since ANSI calls them catalog clusters and cluster is a loaded term in the modern infrastructure we will call it "catalog".
Neon CLI allows you to operate database clusters (catalog clusters) and their commit history locally and in the cloud. Since ANSI calls them catalog clusters and cluster is a loaded term in the modern infrastructure we will call it "catalog".
# CLI v2 (after chatting with Carl)
Zenith introduces the notion of a repository.
Neon introduces the notion of a repository.
```bash
zenith init
zenith clone zenith://zenith.tech/piedpiper/northwind -- clones a repo to the northwind directory
neon init
neon clone neon://neon.tech/piedpiper/northwind -- clones a repo to the northwind directory
```
Once you have a cluster catalog you can explore it
```bash
zenith log -- returns a list of commits
zenith status -- returns if there are changes in the catalog that can be committed
zenith commit -- commits the changes and generates a new commit hash
zenith branch experimental <hash> -- creates a branch called testdb based on a given commit hash
neon log -- returns a list of commits
neon status -- returns if there are changes in the catalog that can be committed
neon commit -- commits the changes and generates a new commit hash
neon branch experimental <hash> -- creates a branch called testdb based on a given commit hash
```
To make changes in the catalog you need to run compute nodes
```bash
-- here is how you a compute node
zenith start /home/pipedpiper/northwind:main -- starts a compute instance
zenith start zenith://zenith.tech/northwind:main -- starts a compute instance in the cloud
neon start /home/pipedpiper/northwind:main -- starts a compute instance
neon start neon://neon.tech/northwind:main -- starts a compute instance in the cloud
-- you can start a compute node against any hash or branch
zenith start /home/pipedpiper/northwind:experimental --port 8008 -- start another compute instance (on different port)
neon start /home/pipedpiper/northwind:experimental --port 8008 -- start another compute instance (on different port)
-- you can start a compute node against any hash or branch
zenith start /home/pipedpiper/northwind:<hash> --port 8009 -- start another compute instance (on different port)
neon start /home/pipedpiper/northwind:<hash> --port 8009 -- start another compute instance (on different port)
-- After running some DML you can run
-- zenith status and see how there are two WAL streams one on top of
-- neon status and see how there are two WAL streams one on top of
@@ -7,22 +7,22 @@ Here is a proposal about implementing push/pull mechanics between pageservers. W
The origin represents connection info for some remote pageserver. Let's use here same commands as git uses except using explicit list subcommand (git uses `origin -v` for that).
```
zenith origin add <name> <connection_uri>
zenith origin list
zenith origin remove <name>
neon origin add <name> <connection_uri>
neon origin list
neon origin remove <name>
```
Connection URI a string of form `postgresql://user:pass@hostname:port` (https://www.postgresql.org/docs/13/libpq-connect.html#id-1.7.3.8.3.6). We can start with libpq password auth and later add support for client certs or require ssh as transport or invent some other kind of transport.
Behind the scenes, this commands may update toml file inside .zenith directory.
Behind the scenes, this commands may update toml file inside .neon directory.
## Push
### Pushing branch
```
zenith push mybranch cloudserver # push to eponymous branch in cloudserver
zenith push mybranch cloudserver:otherbranch # push to a different branch in cloudserver
neon push mybranch cloudserver # push to eponymous branch in cloudserver
neon push mybranch cloudserver:otherbranch # push to a different branch in cloudserver
```
Exact mechanics would be slightly different in the following situations:
@@ -2,7 +2,7 @@ While working on export/import commands, I understood that they fit really well
We may think about backups as snapshots in a different format (i.e plain pgdata format, basebackup tar format, WAL-G format (if they want to support it) and so on). They use same storage API, the only difference is the code that packs/unpacks files.
Even if zenith aims to maintains durability using it's own snapshots, backups will be useful for uploading data from postgres to zenith.
Even if neon aims to maintains durability using it's own snapshots, backups will be useful for uploading data from postgres to neon.
So here is an attempt to design consistent CLI for different usage scenarios:
@@ -16,8 +16,8 @@ Save`storage_dest` and other parameters in config.
Push snapshots to `storage_dest` in background.
```
zenith init --storage_dest=S3_PREFIX
zenith start
neon init --storage_dest=S3_PREFIX
neon start
```
#### 2. Restart pageserver (manually or crash-recovery).
@@ -25,7 +25,7 @@ Take `storage_dest` from pageserver config, start pageserver from latest snapsho
Push snapshots to `storage_dest` in background.
```
zenith start
neon start
```
#### 3. Import.
@@ -35,22 +35,22 @@ Do not save `snapshot_path` and `snapshot_format` in config, as it is a one-time
Save`storage_dest` parameters in config.
Push snapshots to `storage_dest` in background.
```
//I.e. we want to start zenith on top of existing $PGDATA and use s3 as a persistent storage.
Initially created [here](https://github.com/zenithdb/rfcs/pull/16) by @kelvich.
Initially created [here](https://github.com/neondatabase/rfcs/pull/16) by @kelvich.
That it is an alternative to (014-safekeeper-gossip)[]
@@ -292,4 +292,4 @@ But with an etcd we are in a bit different situation:
1. We don't need persistency and strong consistency guarantees for the data we store in the etcd
2. etcd uses Grpc as a protocol, and messages are pretty simple
So it looks like implementing in-mem store with etcd interface is straightforward thing _if we will want that in future_. At the same time, we can avoid implementing it right now, and we will be able to run local zenith installation with etcd running somewhere in the background (as opposed to building and running console, which in turn requires Postgres).
So it looks like implementing in-mem store with etcd interface is straightforward thing _if we will want that in future_. At the same time, we can avoid implementing it right now, and we will be able to run local neon installation with etcd running somewhere in the background (as opposed to building and running console, which in turn requires Postgres).
_Note: keys for relation metadata are ignored here, as this data will be mirrored to all
shards. For distribution purposes, we only care about user data keys_
The properties we want from our Key->Shard mapping are:
- Locality in `blknum`, such that adjacent `blknum` will usually map to
the same stripe and consequently land on the same shard, even though the overall
collection of blocks in a relation will be spread over many stripes and therefore
many shards.
- Avoid the same blknum on different relations landing on the same stripe, so that
with many small relations we do not end up aliasing data to the same stripe/shard.
- Avoid vulnerability to aliasing in the values of relation identity fields, such that
if there are patterns in the value of `relnode`, these do not manifest as patterns
in data placement.
To accomplish this, the blknum is used to select a stripe, and stripes are
assigned to shards in a pseudorandom order via a hash. The motivation for
pseudo-random distribution (rather than sequential mapping of stripe to shard)
is to avoid I/O hotspots when sequentially reading multiple relations: we don't want
all relations' stripes to touch pageservers in the same order.
To map a `Key` to a shard:
- Hash the `Key` field 4 (relNode).
- Divide field 6 (`blknum`) field by the stripe size in pages, and combine the
hash of this with the hash from the previous step.
- The total hash modulo the shard count gives the shard holding this key.
Why don't we use the other fields in the Key?
- We ignore `forknum` for key mapping, because it distinguishes different classes of data
in the same relation, and we would like to keep the data in a relation together.
- We would like to use spcNode and dbNode, but cannot. Postgres database creation operations can refer to an existing database as a template, such that the created
database's blocks differ only by spcNode and dbNode from the original. To enable running
this type of creation without cross-pageserver communication, we must ensure that these
blocks map to the same shard -- we do this by excluding spcNode and dbNode from the hash.
### Data placement examples
For example, consider the extreme large databases cases of postgres data layout in a system with 8 shards
and a stripe size of 32k pages:
- A single large relation: `blknum` division will break the data up into 4096
stripes, which will be scattered across the shards.
- 4096 relations of of 32k pages each: each relation will map to exactly one stripe,
and that stripe will be placed according to the hash of the key fields 4. The
data placement will be statistically uniform across shards.
Data placement will be more uneven on smaller databases:
- A tenant with 2 shards and 2 relations of one stripe size each: there is a 50% chance
that both relations land on the same shard and no data lands on the other shard.
- A tenant with 8 shards and one relation of size 12 stripes: 4 shards will have double
the data of the other four shards.
These uneven cases for small amounts of data do not matter, as long as the stripe size
is an order of magnitude smaller than the amount of data we are comfortable holding
in a single shard: if our system handles shard sizes up to 10-100GB, then it is not an issue if
a tenant has some shards with 256MB size and some shards with 512MB size, even though
the standard deviation of shard size within the tenant is very high. Our key mapping
scheme provides a statistical guarantee that as the tenant's overall data size increases,
uniformity of placement will improve.
### Important Types
#### `ShardIdentity`
Provides the information needed to know whether a particular key belongs
to a particular shard:
- Layout version
- Stripe size
- Shard count
- Shard index
This structure's size is constant. Note that if we had used a differnet key
mapping scheme such as consistent hashing with explicit hash ranges assigned
to each shard, then the ShardIdentity's size would grow with the shard count: the simpler
key mapping scheme used here enables a small fixed size ShardIdentity.
### Pageserver changes
#### Structural
Everywhere the Pageserver currently deals with Tenants, it will move to dealing with
`TenantShard`s, which are just a `Tenant` plus a `ShardIdentity` telling it which part
of the keyspace it owns. An un-sharded tenant is just a `TenantShard` whose `ShardIdentity`
covers the whole keyspace.
When the pageserver writes layers and index_part.json to remote storage, it must
include the shard index & count in the name, to avoid collisions (the count is
necessary for future-proofing: the count will vary in time). These keys
will also include a generation number: the [generation numbers](025-generation-numbers.md) system will work
exactly the same for TenantShards as it does for Tenants today: each shard will have
its own generation number.
#### Storage Format: Keys
For tenants with >1 shard, layer files implicitly become sparse: within the key
range described in the layer name, the layer file for a shard will only hold the
content relevant to stripes assigned to the shard.
For this reason, the LayerFileName within a tenant is no longer unique: different shards
may use the same LayerFileName to refer to different data. We may solve this simply
by including the shard number in the keys used for layers.
The shard number will be included as a prefix (as part of tenant ID), like this:
This RFC describes a new pageserver API for splitting an existing tenant shard into
multiple shards, and describes how to use this API to safely increase the total
shard count of a tenant.
## Motivation
In the [sharding RFC](031-sharding-static.md), a mechanism was introduced to scale
tenants beyond the capacity of a single pageserver by breaking up the key space
into stripes, and distributing these stripes across many pageservers. However,
the shard count was defined once at tenant creation time and not varied thereafter.
In practice, the expected size of a database is rarely known at creation time, and
it is inefficient to enable sharding for very small tenants: we need to be
able to create a tenant with a small number of shards (such as 1), and later expand
when it becomes clear that the tenant has grown in size to a point where sharding
is beneficial.
### Prior art
Many distributed systems have the problem of choosing how many shards to create for
tenants that do not specify an expected size up-front. There are a couple of general
approaches:
- Write to a key space in order, and start a new shard when the highest key advances
past some point. This doesn't work well for Neon, because we write to our key space
in many different contiguous ranges (per relation), rather than in one contiguous
range. To adapt to this kind of model, we would need a sharding scheme where each
relation had its own range of shards, which would be inefficient for the common
case of databases with many small relations.
- Monitor the system, and automatically re-shard at some size threshold. For
example in Ceph, the [pg_autoscaler](https://github.com/ceph/ceph/blob/49c27499af4ee9a90f69fcc6bf3597999d6efc7b/src/pybind/mgr/pg_autoscaler/module.py)
component monitors the size of each RADOS Pool, and adjusts the number of Placement
Groups (Ceph's shard equivalent).
## Requirements
- A configurable capacity limit per-shard is enforced.
- Changes in shard count do not interrupt service beyond requiring postgres
to reconnect (i.e. milliseconds).
- Human being does not have to choose shard count
## Non Goals
- Shard splitting is always a tenant-global operation: we will not enable splitting
one shard while leaving others intact.
- The inverse operation (shard merging) is not described in this RFC. This is a lower
priority than splitting, because databases grow more often than they shrink, and
a database with many shards will still work properly if the stored data shrinks, just
with slightly more overhead (e.g. redundant WAL replication)
- Shard splitting is only initiated based on capacity bounds, not load. Splitting
a tenant based on load will make sense for some medium-capacity, high-load workloads,
but is more complex to reason about and likely is not desirable until we have
shard merging to reduce the shard count again if the database becomes less busy.
## Impacted Components
pageserver, storage controller
(the _storage controller_ is the evolution of what was called `attachment_service` in our test environment)
## Terminology
**Parent** shards are the shards that exist before a split. **Child** shards are
the new shards created during a split.
**Shard** is synonymous with _tenant shard_.
**Shard Index** is the 2-tuple of shard number and shard count, written in
paths as {:02x}{:02x}, e.g. `0001`.
## Background
In the implementation section, a couple of existing aspects of sharding are important
to remember:
- Shard identifiers contain the shard number and count, so that "shard 0 of 1" (`0001`) is
a distinct shard from "shard 0 of 2" (`0002`). This is the case in key paths, local
storage paths, and remote index metadata.
- Remote layer file paths contain the shard index of the shard that created them, and
remote indices contain the same index to enable building the layer file path. A shard's
index may reference layers that were created by another shard.
- Local tenant shard directories include the shard index. All layers downloaded by
a tenant shard are stored in this shard-prefixed path, even if those layers were
initially created by another shard: tenant shards do not read and write one anothers'
paths.
- The `Tenant` pageserver type represents one tenant _shard_, not the whole tenant.
This is for historical reasons and will be cleaned up in future, but the existing
name is used here to help comprehension when reading code.
## Implementation
Note: this section focuses on the correctness of the core split process. This will
be fairly inefficient in a naive implementation, and several important optimizations
are described in a later section.
There are broadly two parts to the implementation:
1. The pageserver split API, which splits one shard on one pageserver
2. The overall tenant split proccess which is coordinated by the storage controller,
and calls into the pageserver split API as needed.
### Pageserver Split API
The pageserver will expose a new API endpoint at `/v1/tenant/:tenant_shard_id/shard_split`
that takes the new total shard count in the body.
The pageserver split API operates on one tenant shard, on one pageserver. External
coordination is required to use it safely, this is described in the later
'Split procedure' section.
#### Preparation
First identify the shard indices for the new child shards. These are deterministic,
calculated from the parent shard's index, and the number of children being created (this
is an input to the API, and validated to be a power of two). In a trivial example, splitting
0001 in two always results in 0002 and 0102.
Child shard indices are chosen such that the childrens' parts of the keyspace will
be subsets of the parent's parts of the keyspace.
#### Step 1: write new remote indices
In remote storage, splitting is very simple: we may just write new index_part.json
objects for each child shard, containing exactly the same layers as the parent shard.
The children will have more data than they need, but this avoids any exhausive
re-writing or copying of layer files.
The index key path includes a generation number: the parent shard's current
attached generation number will also be used for the child shards' indices. This
makes the operation safely retryable: if everything crashes and restarts, we may
call the split API again on the parent shard, and the result will be some new remote
indices for the child shards, under a higher generation number.
#### Step 2: start new `Tenant` objects
A new `Tenant` object may be instantiated for each child shard, while the parent
shard still exists. When calling the tenant_spawn function for this object,
the remote index from step 1 will be read, and the child shard will start
to ingest WAL to catch up from whatever was in the remote storage at step 1.
We now wait for child shards' WAL ingestion to catch up with the parent shard,
so that we can safely tear down the parent shard without risking an availability
gap to clients reading recent LSNs.
#### Step 3: tear down parent `Tenant` object
Once child shards are running and have caught up with WAL ingest, we no longer
need the parent shard. Note that clients may still be using it -- when we
shut it down, any page_service handlers will also shut down, causing clients
to disconnect. When the client reconnects, it will re-lookup the tenant,
and hit the child shard instead of the parent (shard lookup from page_service
should bias toward higher ShardCount shards).
Note that at this stage the page service client has not yet been notified of
any split. In the trivial single split example:
- Shard 0001 is gone: Tenant object torn down
- Shards 0002 and 0102 are running on the same pageserver where Shard 0001 used to live.
- Clients will continue to connect to that server thinking that shard 0001 is there,
and all requests will work, because any key that was in shard 0001 is definitely
available in either shard 0002 or shard 0102.
- Eventually, the storage controller (not the pageserver) will decide to migrate
some child shards away: at that point it will do a live migration, ensuring
that the client has an updated configuration before it detaches anything
from the original server.
#### Complete
When we send a 200 response to the split request, we are promising the caller:
- That the child shards are persistent in remote storage
- That the parent shard has been shut down
This enables the caller to proceed with the overall shard split operation, which
may involve other shards on other pageservers.
### Storage Controller Split procedure
Splitting a tenant requires calling the pageserver split API, and tracking
enough state to ensure recovery + completion in the event of any component (pageserver
or storage controller) crashing (or request timing out) during the split.
1. call the split API on all existing shards. Ensure that the resulting
child shards are pinned to their pageservers until _all_ the split calls are done.
This pinning may be implemented as a "split bit" on the tenant shards, that
blocks any migrations, and also acts as a sign that if we restart, we must go
through some recovery steps to resume the split.
2. Once all the split calls are done, we may unpin the child shards (clear
the split bit). The split is now complete: subsequent steps are just migrations,
not strictly part of the split.
3. Try to schedule new pageserver locations for the child shards, using
a soft anti-affinity constraint to place shards from the same tenant onto different
pageservers.
Updating computes about the new shard count is not necessary until we migrate
any of the child shards away from the parent's location.
### Recovering from failures
#### Rolling back an incomplete split
An incomplete shard split may be rolled back quite simply, by attaching the parent shards to pageservers,
and detaching child shards. This will lose any WAL ingested into the children after the parents
were detached earlier, but the parents will catch up.
No special pageserver API is needed for this. From the storage controllers point of view, the
procedure is:
1. For all parent shards in the tenant, ensure they are attached
2. For all child shards, ensure they are not attached
3. Drop child shards from the storage controller's database, and clear the split bit on the parent shards.
Any remote storage content for child shards is left behind. This is similar to other cases where
we may leave garbage objects in S3 (e.g. when we upload a layer but crash before uploading an
index that references it). Future online scrub/cleanup functionality can remove these objects, or
they will be removed when the tenant is deleted, as tenant deletion lists all objects in the prefix,
which would include any child shards that were rolled back.
If any timelines had been created on child shards, they will be lost when rolling back. To mitigate
this, we will **block timeline creation during splitting**, so that we can safely roll back until
the split is complete, without risking losing timelines.
Rolling back an incomplete split will happen automatically if a split fails due to some fatal
reason, and will not be accessible via an API:
- A pageserver fails to complete its split API request after too many retries
- A pageserver returns a fatal unexpected error such as 400 or 500
- The storage controller database returns a non-retryable error
- Some internal invariant is violated in the storage controller split code
#### Rolling back a complete split
A complete shard split may be rolled back similarly to an incomplete split, with the following
modifications:
- The parent shards will no longer exist in the storage controller database, so these must
be re-synthesized somehow: the hard part of this is figuring the parent shards' generations. This
may be accomplished either by probing in S3, or by retaining some tombstone state for deleted
shards in the storage controller database.
- Any timelines that were created after the split complete will disappear when rolling back
to the tenant shards. For this reason, rolling back after a complete split should only
be done due to serious issues where loss of recently created timelines is acceptable, or
in cases where we have confirmed that no timelines were created in the intervening period.
- Parent shards' layers must not have been deleted: this property will come "for free" when
we first roll out sharding, by simply not implementing deletion of parent layers after
a split. When we do implement such deletion (see "Cleaning up parent-shard layers" in the
Optimizations section), it should apply a TTL to layers such that we have a
defined walltime window in which rollback will be possible.
The storage controller will expose an API for rolling back a complete split, for use
in the field if we encounter some critical bug with a post-split tenant.
#### Retrying API calls during Pageserver Restart
When a pageserver restarts during a split API call, it may witness on-disk content for both parent and
child shards from an ongoing split. This does not intrinsically break anything, and the
pageserver may include all these shards in its `/re-attach` request to the storage controller.
In order to support such restarts, it is important that the storage controller stores
persistent records of each child shard before it calls into a pageserver, as these child shards
may require generation increments via a `/re-attach` request.
The pageserver restart will also result in a failed API call from the storage controller's point
of view. Recall that if _any_ pageserver fails to split, the overall split operation may not
complete, and all shards must remain pinned to their current pageserver locations until the
split is done.
The pageserver API calls during splitting will retry on transient errors, so that
short availability gaps do not result in a failure of the overall operation. The
split in progress will be automatically rolled back if the threshold for API
retries is reached (e.g. if a pageserver stays offline for longer than a typical
restart).
#### Rollback on Storage Controller Restart
On startup, the storage controller will inspect the split bit for tenant shards that
it loads from the database. If any splits are in progress:
- Database content will be reverted to the parent shards
- Child shards will be dropped from memory
- The parent and child shards will be included in the general startup reconciliation that
the storage controller does: any child shards will be detached from pageservers because
they don't exist in the storage controller's expected set of shards, and parent shards
will be attached if they aren't already.
#### Storage controller API request failures/retries
The split request handler will implement idempotency: if the [`Tenant`] requested to split
doesn't exist, we will check for the would-be child shards, and if they already exist,
we consider the request complete.
If a request is retried while the original request is still underway, then the split
request handler will notice an InProgress marker in TenantManager, and return 503
to encourage the client to backoff/retry. This is the same as the general pageserver
API handling for calls that try to act on an InProgress shard.
#### Compute start/restart during a split
If a compute starts up during split, it will be configured with the old sharding
configuration. This will work for reads irrespective of the progress of the split
as long as no child hards have been migrated away from their original location, and
this is guaranteed in the split procedure (see earlier section).
#### Pageserver fails permanently during a split
If a pageserver permanently fails (i.e. the storage controller availability state for it
goes to Offline) while a split is in progress, the splitting operation will roll back, and
during the roll back it will skip any API calls to the offline pageserver. If the offline
pageserver becomes available again, any stale locations will be cleaned up via the normal reconciliation process (the `/re-attach` API).
### Handling secondary locations
For correctness, it is not necessary to split secondary locations. We can simply detach
the secondary locations for parent shards, and then attach new secondary locations
for child shards.
Clearly this is not optimal, as it will result in re-downloads of layer files that
were already present on disk. See "Splitting secondary locations"
### Conditions to trigger a split
The pageserver will expose a new API for reporting on shards that are candidates
for split: this will return a top-N report of the largest tenant shards by
physical size (remote size). This should exclude any tenants that are already
The storage controller will poll that API across all pageservers it manages at some appropriate interval (e.g. 60 seconds).
A split operation will be started when the tenant exceeds some threshold. This threshold
should be _less than_ how large we actually want shards to be, perhaps much less. That's to
minimize the amount of work involved in splitting -- if we want 100GiB shards, we shouldn't
wait for a tenant to exceed 100GiB before we split anything. Some data analysis of existing
tenant size distribution may be useful here: if we can make a statement like "usually, if
a tenant has exceeded 20GiB they're probably going to exceed 100GiB later", then we might
make our policy to split a tenant at 20GiB.
The finest split we can do is by factors of two, but we can do higher-cardinality splits
too, and this will help to reduce the overhead of repeatedly re-splitting a tenant
as it grows. An example of a very simple heuristic for early deployment of the splitting
feature would be: "Split tenants into 8 shards when their physical size exceeds 64GiB": that
would give us two kinds of tenant (1 shard and 8 shards), and the confidence that once we had
split a tenant, it will not need re-splitting soon after.
## Optimizations
### Flush parent shard to remote storage during split
Any data that is in WAL but not remote storage at time of split will need
to be replayed by child shards when they start for the first time. To minimize
this work, we may flush the parent shard to remote storage before writing the
remote indices for child shards.
It is important that this flush is subject to some time bounds: we may be splitting
in response to a surge of write ingest, so it may be time-critical to split. A
few seconds to flush latest data should be sufficient to optimize common cases without
running the risk of holding up a split for a harmful length of time when a parent
shard is being written heavily. If the flush doesn't complete in time, we may proceed
to shut down the parent shard and carry on with the split.
### Hard linking parent layers into child shard directories
Before we start the Tenant objects for child shards, we may pre-populate their
local storage directories with hard links to the layer files already present
in the parent shard's local directory. When the child shard starts and downloads
its remote index, it will find all those layer files already present on local disk.
This avoids wasting download capacity and makes splitting faster, but more importantly
it avoids taking up a factor of N more disk space when splitting 1 shard into N.
This mechanism will work well in typical flows where shards are migrated away
promptly after a split, but for the general case including what happens when
layers are evicted and re-downloaded after a split, see the 'Proactive compaction'
section below.
### Filtering during compaction
Compaction, especially image layer generation, should skip any keys that are
present in a shard's layer files, but do not match the shard's ShardIdentity's
is_key_local() check. This avoids carrying around data for longer than necessary
in post-split compactions.
This was already implemented in https://github.com/neondatabase/neon/pull/6246
### Proactive compaction
In remote storage, there is little reason to rewrite any data on a shard split:
all the children can reference parent layers via the very cheap write of the child
index_part.json.
In local storage, things are more nuanced. During the initial split there is no
capacity cost to duplicating parent layers, if we implement the hard linking
optimization described above. However, as soon as any layers are evicted from
local disk and re-downloaded, the downloaded layers will not be hard-links any more:
they'll have real capacity footprint. That isn't a problem if we migrate child shards
away from the parent node swiftly, but it risks a significant over-use of local disk
space if we do not.
For example, if we did an 8-way split of a shard, and then _didn't_ migrate 7 of
the shards elsewhere, then churned all the layers in all the shards via eviction,
then we would blow up the storage capacity used on the node by 8x. If we're splitting
a 100GB shard, that could take the pageserver to the point of exhausting disk space.
To avoid this scenario, we could implement a special compaction mode where we just
read historic layers, drop unwanted keys, and write back the layer file. This
is pretty expensive, but useful if we have split a large shard and are not going to
migrate the child shards away.
The heuristic conditions for triggering such a compaction are:
- A) eviction plus time: if a child shard
has existed for more than a time threshold, and has been requested to perform at least one eviction, then it becomes urgent for this child shard to execute a proactive compaction to reduce its storage footprint, at the cost of I/O load.
- B) resident size plus time: we may inspect the resident layers and calculate how
many of them include the overhead of storing pre-split keys. After some time
threshold (different to the one in case A) we still have such layers occupying
local disk space, then we should proactively compact them.
### Cleaning up parent-shard layers
It is functionally harmless to leave parent shard layers in remote storage indefinitely.
They would be cleaned up in the event of the tenant's deletion.
As an optimization to avoid leaking remote storage capacity (which costs money), we may
lazily clean up parent shard layers once no child shards reference them.
This may be done _very_ lazily: e.g. check every PITR interval. The cleanup procedure is:
- list all the key prefixes beginning with the tenant ID, and select those shard prefixes
which do not belong to the most-recently-split set of shards (_ancestral shards_, i.e. `shard*count < max(shard_count) over all shards)`, and those shard prefixes which do have the latest shard count (_current shards_)
- If there are no _ancestral shard_ prefixes found, we have nothing to clean up and
may drop out now.
- find the latest-generation index for each _current shard_, read all and accumulate the set of layers belonging to ancestral shards referenced by these indices.
- for all ancestral shards, list objects in the prefix and delete any layer which was not
referenced by a current shard.
If this cleanup is scheduled for 1-2 PITR periods after the split, there is a good chance that child shards will have written their own image layers covering the whole keyspace, such that all parent shard layers will be deletable.
The cleanup may be done by the scrubber (external process), or we may choose to have
the zeroth shard in the latest generation do the work -- there is no obstacle to one shard
reading the other shard's indices at runtime, and we do not require visibility of the
latest index writes.
Cleanup should be artificially delayed by some period (for example 24 hours) to ensure
that we retain the option to roll back a split in case of bugs.
### Splitting secondary locations
We may implement a pageserver API similar to the main splitting API, which does a simpler
operation for secondary locations: it would not write anything to S3, instead it would simply
create the child shard directory on local disk, hard link in directories from the parent,
and set up the in memory (TenantSlot) state for the children.
Similar to attached locations, a subset of secondary locations will probably need re-locating
after the split is complete, to avoid leaving multiple child shards on the same pageservers,
where they may use excessive space for the tenant.
## FAQ/Alternatives
### What should the thresholds be set to?
Shard size limit: the pre-sharding default capacity quota for databases was 200GiB, so this could be a starting point for the per-shard size limit.
Max shard count:
- The safekeeper overhead to sharding is currently O(N) network bandwidth because
the un-filtered WAL is sent to all shards. To avoid this growing out of control,
a limit of 8 shards should be temporarily imposed until WAL filtering is implemented
on the safekeeper.
- there is also little benefit to increasing the shard count beyond the number
of pageservers in a region.
### Is it worth just rewriting all the data during a split to simplify reasoning about space?
The storage controller sits between administrative API clients and pageservers, and handles the details of mapping tenants to pageserver tenant shards. For example, creating a tenant is one API call to the storage controller,
which is mapped into many API calls to many pageservers (for multiple shards, and for secondary locations).
It implements a pageserver-compatible API that may be used for CRUD operations on tenants and timelines, translating these requests into appropriate operations on the shards within a tenant, which may be on many different pageservers. Using this API, the storage controller may be used in the same way as the pageserver's administrative HTTP API, hiding
the underlying details of how data is spread across multiple nodes.
The storage controller also manages generations, high availability (via secondary locations) and live migrations for tenants under its management. This is done with a reconciliation loop pattern, where tenants have an “intent” state and a “reconcile” task that tries to make the outside world match the intent.
## APIs
The storage controller’s HTTP server implements four logically separate APIs:
-`/v1/...` path is the pageserver-compatible API. This has to be at the path root because that’s where clients expect to find it on a pageserver.
-`/control/v1/...` path is the storage controller’s API, which enables operations such as registering and management pageservers, or executing shard splits.
-`/debug/v1/...` path contains endpoints which are either exclusively used in tests, or are for use by engineers when supporting a deployed system.
-`/upcall/v1/...` path contains endpoints that are called by pageservers. This includes the `/re-attach` and `/validate` APIs used by pageservers
to ensure data safety with generation numbers.
The API is authenticated with a JWT token, and tokens must have scope `pageserverapi` (i.e. the same scope as pageservers’ APIs).
See the `http.rs` file in the source for where the HTTP APIs are implemented.
## Database
The storage controller uses a postgres database to persist a subset of its state. Note that the storage controller does _not_ keep all its state in the database: this is a design choice to enable most operations to be done efficiently in memory, rather than having to read from the database. See `persistence.rs` for a more comprehensive comment explaining what we do and do not persist: a useful metaphor is that we persist objects like tenants and nodes, but we do not
persist the _relationships_ between them: the attachment state of a tenant's shards to nodes is kept in memory and
rebuilt on startup.
The file `persistence.rs` contains all the code for accessing the database, and has a large doc comment that goes into more detail about exactly what we persist and why.
The `diesel` crate is used for defining models & migrations.
Running a local cluster with `cargo neon` automatically starts a vanilla postgress process to host the storage controller’s database.
### Diesel tip: migrations
If you need to modify the database schema, here’s how to create a migration:
- Install the diesel CLI with `cargo install diesel_cli`
- Use `diesel migration generate <name>` to create a new migration
- Populate the SQL files in the `migrations/` subdirectory
- Use `DATABASE_URL=... diesel migration run` to apply the migration you just wrote: this will update the `[schema.rs](http://schema.rs)` file automatically.
- This requires a running database: the easiest way to do that is to just run `cargo neon init ; cargo neon start`, which will leave a database available at `postgresql://localhost:1235/attachment_service`
- Commit the migration files and the changes to schema.rs
- If you need to iterate, you can rewind migrations with `diesel migration revert -a` and then `diesel migration run` again.
- The migrations are build into the storage controller binary, and automatically run at startup after it is deployed, so once you’ve committed a migration no further steps are needed.
## storcon_cli
The `storcon_cli` tool enables interactive management of the storage controller. This is usually
only necessary for debug, but may also be used to manage nodes (e.g. marking a node as offline).
`storcon_cli --help` includes details on commands.
# Deploying
This section is aimed at engineers deploying the storage controller outside of Neon's cloud platform, as
part of a self-hosted system.
_General note: since the default `neon_local` environment includes a storage controller, this is a useful
reference when figuring out deployment._
## Database
It is **essential** that the database used by the storage controller is durable (**do not store it on ephemeral
local disk**). This database contains pageserver generation numbers, which are essential to data safety on the pageserver.
The resource requirements for the database are very low: a single CPU core and 1GiB of memory should work well for most deployments. The physical size of the database is typically under a gigabyte.
Set the URL to the database using the `--database-url` CLI option.
There is no need to run migrations manually: the storage controller automatically applies migrations
when it starts up.
## Configure pageservers to use the storage controller
1. The pageserver `control_plane_api` and `control_plane_api_token` should be set in the `pageserver.toml` file. The API setting should
point to the "upcall" prefix, for example `http://127.0.0.1:1234/upcall/v1/` is used in neon_local clusters.
2. Create a `metadata.json` file in the same directory as `pageserver.toml`: this enables the pageserver to automatically register itself
with the storage controller when it starts up. See the example below for the format of this file.
This is a library for running simulations of distributed systems. The main idea is borrowed from [FoundationDB](https://www.youtube.com/watch?v=4fFDFbi3toc).
Each node runs as a separate thread. This library was not optimized for speed yet, but it's already much faster than running usual intergration tests in real time, because it uses virtual simulation time and can fast-forward time to skip intervals where all nodes are doing nothing but sleeping or waiting for something.
The original purpose for this library is to test walproposer and safekeeper implementation working together, in a scenarios close to the real world environment. This simulator is determenistic and can inject failures in networking without waiting minutes of wall-time to trigger timeout, which makes it easier to find bugs in our consensus implementation compared to using integration tests.
/// Describes random delays and failures. Delay will be uniformly distributed in [min, max].
/// Connection failure will occur with the probablity fail_prob.
#[derive(Clone, Debug)]
pubstructDelay{
pubmin: u64,
pubmax: u64,
pubfail_prob: f64,// [0; 1]
}
implDelay{
/// Create a struct with no delay, no failures.
pubfnempty()-> Delay{
Delay{
min: 0,
max: 0,
fail_prob: 0.0,
}
}
/// Create a struct with a fixed delay.
pubfnfixed(ms: u64)-> Delay{
Delay{
min: ms,
max: ms,
fail_prob: 0.0,
}
}
/// Generate a random delay in range [min, max]. Return None if the
/// message should be dropped.
pubfndelay(&self,rng: &mutStdRng)-> Option<u64>{
ifrng.gen_bool(self.fail_prob){
returnNone;
}
Some(rng.gen_range(self.min..=self.max))
}
}
/// Describes network settings. All network packets will be subjected to the same delays and failures.
#[derive(Clone, Debug)]
pubstructNetworkOptions{
/// Connection will be automatically closed after this timeout if no data is received.
pubkeepalive_timeout: Option<u64>,
/// New connections will be delayed by this amount of time.
pubconnect_delay: Delay,
/// Each message will be delayed by this amount of time.
pubsend_delay: Delay,
}
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
