## Problem
The endpoint filter cache is still unused because it's not yet reliable
enough to be used. It only consumes a lot of memory.
## Summary of changes
Remove the code. Needs a new design.
neondatabase/cloud#30634
See #11992 and #11961 for some examples of usecases.
This introduces a JSON serialization lib, designed for more flexibility
than serde_json offers.
## Dynamic construction
Sometimes you have dynamic values you want to serialize, that are not
already in a serde-aware model like a struct or a Vec etc. To achieve
this with serde, you need to implement a lot of different traits on a
lot of different new-types. Because of this, it's often easier to
give-in and pull all the data into a serde-aware model
(serde_json::Value or some intermediate struct), but that is often not
very efficient.
This crate allows full control over the JSON encoding without needing to
implement any extra traits. Just call the relevant functions, and it
will guarantee a correctly encoded JSON value.
## Async construction
Similar to the above, sometimes the values arrive asynchronously. Often
collecting those values in memory is more expensive than writing them as
JSON, since the overheads of `Vec` and `String` is much higher, however
there are exceptions.
Serializing to JSON all in one go is also more CPU intensive and can
cause lag spikes, whereas serializing values incrementally spreads out
the CPU load and reduces lag.
## Problem
We don't log gRPC request errors on the server.
Touches #11728.
## Summary of changes
Automatically log non-OK gRPC response statuses in the observability
middleware, and add corresponding logging for the `get_pages` stream.
Also adds the peer address and gRPC method to the gRPC tracing span.
Example output:
```
2025-07-02T20:18:16.813718Z WARN grpc:pageservice{peer=127.0.0.1:56698 method=CheckRelExists tenant_id=c7b45faa1924b1958f05c5fdee8b0d04 timeline_id=4a36ee64fd2f97781b9dcc2c3cddd51b shard_id=0000}: request failed with NotFound: Tenant c7b45faa1924b1958f05c5fdee8b0d04 not found
```
## Problem
Names are not consistent between safekeeper migration RFC and the actual
implementation.
It's not used anywhere in production yet, so it's safe to rename. We
don't need to worry about backward compatibility.
- Follow up on https://github.com/neondatabase/neon/pull/12432
## Summary of changes
- rename term -> last_log_term in TimelineMembershipSwitchResponse
- add missing fields to TimelineMembershipSwitchResponse in python
## Problem
The safekeeper migration code/logic slightly diverges from the initial
RFC. This PR aims to address these differences.
- Part of https://github.com/neondatabase/neon/issues/12192
## Summary of changes
- Adjust the RFC to reflect that we implemented the safekeeper
reconciler with in-memory queue.
- Add `sk_set_notified_generation` field to the `timelines` table in the
RFC to address the "finish migration atomically" problem.
- Describe how we are going to make the timeline migration handler fully
retriable with in-memory reconciler queue.
- Unify type/field/method names in the code and RFC.
- Fix typos
This doesn't do anything interesting yet, but demonstrates linking Rust
code to the neon Postgres extension, so that we can review and test
drive just the build process changes independently.
Mostly a revert of https://github.com/neondatabase/neon/pull/11851 and
https://github.com/neondatabase/neon/pull/12330 .
Christian suggested reverting his PR to fix the issue
https://github.com/neondatabase/neon/issues/12369 .
Alternatives considered:
1. I have originally wanted to introduce cancellation tokens to
`RequestContext`, but in the end I gave up on them because I didn't find
a select-free way of preventing
`test_layer_download_cancelled_by_config_location` from hanging.
Namely if I put a select around the `get_or_maybe_download` invocation
in `get_values_reconstruct_data`, it wouldn't hang, but if I put it
around the `download_init_and_wait` invocation in
`get_or_maybe_download`, the test would still hang. Not sure why, even
though I made the attached child function of the `RequestContext` create
a child token.
2. Introduction of a `download_cancel` cancellation token as a child of
a timeline token, putting it into `RemoteTimelineClient` together with
the main token, and then putting it into the whole
`RemoteTimelineClient` read path.
3. Greater refactorings, like to make cancellation tokens follow a DAG
structure so you can have tokens cancelled either by say timeline
shutting down or a request ending. It doesn't just represent an effort
that we don't have the engineering budget for, it also causes
interesting questions like what to do about batching (do you cancel the
entire request if only some requests get cancelled?).
We might see a reemergence of
https://github.com/neondatabase/neon/issues/11762, but given that we
have https://github.com/neondatabase/neon/pull/11853 and
https://github.com/neondatabase/neon/pull/12376 now, it is possible that
it will not come back. Looking at some code, it might actually fix the
locations where the error pops up. Let's see.
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
- Add ComputeSpec flag `offload_lfc_interval_seconds` controlling
whether LFC should be offloaded to endpoint storage. Default value
(None) means "don't offload".
- Add glue code around it for `neon_local` and integration tests.
- Add `autoprewarm` mode for `test_lfc_prewarm` testing
`offload_lfc_interval_seconds` and `autoprewarm` flags in conjunction.
- Rename `compute_ctl_lfc_prewarm_requests_total` and
`compute_ctl_lfc_offload_requests_total` to
`compute_ctl_lfc_prewarms_total`
and `compute_ctl_lfc_offloads_total` to reflect we count prewarms and
offloads, not `compute_ctl` requests of those.
Don't count request in metrics if there is a prewarm/offload already
ongoing.
https://github.com/neondatabase/cloud/issues/19011
Resolves: https://github.com/neondatabase/cloud/issues/30770
Don't build walproposer-lib as a separate job. It only takes a few
seconds, after you have built all its dependencies.
Don't cache the Neon Pg extensions in the per-postgres-version caches.
This is in preparation for the communicator project, which will
introduce Rust parts to the Neon Pg extension, which complicates the
build process. With that, the 'make neon-pg-ext' step requires some of
the Rust bits to be built already, or it will build them on the spot,
which in turn requires all the Rust sources to be present, and we don't
want to repeat that part for each Postgres version anyway. To prepare
for that, rely on "make all" to build the neon extension and the rust
bits in the correct order instead. Building the neon extension doesn't
currently take very long anyway after you have built Postgres itself, so
you don't gain much by caching it. See
https://github.com/neondatabase/neon/pull/12266.
Add an explicit "rustup update" step to update the toolchain. It's not
strictly necessary right now, because currently "make all" will only
invoke "cargo build" once and the race condition described in the
comment doesn't happen. But prepare for the future.
To further simplify the build, get rid of the separate 'build-postgres'
jobs too, and just build Postgres as a step in the main job. That makes
the overall workflow run longer, because we no longer build all the
postgres versions in parallel (although you still get intra-runner
parallelism thanks to `make -j`), but that's acceptable. In the
cache-hit case, it might even be a little faster because there is less
overhead from launching jobs, and in the cache-miss case, it's maybe
5-10 minutes slower altogether.
---------
Co-authored-by: Alexander Bayandin <alexander@neon.tech>
## Problem
The current deletion operation is synchronous and blocking, which is
unsuitable for potentially long-running tasks like. In such cases, the
standard HTTP request-response pattern is not a good fit.
## Summary of Changes
- Added new `storcon_cli` commands: `NodeStartDelete` and
`NodeCancelDelete` to initiate and cancel deletion asynchronously.
- Added corresponding `storcon` HTTP handlers to support the new
start/cancel deletion flow.
- Introduced a new type of background operation: `Delete`, to track and
manage the deletion process outside the request lifecycle.
---------
Co-authored-by: Aleksandr Sarantsev <aleksandr.sarantsev@databricks.com>
## Problem
I suspect that the pageservers get stuck on receiving broker updates.
## Summary of changes
This is a an opportunistic (staging only) patch that resets the
susbscription
stream if it's been idle for a while. This won't go to prod in this
form.
I'll revert or update it before Friday.
## Problem
DEBUG_LOCAL_COMPARE mode allows to detect data corruption.
But it requires rebuild of neon extension (and so requires special
image) and significantly slowdown execution because always fetch pages
from page server.
## Summary of changes
Introduce new GUC `neon.debug_compare_local`, accepting the following
values: " none", "prefetch", "lfc", "all" (by default it is definitely
disabled).
In mode less than "all", neon SMGR will not fetch page from PS if it is
found in local caches.
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
The background installed extensions worker relied on `interval.tick()`
to go to sleep for a period of time. This can lead to bugs due to the
interval being updated at the end of the loop as the first tick is
[instantaneous](https://docs.rs/tokio/latest/tokio/time/struct.Interval.html#method.tick).
## Summary of changes
Changed it to a `tokio::time::sleep` to prevent this issue. Now it puts
the thread to sleep and only wakes up after the specified duration
## Problem
Better investigate memory usage during backpressure
## Summary of changes
Print open layer size if backpressure is activated
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
Ancestor detach from a previously detached parent when there were no
writes panics since it tries to upload the tombstone layer twice.
## Summary of Changes
If we're gonna copy the tombstone from the ancestor, don't bother
creating it.
Fixes https://github.com/neondatabase/neon/issues/12458
When deploying new safekeepers, we don't immediately want to send
traffic to them. Maybe they are not ready yet by the time the deploy
script is registering them with the storage controller.
For pageservers, the storcon solves the problem by not scheduling stuff
to them unless there has been a positive heartbeat response. We can't do
the same for safekeepers though, otherwise a single down safekeeper
would mean we can't create new timelines in smaller regions where there
is only three safekeepers in total.
So far we have created safekeepers as `pause` but this adds a manual
step to safekeeper deployment which is prone to oversight. We want
things to be automatted. So we introduce a new state `activating` that
acts just like `pause`, except that we automatically transition the
policy to `active` once we get a positive heartbeat from the safekeeper.
For `pause`, we always keep the safekeeper paused.
## Problem
Currently, if `storcon` (storage controller) reconciliations repeatedly
fail, the system will indefinitely freeze optimizations. This can result
in optimization starvation for several days until the reconciliation
issues are manually resolved. To mitigate this, we should detect
persistently failing reconciliations and exclude them from influencing
the optimization decision.
## Summary of Changes
- A tenant shard reconciliation is now considered "keep-failing" if it
fails 5 consecutive times. These failures are excluded from the
optimization readiness check.
- Added a new metric: `storage_controller_keep_failing_reconciles` to
monitor such cases.
- Added a warning log message when a reconciliation is marked as
"keep-failing".
---------
Co-authored-by: Aleksandr Sarantsev <aleksandr.sarantsev@databricks.com>
## TLDR
This PR is a no-op. The changes are disabled by default.
## Problem
I. Currently we don't have a way to detect disk I/O failures from WAL
operations.
II.
We observe that the offloader fails to upload a segment due to race
conditions on XLOG SWITCH and PG start streaming WALs. wal_backup task
continously failing to upload a full segment while the segment remains
partial on the disk.
The consequence is that commit_lsn for all SKs move forward but
backup_lsn stays the same. Then, all SKs run out of disk space.
III.
We have discovered SK bugs where the WAL offload owner cannot keep up
with WAL backup/upload to S3, which results in an unbounded accumulation
of WAL segment files on the Safekeeper's disk until the disk becomes
full. This is a somewhat dangerous operation that is hard to recover
from because the Safekeeper cannot write its control files when it is
out of disk space. There are actually 2 problems here:
1. A single problematic timeline can take over the entire disk for the
SK
2. Once out of disk, it's difficult to recover SK
IV.
Neon reports certain storage errors as "critical" errors using a marco,
which will increment a counter/metric that can be used to raise alerts.
However, this metric isn't sliced by tenant and/or timeline today. We
need the tenant/timeline dimension to better respond to incidents and
for blast radius analysis.
## Summary of changes
I.
The PR adds a `safekeeper_wal_disk_io_errors ` which is incremented when
SK fails to create or flush WALs.
II.
To mitigate this issue, we will re-elect a new offloader if the current
offloader is lagging behind too much.
Each SK makes the decision locally but they are aware of each other's
commit and backup lsns.
The new algorithm is
- determine_offloader will pick a SK. say SK-1.
- Each SK checks
-- if commit_lsn - back_lsn > threshold,
-- -- remove SK-1 from the candidate and call determine_offloader again.
SK-1 will step down and all SKs will elect the same leader again.
After the backup is caught up, the leader will become SK-1 again.
This also helps when SK-1 is slow to backup.
I'll set the reelect backup lag to 4 GB later. Setting to 128 MB in dev
to trigger the code more frequently.
III.
This change addresses problem no. 1 by having the Safekeeper perform a
timeline disk utilization check check when processing WAL proposal
messages from Postgres/compute. The Safekeeper now rejects the WAL
proposal message, effectively stops writing more WAL for the timeline to
disk, if the existing WAL files for the timeline on the SK disk exceeds
a certain size (the default threshold is 100GB). The disk utilization is
calculated based on a `last_removed_segno` variable tracked by the
background task removing WAL files, which produces an accurate and
conservative estimate (>= than actual disk usage) of the actual disk
usage.
IV.
* Add a new metric `hadron_critical_storage_event_count` that has the
`tenant_shard_id` and `timeline_id` as dimensions.
* Modified the `crtitical!` marco to include tenant_id and timeline_id
as additional arguments and adapted existing call sites to populate the
tenant shard and timeline ID fields. The `critical!` marco invocation
now increments the `hadron_critical_storage_event_count` with the extra
dimensions. (In SK there isn't the notion of a tenant-shard, so just the
tenant ID is recorded in lieu of tenant shard ID.)
I considered adding a separate marco to avoid merge conflicts, but I
think in this case (detecting critical errors) conflicts are probably
more desirable so that we can be aware whenever Neon adds another
`critical!` invocation in their code.
---------
Co-authored-by: Chen Luo <chen.luo@databricks.com>
Co-authored-by: Haoyu Huang <haoyu.huang@databricks.com>
Co-authored-by: William Huang <william.huang@databricks.com>
## Problem
integrating subzero requires a bit of refactoring. To make the
integration PR a bit more manageable, the refactoring is done in this
separate PR.
## Summary of changes
* move common types/functions used in sql_over_http to errors.rs and
http_util.rs
* add the "Local" auth backend to proxy (similar to local_proxy), useful
in local testing
* change the Connect and Send type for the http client to allow for
custom body when making post requests to local_proxy from the proxy
---------
Co-authored-by: Ruslan Talpa <ruslan.talpa@databricks.com>
## Problem
`keep_connection` does not exit, so it was never setting
`credentials_refreshed`.
## Summary of changes
Set `credentials_refreshed` to true when we first establish a
connection, and after we re-authenticate the connection.
## Problem
`keep_connection` does not exit, so it was never setting
`credentials_refreshed`.
## Summary of changes
Set `credentials_refreshed` to true when we first establish a
connection, and after we re-authenticate the connection.
## Problem
#12450
## Summary of changes
Instead of `#[arg(default_value_t = typed_default_value)]`, we use
`#[arg(default_value = "str that deserializes into the value")]`,
because apparently you can't convince clap to _not_ deserialize from the
Display implementation of an imported enum.
Health checks for pg-sni-router open a TCP connection and immediately
close it again. This is noisy. We will filter out any EOF errors on the
first message.
"acquired permit" debug log is incorrect since it logs when we timedout
as well. This fixes the debug log.
## Problem
When local proxy is configured with TLS, the certificate does not match
the endpoint string. This currently returns an error.
## Summary of changes
I don't think this code is necessary anymore, taking the prefix from the
hostname is good enough (and is equivalent to what `endpoint_sni` was
doing) and we ignore checking the domain suffix.
## Problem
Neon currently implements several features that guarantee high uptime of
compute nodes:
1. Storage high-availability (HA), i.e. each tenant shard has a
secondary pageserver location, so we can quickly switch over compute to
it in case of primary pageserver failure.
2. Fast compute provisioning, i.e. we have a fleet of pre-created empty
computes, that are ready to serve workload, so restarting unresponsive
compute is very fast.
3. Preemptive NeonVM compute provisioning in case of k8s node
unavailability.
This helps us to be well-within the uptime SLO of 99.95% most of the
time. Problems begin when we go up to multi-TB workloads and 32-64 CU
computes. During restart, compute looses all caches: LFC, shared
buffers, file system cache. Depending on the workload, it can take a lot
of time to warm up the caches, so that performance could be degraded and
might be even unacceptable for certain workloads. The latter means that
although current approach works well for small to
medium workloads, we still have to do some additional work to avoid
performance degradation after restart of large instances.
[Rendered
version](https://github.com/neondatabase/neon/blob/alexk/pg-prewarm-rfc/docs/rfcs/2025-03-17-compute-prewarm.md)
Part of https://github.com/neondatabase/cloud/issues/19011
## Problem
The billing team wants to change the billing events pipeline and use a
common events format in S3 buckets across different event producers.
## Summary of changes
Change the events storage format for billing events from JSON to NDJSON.
Also partition files by hours, rather than days.
Resolves: https://github.com/neondatabase/cloud/issues/29995
We would like to move towards an arena system for JSON encoding the
responses. This change pushes an "out" parameter into the pg-test to
json routines to make swapping in an arena system easier in the future.
(see #11992)
This additionally removes the redundant `column: &[Type]` argument, as
well as rewriting the pg_array parser.
---
I rewrote the pg_array parser since while making these changes I found
it hard to reason about. I went back to the specification and rewrote it
from scratch. There's 4 separate routines:
1. pg_array_parse - checks for any prelude (multidimensional array
ranges)
2. pg_array_parse_inner - only deals with the arrays themselves
3. pg_array_parse_item - parses a single item from the array, this might
be quoted, unquoted, or another nested array.
4. pg_array_parse_quoted - parses a quoted string, following the
relevant string escaping rules.
Don't print errors like:
```
Compaction failed 1 times, retrying in 2s: Failed to offload timeline: Unexpected offload error: Timeline deletion is already in progress
```
Print it at info log level instead.
https://github.com/neondatabase/cloud/issues/30666
## Problem
https://github.com/neondatabase/neon/pull/11712 changed how computes are
started in the test: the lsn is specified, making them read-only static
replicas. Lsn is `last_record_lsn` from pageserver. It works fine with
read-only branches (because their `last_record_lsn` is equal to
`start_lsn` and always valid). But with writable timelines, the
`last_record_lsn` on the pageserver might be stale.
Particularly in this test, after the `detach_branch` operation, the
tenant is reset on the pagesever. It leads to `last_record_lsn` going
back to `disk_consistent_lsn`, so basically rolling back some recent
writes.
If we start a primary compute, it will start at safekeepers' commit Lsn,
which is the correct one , and will wait till pageserver catches up with
this Lsn after reset.
- Closes: https://github.com/neondatabase/neon/issues/12365
## Summary of changes
- Start `primary` compute for writable timelines.
## Problem
Part of #11813
## Summary of changes
* Compute tenant remote size in the housekeeping loop.
* Add a new `TenantFeatureResolver` struct to cache the tenant-specific
properties.
* Evaluate feature flag based on the remote size.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
close https://github.com/neondatabase/neon/issues/11528
## Summary of changes
Gives us better observability of compaction progress.
- Image creation: num of partition processed / total partition
- Gc-compaction: index of the in the queue / total items for a full
compaction
- Shard ancestor compaction: layers to rewrite / total layers
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
L0 compaction currently holds the read lock for a long region while it
doesn't need to.
## Summary of changes
This patch reduces the one long contention region into 2 short ones:
gather the layers to compact at the beginning, and several short read
locks when querying the image coverage.
Co-Authored-By: Chen Luo
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
This PR works on adding dynamic masking support for `anon` v2. It
currently only supports static masking.
## Summary of changes
Added a security definer function that sets the dynamic masking guc to
`true` with superuser permissions.
Added a security definer function that adds `anon` to
`session_preload_libraries` if it's not already present.
Related to: https://github.com/neondatabase/cloud/issues/20456
## Problem
Currently we do not have metrics for autovacuum.
## Summary of changes
Added a metric that extracts the top 5 DBs with oldest mxid and frozen
xid. Tables that were vacuumed recently should have younger value (or
younger age).
Related Issue: https://github.com/neondatabase/cloud/issues/27296
## Problem
#8843
User initiated sql queries are being classified as "postgres" errors,
whereas they're really user errors.
## Summary of changes
Classify user-initiated postgres errors as user errors if they are
related to a sql query that we ran on their behalf. Do not log those
errors.
Add support to transport syslogs over TLS. Since TLS params essentially
require passing host and port separately, add a boolean flag to the
configuration template and also use the same `action` format for
plaintext logs. This allows seamless transition.
The plaintext host:port is picked from `AUDIT_LOGGING_ENDPOINT` (as
earlier) and from `AUDIT_LOGGING_TLS_ENDPOINT`. The TLS host:port is
used when defined and non-empty.
`remote_endpoint` is split separately to hostname and port as required
by `omfwd` module.
Also the address parsing and config content generation are split to more
testable functions with basic tests added.
## Problem
We persist safekeeper host/port in the storcon DB after
https://github.com/neondatabase/neon/pull/11712, so the storcon fails to
ping safekeepers in the compatibility tests, where we start the cluster
from the snapshot.
PR also adds some small code improvements related to the test failure.
- Closes: https://github.com/neondatabase/neon/issues/12339
## Summary of changes
- Update safekeeper ports in the storcon DB when starting the neon from
the dir (snapshot)
- Fail the response on all not-success codes (e.g. 3xx). Should not
happen, but just to be more safe.
- Add `neon_previous/` to .gitignore to make it easier to run compat
tests.
- Add missing EXPORT to the instruction for running compat tests
## Problem
Previously, the background worker that collects the list of installed
extensions across DBs had a timeout set to 1 hour. This cause a problem
with computes that had a `suspend_timeout` > 1 hour as this collection
was treated as activity, preventing compute shutdown.
Issue: https://github.com/neondatabase/cloud/issues/30147
## Summary of changes
Passing the `suspend_timeout` as part of the `ComputeSpec` so that any
updates to this are taken into account by the background worker and
updates its collection interval.
If a hardlink operation inside `detach_ancestor` fails due to the layer
already existing, we delete the layer to make sure the source is one we
know about, and then retry.
But we deleted the wrong file, namely, the one we wanted to use as the
source of the hardlink. As a result, the follow up hard link operation
failed. Our PR corrects this mistake.
## Problem
We don't notify cplane about safekeeper membership change yet. Without
the notification the compute needs to know all the safekeepers on the
cluster to be able to speak to them. Change notifications will allow to
avoid it.
- Closes: https://github.com/neondatabase/neon/issues/12188
## Summary of changes
- Implement `notify_safekeepers` method in `ComputeHook`
- Notify cplane about safekeepers in `safekeeper_migrate` handler.
- Update the test to make sure notifications work.
## Out of scope
- There is `cplane_notified_generation` field in `timelines` table in
strocon's database. It's not needed now, so it's not updated in the PR.
Probably we can remove it.
- e2e tests to make sure it works with a production cplane
## Problem
The gRPC API does not provide LSN leases.
## Summary of changes
* Add LSN lease support to the gRPC API.
* Use gRPC LSN leases for static computes with `grpc://` connstrings.
* Move `PageserverProtocol` into the `compute_api::spec` module and
reuse it.
## Problem
Location config changes can currently result in changes to the shard
identity. Such changes will cause data corruption, as seen with #12217.
Resolves#12227.
Requires #12377.
## Summary of changes
Assert that the shard identity does not change on location config
updates and on (re)attach.
This is currently asserted with `critical!`, in case it misfires in
production. Later, we should reject such requests with an error and turn
this into a proper assertion.
Implicit mapping to an `anyhow::Error` when we do `?` is discouraged
because tooling to find those places isn't great.
As a drive-by, also make SplitImageLayerWriter::new infallible and sync.
I think we should also make ImageLayerWriter::new completely lazy,
then `BatchLayerWriter:new` infallible and async.
Reverts neondatabase/neon#11663 and
https://github.com/neondatabase/neon/pull/11265/
Step Security is not yet approved by Databricks team, in order to
prevent issues during Github org migration, I'll revert this PR to use
the previous action instead of Step Security maintained action.
## Problem
Similarly to #12217, the following endpoints may result in a stripe size
mismatch between the storage controller and Pageserver if an unsharded
tenant has a different stripe size set than the default. This can lead
to data corruption if the tenant is later manually split without
specifying an explicit stripe size, since the storage controller and
Pageserver will apply different defaults. This commonly happens with
tenants that were created before the default stripe size was changed
from 32k to 2k.
* `PUT /v1/tenant/config`
* `PATCH /v1/tenant/config`
These endpoints are no longer in regular production use (they were used
when cplane still managed Pageserver directly), but can still be called
manually or by tests.
## Summary of changes
Retain the current shard parameters when updating the location config in
`PUT | PATCH /v1/tenant/config`.
Also opportunistically derive `Copy` for `ShardParameters`.
## Problem
Some of the design decisions in PR #12256 were influenced by the
requirements of consistency tests. These decisions introduced
intermediate logic that is no longer needed and should be cleaned up.
## Summary of Changes
- Remove the `feature("testing")` flag related to
`kick_secondary_download`.
- Set the default value of `kick_secondary_download` back to false,
reflecting the intended production behavior.
Co-authored-by: Aleksandr Sarantsev <aleksandr.sarantsev@databricks.com>
## Problem
- Closes: https://github.com/neondatabase/neon/issues/12298
## Summary of changes
- Set `timeline_safekeeper_count` in `neon_local` if we have less than 3
safekeepers
- Remove `cfg!(feature = "testing")` code from
`safekeepers_for_new_timeline`
- Change `timeline_safekeeper_count` type to `usize`
## Problem
`compute_ctl` should support gRPC base backups.
Requires #12111.
Requires #12243.
Touches #11926.
## Summary of changes
Support `grpc://` connstrings for `compute_ctl` base backups.
## Problem
The problem has been well described in already-commited PR #11853.
tl;dr: BufferedWriter is sensitive to cancellation, which the previous
approach was not.
The write path was most affected (ingest & compaction), which was mostly
fixed in #11853:
it introduced `PutError` and mapped instances of `PutError` that were
due to cancellation of underlying buffered writer into
`CreateImageLayersError::Cancelled`.
However, there is a long tail of remaining errors that weren't caught by
#11853 that result in `CompactionError::Other`s, which we log with great
noise.
## Solution
The stack trace logging for CompactionError::Other added in #11853
allows us to chop away at that long tail using the following pattern:
- look at the stack trace
- from leaf up, identify the place where we incorrectly map from the
distinguished variant X indicating cancellation to an `anyhow::Error`
- follow that anyhow further up, ensuring it stays the same anyhow all
the way up in the `CompactionError::Other`
- since it stayed one anyhow chain all the way up, root_cause() will
yield us X
- so, in `log_compaction_error`, add an additional `downcast_ref` check
for X
This PR specifically adds checks for
- the flush task cancelling (FlushTaskError, BlobWriterError)
- opening of the layer writer (GateError)
That should cover all the reports in issues
- https://github.com/neondatabase/cloud/issues/29434
- https://github.com/neondatabase/neon/issues/12162
## Refs
- follow-up to #11853
- fixup of / fixes https://github.com/neondatabase/neon/issues/11762
- fixes https://github.com/neondatabase/neon/issues/12162
- refs https://github.com/neondatabase/cloud/issues/29434
We don't have cancellation support for timeline deletions. In other
words, timeline deletion might still go on in an older generation while
we are attaching it in a newer generation already, because the
cancellation simply hasn't reached the deletion code.
This has caused us to hit a situation with offloaded timelines in which
the timeline was in an unrecoverable state: always returning an accepted
response, but never a 404 like it should be.
The detailed description can be found in
[here](https://github.com/neondatabase/cloud/issues/30406#issuecomment-3008667859)
(private repo link).
TLDR:
1. we ask to delete timeline on old pageserver/generation, starts
process in background
2. the storcon migrates the tenant to a different pageserver.
- during attach, the pageserver still finds an index part, so it adds it
to `offloaded_timelines`
4. the timeline deletion finishes, removing the index part in S3
5. there is a retry of the timeline deletion endpoint, sent to the new
pageserver location. it is bound to fail however:
- as the index part is gone, we print `Timeline already deleted in
remote storage`.
- the problem is that we then return an accepted response code, and not
a 404.
- this confuses the code calling us. it thinks the timeline is not
deleted, so keeps retrying.
- this state never gets recovered from until a reset/detach, because of
the `offloaded_timelines` entry staying there.
This is where this PR fixes things: if no index part can be found, we
can safely assume that the timeline is gone in S3 (it's the last thing
to be deleted), so we can remove it from `offloaded_timelines` and
trigger a reupload of the manifest. Subsequent retries will pick that
up.
Why not improve the cancellation support? It is a more disruptive code
change, that might have its own risks. So we don't do it for now.
Fixes https://github.com/neondatabase/cloud/issues/30406
Mainly for general readability. Some notable changes:
- Postgres can be built without the rest of the repository, and in
particular without any of the Rust bits. Some CI scripts took advantage
of that, so let's make that more explicit by separating those parts.
Also add an explicit comment about that in the new postgres.mk file.
- Add a new PG_INSTALL_CACHED variable. If it's set, `make all` and
other top-Makefile targets skip checking if Postgres is up-to-date. This
is also to be used in CI scripts that build and cache Postgres as
separate steps. (It is currently only used in the macos walproposer-lib
rule, but stay tuned for more.)
- Introduce a POSTGRES_VERSIONS variable that lists all supported
PostgreSQL versions. Refactor a few Makefile rules to use that.
## Problem
In large oltp test run
https://github.com/neondatabase/neon/actions/runs/15905488707/job/44859116742
we see that the `Benchmark database maintenance` step is skipped in all
3 strategy variants, however it should be executed in two.
This is due to treating the `test_maintenance` boolean type in the
strategy in the condition of the `Benchmark database maintenance` step
## Summary of changes
Use a boolean condition instead of a string comparison
## Test run from this pull request branch
https://github.com/neondatabase/neon/actions/runs/15923605412
## Problem
When resolving a shard during a split we might have multiple attached
shards with the old shard count (i.e. not all of them are marked in
progress and ignored). Hence, we can compute the desired shard number
based on the old shard count and misroute the request.
## Summary of Changes
Recompute the desired shard every time the shard count changes during
the iteration
# Problem
In #12335 I moved the `authenticate` method outside of the
`connect_to_compute` loop. This triggered [e2e tests to become
flaky](https://github.com/neondatabase/cloud/pull/30533). This
highlighted an edge case we forgot to consider with that change.
When we connect to compute, the compute IP might be cached. This cache
hit might however be stale. Because we can't validate the IP is
associated with a specific compute-id☨, we will succeed the
connect_to_compute operation and fail when it comes to password
authentication☨☨. Before the change, we were invalidating the cache and
triggering wake_compute if the authentication failed.
Additionally, I noticed some faulty logic I introduced 1 year ago
https://github.com/neondatabase/neon/pull/8141/files#diff-5491e3afe62d8c5c77178149c665603b29d88d3ec2e47fc1b3bb119a0a970afaL145-R147
☨ We can when we roll out TLS, as the certificate common name includes
the compute-id.
☨☨ Technically password authentication could pass for the wrong compute,
but I think this would only happen in the very very rare event that the
IP got reused **and** the compute's endpoint happened to be a
branch/replica.
# Solution
1. Fix the broken logic
2. Simplify cache invalidation (I don't know why it was so convoluted)
3. Add a loop around connect_to_compute + authenticate to re-introduce
the wake_compute invalidation we accidentally removed.
I went with this approach to try and avoid interfering with
https://github.com/neondatabase/neon/compare/main...cloneable/proxy-pglb-connect-compute-split.
The changes made in commit 3 will move into `handle_client_request` I
suspect,
## Problem
Some pageservers hit `max_size_entries` limit in staging with only ~25
MiB storage used by basebackup cache. The limit is too strict. It should
be safe to relax it.
- Part of https://github.com/neondatabase/cloud/issues/29353
## Summary of changes
- Increase the default `max_size_entries` from 1000 to 10000
## Problem
Druing shard splits we shut down the remote client early and allow the
parent shard to keep ingesting data. While ingesting data, the wal
receiver task may wait for the current flush to complete in order to
apply backpressure. Notifications are delivered via
`Timeline::layer_flush_done_tx`.
When the remote client was being shut down the flush loop exited
whithout delivering a notification. This left
`Timeline::wait_flush_completion` hanging indefinitely which blocked the
shutdown of the wal receiver task, and, hence, the shard split.
## Summary of Changes
Deliver a final notification when the flush loop is shutting down
without the timeline cancel cancellation token having fired. I tried
writing a test for this, but got stuck in failpoint hell and decided
it's not worth it.
`test_sharding_autosplit`, which reproduces this reliably in CI, passed
with the proposed fix in
https://github.com/neondatabase/neon/pull/12304.
Closes https://github.com/neondatabase/neon/issues/12060
Make the safekeeper `pull_timeline` endpoint support timelines that
haven't had any writes yet. In the storcon managed sk timelines world,
if a safekeeper goes down temporarily, the storcon will schedule a
`pull_timeline` call. There is no guarantee however that by when the
safekeeper is online again, there have been writes to the timeline yet.
The `snapshot` endpoint gives an error if the timeline hasn't had
writes, so we avoid calling it if `timeline_start_lsn` indicates a
freshly created timeline.
Fixes#11422
Part of #11670
## Problem
gRPC base backups do not use the base backup cache.
Touches https://github.com/neondatabase/neon/issues/11728.
## Summary of changes
Integrate gRPC base backups with the base backup cache.
Also fixes a bug where the base backup cache did not differentiate
between primary/replica base backups (at least I think that's a bug?).
## Problem
In our infra config, we have to split server_api_key and other fields in
two files: the former one in the sops file, and the latter one in the
normal config. It creates the situation that we might misconfigure some
regions that it only has part of the fields available, causing
storcon/pageserver refuse to start.
## Summary of changes
Allow PostHog config to have part of the fields available. Parse it
later.
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
Basebackup cache now uses unbounded channel for prepare requests. In
theory it can grow large if the cache is hung and does not process the
requests.
- Part of https://github.com/neondatabase/cloud/issues/29353
## Summary of changes
- Replace an unbounded channel with a bounded one, the size is
configurable.
- Add `pageserver_basebackup_cache_prepare_queue_size` to observe the
size of the queue.
- Refactor a bit to move all metrics logic to `basebackup_cache.rs`
I was looking at how we could expose our proxy config as toml again, and
as I was writing out the schema format, I noticed some cruft in our CLI
args that no longer seem to be in use.
The redis change is the most complex, but I am pretty sure it's sound.
Since https://github.com/neondatabase/cloud/pull/15613 cplane longer
publishes to the global redis instance.
## Problem
If LFC generation is changed then `lfc_readv_select` will return -1 but
pages are still marked as available in bitmap.
## Summary of changes
Update bitmap after generation check.
Co-authored-by: Kosntantin Knizhnik <konstantin.knizhnik@databricks.com>
## Problem
Fix for https://github.com/neondatabase/neon/pull/12324
## Summary of changes
Need `serde(default)` to allow this field not present in the config,
otherwise there will be a config deserialization error.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
gRPC base backups use gRPC compression. However, this has two problems:
* Base backup caching will cache compressed base backups (making gRPC
compression pointless).
* Tonic does not support varying the compression level, and zstd default
level is 10% slower than gzip fastest level.
Touches https://github.com/neondatabase/neon/issues/11728.
Touches https://github.com/neondatabase/cloud/issues/29353.
## Summary of changes
This patch adds a gRPC parameter `BaseBackupRequest::compression`
specifying the compression algorithm. It also moves compression into
`send_basebackup_tarball` to reduce code duplication.
A follow-up PR will integrate the base backup cache with gRPC.
## Summary
A design for a storage system that allows storage of files required to
make
Neon's Endpoints have a better experience at or after a reboot.
## Motivation
Several systems inside PostgreSQL (and Neon) need some persistent
storage for
optimal workings across reboots and restarts, but still work without.
Examples are the cumulative statistics file in `pg_stat/global.stat`,
`pg_stat_statements`' `pg_stat/pg_stat_statements.stat`, and
`pg_prewarm`'s
`autoprewarm.blocks`. We need a storage system that can store and manage
these files for each Endpoint.
[GH rendered
file](https://github.com/neondatabase/neon/blob/MMeent/rfc-unlogged-file/docs/rfcs/040-Endpoint-Persistent-Unlogged-Files-Storage.md)
Part of https://github.com/neondatabase/cloud/issues/24225
## Problem
part of https://github.com/neondatabase/neon/issues/11813
## Summary of changes
It costs $$$ to directly retrieve the feature flags from the pageserver.
Therefore, this patch adds new APIs to retrieve the spec from the
storcon and updates it via pageserver.
* Storcon retrieves the feature flag and send it to the pageservers.
* If the feature flag gets updated outside of the normal refresh loop of
the pageserver, pageserver won't fetch the flags on its own as long as
the last updated time <= refresh_period.
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
https://github.com/neondatabase/cloud/issues/30539
If the current leader cancels the `call` function, then it has removed
the jobs from the queue, but will never finish sending the responses.
Because of this, it is not cancellation safe.
## Summary of changes
Document these functions as not cancellation safe. Move cancellation of
the queued jobs into the queue itself.
## Alternatives considered
1. We could spawn the task that runs the batch, since that won't get
cancelled.
* This requires `fn call(self: Arc<Self>)` or `fn call(&'static self)`.
2. We could add another scopeguard and return the requests back to the
queue.
* This requires that requests are always retry safe, and also requires
requests to be `Clone`.
## Problem
https://github.com/neondatabase/cloud/issues/30539
If the current leader cancels the `call` function, then it has removed
the jobs from the queue, but will never finish sending the responses.
Because of this, it is not cancellation safe.
## Summary of changes
Document these functions as not cancellation safe. Move cancellation of
the queued jobs into the queue itself.
## Alternatives considered
1. We could spawn the task that runs the batch, since that won't get
cancelled.
* This requires `fn call(self: Arc<Self>)` or `fn call(&'static self)`.
2. We could add another scopeguard and return the requests back to the
queue.
* This requires that requests are always retry safe, and also requires
requests to be `Clone`.
## Problem
While working more on TLS to compute, I realised that Console Redirect
-> pg-sni-router -> compute would break if channel binding was set to
prefer. This is because the channel binding data would differ between
Console Redirect -> pg-sni-router vs pg-sni-router -> compute.
I also noticed that I actually disabled channel binding in #12145, since
`connect_raw` would think that the connection didn't support TLS.
## Summary of changes
Make sure we specify the channel binding.
Make sure that `connect_raw` can see if we have TLS support.
In `test_layer_download_cancelled_by_config_location`, we simulate hung
downloads via the `before-downloading-layer-stream-pausable` failpoint.
Then, we cancel a timeline via the `location_config` endpoint.
With the new default as of
https://github.com/neondatabase/neon/pull/11712, we would be creating
the timeline on safekeepers regardless if there have been writes or not,
and it turns out the test relied on the timeline not existing on
safekeepers, due to a cancellation bug:
* as established before, the test makes the read path hang
* the timeline cancellation function first cancels the walreceiver, and
only then cancels the timeline's token
* `WalIngest::new` is requesting a checkpoint, which hits the read path
* at cancellation time, we'd be hanging inside the read, not seeing the
cancellation of the walreceiver
* the test would time out due to the hang
This is probably also reproducible in the wild when there is S3
unavailabilies or bottlenecks. So we thought that it's worthwhile to fix
the hang issue. The approach chosen in the end involves the
`tokio::select` macro.
In PR 11712, we originally punted on the test due to the hang and opted
it out from the new default, but now we can use the new default.
Part of https://github.com/neondatabase/neon/issues/12299
pgaudit can spam logs due to all the monitoring that we do. Logs from
these connections are not necessary for HIPPA compliance, so we can stop
logging from those connections.
Part-of: https://github.com/neondatabase/cloud/issues/29574
Signed-off-by: Tristan Partin <tristan@neon.tech>
This makes it possible for the compiler to validate that a match block
matched all PostgreSQL versions we support.
## Problem
We did not have a complete picture about which places we had to test
against PG versions, and what format these versions were: The full PG
version ID format (Major/minor/bugfix `MMmmbb`) as transfered in
protocol messages, or only the Major release version (`MM`). This meant
type confusion was rampant.
With this change, it becomes easier to develop new version-dependent
features, by making type and niche confusion impossible.
## Summary of changes
Every use of `pg_version` is now typed as either `PgVersionId` (u32,
valued in decimal `MMmmbb`) or PgMajorVersion (an enum, with a value for
every major version we support, serialized and stored like a u32 with
the value of that major version)
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
The billing team wants to change the billing events pipeline and use a
common events format in S3 buckets across different event producers.
## Summary of changes
Change the events storage format for billing events from JSON to NDJSON.
Resolves: https://github.com/neondatabase/cloud/issues/29994
## Problem
PGLB will do the connect_to_compute logic, neonkeeper will do the
session establishment logic. We should split it.
## Summary of changes
Moves postgres authentication to compute to a separate routine that
happens after connect_to_compute.
Like #9931 but without rebasing upstream just yet, to try and minimise
the differences.
Removes all proxy-specific commits from the rust-postgres fork, now that
proxy no longer depends on them. Merging upstream changes to come later.
Closes#9387.
## Problem
`BufferedWriter` cannot proceed while the owned buffer is flushing to
disk. We want to implement double buffering so that the flush can happen
in the background. See #9387.
## Summary of changes
- Maintain two owned buffers in `BufferedWriter`.
- The writer is in charge of copying the data into owned, aligned
buffer, once full, submit it to the flush task.
- The flush background task is in charge of flushing the owned buffer to
disk, and returned the buffer to the writer for reuse.
- The writer and the flush background task communicate through a
bi-directional channel.
For in-memory layer, we also need to be able to read from the buffered
writer in `get_values_reconstruct_data`. To handle this case, we did the
following
- Use replace `VirtualFile::write_all` with `VirtualFile::write_all_at`,
and use `Arc` to share it between writer and background task.
- leverage `IoBufferMut::freeze` to get a cheaply clonable `IoBuffer`,
one clone will be submitted to the channel, the other clone will be
saved within the writer to serve reads. When we want to reuse the
buffer, we can invoke `IoBuffer::into_mut`, which gives us back the
mutable aligned buffer.
- InMemoryLayer reads is now aware of the maybe_flushed part of the
buffer.
**Caveat**
- We removed the owned version of write, because this interface does not
work well with buffer alignment. The result is that without direct IO
enabled,
[`download_object`](a439d57050/pageserver/src/tenant/remote_timeline_client/download.rs (L243))
does one more memcpy than before this PR due to the switch to use
`_borrowed` version of the write.
- "Bypass aligned part of write" could be implemented later to avoid
large amount of memcpy.
**Testing**
- use an oneshot channel based control mechanism to make flush behavior
deterministic in test.
- test reading from `EphemeralFile` when the last submitted buffer is
not flushed, in-progress, and done flushing to disk.
## Performance
We see performance improvement for small values, and regression on big
values, likely due to being CPU bound + disk write latency.
[Results](https://www.notion.so/neondatabase/Benchmarking-New-BufferedWriter-11-20-2024-143f189e0047805ba99acda89f984d51?pvs=4)
## 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
---------
Signed-off-by: Yuchen Liang <yuchen@neon.tech>
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
We have a scale test for the storage controller which also acts as a
good stress test for scheduling stability. However, it created nodes
with no AZs set.
## Summary of changes
- Bump node count to 6 and set AZs on them.
This is a precursor to other AZ-related PRs, to make sure any new code
that's landed is getting scale tested in an AZ-aware environment.
## Problem
We practice a manual release flow for the compute module. This will
allow automation of the compute release process.
## Summary of changes
The workflow was modified to make a compute release automatically on the
branch release-compute.
## 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
## Problem
Reqwest errors don't include details about the inner source error. This
means that we get opaque errors like:
```
receive body: error sending request for url (http://localhost:9898/v1/location_config)
```
Instead of the more helpful:
```
receive body: error sending request for url (http://localhost:9898/v1/location_config): operation timed out
```
Touches #9801.
## Summary of changes
Include the source error for `reqwest::Error` wherever it's displayed.
## Problem
When client specifies `application_name`, pgbouncer propagates it to the
Postgres. Yet, if client doesn't do it, we have hard time figuring out
who opens a lot of Postgres connections (including the `cloud_admin`
ones).
See this investigation as an example:
https://neondb.slack.com/archives/C0836R0RZ0D
## Summary of changes
I haven't found this documented, but it looks like pgbouncer accepts
standard Postgres connstring parameters in the connstring in the
`[databases]` section, so put the default `application_name=pgbouncer`
there. That way, we will always see who opens Postgres connections. I
did tests, and if client specifies a `application_name`, pgbouncer
overrides this default, so it only works if it's not specified or set to
blank `&application_name=` in the connection string.
This is the last place we could potentially open some Postgres
connections without `application_name`. Everything else should be either
of two:
1. Direct client connections without `application_name`, but these
should be strictly non-`cloud_admin` ones
2. Some ad-hoc internal connections, so if we see spikes of unidentified
`cloud_admin` connections, we will need to investigate it again.
Fixesneondatabase/cloud#20948
(stacked on #9990 and #9995)
Partially fixes#1287 with a custom option field to enable the fixed
behaviour. This allows us to gradually roll out the fix without silently
changing the observed behaviour for our customers.
related to https://github.com/neondatabase/cloud/issues/15284
## Problem
During deploys, we see a lot of 500 errors due to heapmap uploads for
inactive tenants. These should be 503s instead.
Resolves#9574.
## Summary of changes
Make the secondary tenant scheduler use `ApiError` rather than
`anyhow::Error`, to propagate the tenant error and convert it to an
appropriate status code.
## Problem
we tried different parallelism settings for ingest bench
## Summary of changes
the following settings seem optimal after merging
- SK side Wal filtering
- batched getpages
Settings:
- effective_io_concurrency 100
- concurrency limit 200 (different from Prod!)
- jobs 4, maintenance workers 7
- 10 GB chunk size
## Problem
```
2024-12-03T15:42:46.5978335Z + poetry run python /__w/neon/neon/scripts/ingest_perf_test_result.py --ingest /__w/neon/neon/test_runner/perf-report-local
2024-12-03T15:42:49.5325077Z Traceback (most recent call last):
2024-12-03T15:42:49.5325603Z File "/__w/neon/neon/scripts/ingest_perf_test_result.py", line 165, in <module>
2024-12-03T15:42:49.5326029Z main()
2024-12-03T15:42:49.5326316Z File "/__w/neon/neon/scripts/ingest_perf_test_result.py", line 155, in main
2024-12-03T15:42:49.5326739Z ingested = ingest_perf_test_result(cur, item, recorded_at_timestamp)
2024-12-03T15:42:49.5327488Z ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2024-12-03T15:42:49.5327914Z File "/__w/neon/neon/scripts/ingest_perf_test_result.py", line 99, in ingest_perf_test_result
2024-12-03T15:42:49.5328321Z psycopg2.extras.execute_values(
2024-12-03T15:42:49.5328940Z File "/github/home/.cache/pypoetry/virtualenvs/non-package-mode-_pxWMzVK-py3.11/lib/python3.11/site-packages/psycopg2/extras.py", line 1299, in execute_values
2024-12-03T15:42:49.5335618Z cur.execute(b''.join(parts))
2024-12-03T15:42:49.5335967Z psycopg2.errors.InvalidTextRepresentation: invalid input syntax for type numeric: "concurrent-futures"
2024-12-03T15:42:49.5336287Z LINE 57: 'concurrent-futures',
2024-12-03T15:42:49.5336462Z ^
```
## Summary of changes
- `test_page_service_batching`: save non-numeric params as `labels`
- Add a runtime check that `metric_value` is NUMERIC
Before this PR, some override callbacks used `.default()`, others
used `.setdefault()`.
As of this PR, all callbacks use `.setdefault()` which I think is least
prone to failure.
Aligning on a single way will set the right example for future tests
that need such customization.
The `test_pageserver_getpage_throttle.py` technically is a change in
behavior: before, it replaced the `tenant_config` field, now it just
configures the throttle. This is what I believe is intended anyway.
Support tenant manifests in the storage scrubber:
* list the manifests, order them by generation
* delete all manifests except for the two most recent generations
* for the latest manifest: try parsing it.
I've tested this patch by running the against a staging bucket and it
successfully deleted stuff (and avoided deleting the latest two
generations).
In follow-up work, we might want to also check some invariants of the
manifest, as mentioned in #8088.
Part of #9386
Part of #8088
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
The Pageserver signal handler would only respond to a single signal and
initiate shutdown. Subsequent signals were ignored. This meant that a
`SIGQUIT` sent after a `SIGTERM` had no effect (e.g. in the case of a
slow or stalled shutdown). The `test_runner` uses this to force shutdown
if graceful shutdown is slow.
Touches #9740.
## Summary of changes
Keep responding to signals after the initial shutdown signal has been
received.
Arguably, the `test_runner` should also use `SIGKILL` rather than
`SIGQUIT` in this case, but it seems reasonable to respond to `SIGQUIT`
regardless.
Keeping the `mock` postgres cplane adaptor using "stock" tokio-postgres
allows us to remove a lot of dead weight from our actual postgres
connection logic.
## Problem
We saw a peculiar case where a pageserver apparently got a 0-tenant
response to `/re-attach` but we couldn't see the request landing on a
storage controller. It was hard to confirm retrospectively that the
pageserver was configured properly at the moment it sent the request.
## Summary of changes
- Log the URL to which we are sending the request
- Log the NodeId and metadata that we sent
## Problem
Sharded tenants should be run in a single AZ for best performance, so
that computes have AZ-local latency to all the shards.
Part of https://github.com/neondatabase/neon/issues/8264
## Summary of changes
- When we split a tenant, instead of updating each shard's preferred AZ
to wherever it is scheduled, propagate the preferred AZ from the parent.
- Drop the check in `test_shard_preferred_azs` that asserts shards end
up in their preferred AZ: this will not be true again until the
optimize_attachment logic is updated to make this so. The existing check
wasn't testing anything about scheduling, it was just asserting that we
set preferred AZ in a way that matches the way things happen to be
scheduled at time of split.
## Problem
In the batching PR
- https://github.com/neondatabase/neon/pull/9870
I stopped deducting the time-spent-in-throttle fro latency metrics,
i.e.,
- smgr latency metrics (`SmgrOpTimer`)
- basebackup latency (+scan latency, which I think is part of
basebackup).
The reason for stopping the deduction was that with the introduction of
batching, the trick with tracking time-spent-in-throttle inside
RequestContext and swap-replacing it from the `impl Drop for
SmgrOpTimer` no longer worked with >1 requests in a batch.
However, deducting time-spent-in-throttle is desirable because our
internal latency SLO definition does not account for throttling.
## Summary of changes
- Redefine throttling to be a page_service pagestream request throttle
instead of a throttle for repository `Key` reads through `Timeline::get`
/ `Timeline::get_vectored`.
- This means reads done by `basebackup` are no longer subject to any
throttle.
- The throttle applies after batching, before handling of the request.
- Drive-by fix: make throttle sensitive to cancellation.
- Rename metric label `kind` from `timeline_get` to `pagestream` to
reflect the new scope of throttling.
To avoid config format breakage, we leave the config field named
`timeline_get_throttle` and ignore the `task_kinds` field.
This will be cleaned up in a future PR.
## Trade-Offs
Ideally, we would apply the throttle before reading a request off the
connection, so that we queue the minimal amount of work inside the
process.
However, that's not possible because we need to do shard routing.
The redefinition of the throttle to limit pagestream request rate
instead of repository `Key` rate comes with several downsides:
- We're no longer able to use the throttle mechanism for other other
tasks, e.g. image layer creation.
However, in practice, we never used that capability anyways.
- We no longer throttle basebackup.
## Problem
`test_sharded_ingest` ingests a lot of data, which can cause shutdown to
be slow e.g. due to local "S3 uploads" or compactions. This can cause
test flakes during teardown.
Resolves#9740.
## Summary of changes
Perform an immediate shutdown of the cluster.
## Problem
We don't have good observability for memory usage. This would be useful
e.g. to debug OOM incidents or optimize performance or resource usage.
We would also like to use continuous profiling with e.g. [Grafana Cloud
Profiles](https://grafana.com/products/cloud/profiles-for-continuous-profiling/)
(see https://github.com/neondatabase/cloud/issues/14888).
This PR is intended as a proof of concept, to try it out in staging and
drive further discussions about profiling more broadly.
Touches https://github.com/neondatabase/neon/issues/9534.
Touches https://github.com/neondatabase/cloud/issues/14888.
Depends on #9779.
Depends on #9780.
## Summary of changes
Adds a HTTP route `/profile/heap` that takes a heap profile and returns
it. Query parameters:
* `format`: output format (`jemalloc` or `pprof`; default `pprof`).
Unlike CPU profiles (see #9764), heap profiles are not symbolized and
require the original binary to translate addresses to function names. To
make this work with Grafana, we'll probably have to symbolize the
process server-side -- this is left as future work, as is other output
formats like SVG.
Heap profiles don't work on macOS due to limitations in jemalloc.
## Problem
The extensions for Postgres v17 are ready but we do not test the
extensions shipped with v17
## Summary of changes
Build the test image based on Postgres v17. Run the tests for v17.
---------
Co-authored-by: Anastasia Lubennikova <anastasia@neon.tech>
This PR
- fixes smgr metrics https://github.com/neondatabase/neon/issues/9925
- adds an additional startup log line logging the current batching
config
- adds a histogram of batch sizes global and per-tenant
- adds a metric exposing the current batching config
The issue described #9925 is that before this PR, request latency was
only observed *after* batching.
This means that smgr latency metrics (most importantly getpage latency)
don't account for
- `wait_lsn` time
- time spent waiting for batch to fill up / the executor stage to pick
up the batch.
The fix is to use a per-request batching timer, like we did before the
initial batching PR.
We funnel those timers through the entire request lifecycle.
I noticed that even before the initial batching changes, we weren't
accounting for the time spent writing & flushing the response to the
wire.
This PR drive-by fixes that deficiency by dropping the timers at the
very end of processing the batch, i.e., after the `pgb.flush()` call.
I was **unable to maintain the behavior that we deduct
time-spent-in-throttle from various latency metrics.
The reason is that we're using a *single* counter in `RequestContext` to
track micros spent in throttle.
But there are *N* metrics timers in the batch, one per request.
As a consequence, the practice of consuming the counter in the drop
handler of each timer no longer works because all but the first timer
will encounter error `close() called on closed state`.
A failed attempt to maintain the current behavior can be found in
https://github.com/neondatabase/neon/pull/9951.
So, this PR remvoes the deduction behavior from all metrics.
I started a discussion on Slack about it the implications this has for
our internal SLO calculation:
https://neondb.slack.com/archives/C033RQ5SPDH/p1732910861704029
# Refs
- fixes https://github.com/neondatabase/neon/issues/9925
- sub-issue https://github.com/neondatabase/neon/issues/9377
- epic: https://github.com/neondatabase/neon/issues/9376
Before this PR, the storcon_cli didn't have a way to show the
tenant-wide information of the TenantDescribeResponse.
Sadly, the `Serialize` impl for the tenant config doesn't skip on
`None`, so, the output becomes a bit bloated.
Maybe we can use `skip_serializing_if(Option::is_none)` in the future.
=> https://github.com/neondatabase/neon/issues/9983
## Problem
I was touching `test_storage_controller_node_deletion` because for AZ
scheduling work I was adding a change to the storage controller (kick
secondaries during optimisation) that made a FIXME in this test defunct.
While looking at it I also realized that we can easily fix the way node
deletion currently doesn't use a proper ScheduleContext, using the
iterator type recently added for that purpose.
## Summary of changes
- A testing-only behavior in storage controller where if a secondary
location isn't yet ready during optimisation, it will be actively
polled.
- Remove workaround in `test_storage_controller_node_deletion` that
previously was needed because optimisation would get stuck on cold
secondaries.
- Update node deletion code to use a `TenantShardContextIterator` and
thereby a proper ScheduleContext
## Problem
After enabling LFC in tests and lowering `shared_buffers` we started
having more problems with `test_pg_regress`.
## Summary of changes
Set `shared_buffers` to 1MB to both exercise getPage requests/LFC, and
still have enough room for Postgres to operate. Everything smaller might
be not enough for Postgres under load, and can cause errors like 'no
unpinned buffers available'.
See Konstantin's comment [1] as well.
Fixes#9956
[1]:
https://github.com/neondatabase/neon/issues/9956#issuecomment-2511608097
On reconfigure, we no longer passed a port for the extension server
which caused us to not write out the neon.extension_server_port line.
Thus, Postgres thought we were setting the port to the default value of
0. PGC_POSTMASTER GUCs cannot be set at runtime, which causes the
following log messages:
> LOG: parameter "neon.extension_server_port" cannot be changed without
restarting the server
> LOG: configuration file
"/var/db/postgres/compute/pgdata/postgresql.conf" contains errors;
unaffected changes were applied
Fixes: https://github.com/neondatabase/neon/issues/9945
Signed-off-by: Tristan Partin <tristan@neon.tech>
The spec was written for the buggy protocol which we had before the one
more similar to Raft was implemented. Update the spec with what we
currently have.
ref https://github.com/neondatabase/neon/issues/8699
## Problem
The credentials providers tries to connect to AWS STS even when we use
plain Redis connections.
## Summary of changes
* Construct the CredentialsProvider only when needed ("irsa").
## Problem
`if: ${{ github.event.schedule }}` gets skipped if a previous step has
failed, but we want to run the step for both `success` and `failure`
## Summary of changes
- Add `!cancelled()` to notification step if-condition, to skip only
cancelled jobs
Fixes https://github.com/neondatabase/cloud/issues/20973.
This refactors `connect_raw` in order to return direct access to the
delayed notices.
I cannot find a way to test this with psycopg2 unfortunately, although
testing it with psql does return the expected results.
## Problem
We can't easily tell how far the state of shards is from their AZ
preferences. This can be a cause of performance issues, so it's
important for diagnosability that we can tell easily if there are
significant numbers of shards that aren't running in their preferred AZ.
Related: https://github.com/neondatabase/cloud/issues/15413
## Summary of changes
- In reconcile_all, count shards that are scheduled into the wrong AZ
(if they have a preference), and publish it as a prometheus gauge.
- Also calculate a statistic for how many shards wanted to reconcile but
couldn't.
This is clearly a lazy calculation: reconcile all only runs
periodically. But that's okay: shards in the wrong AZ is something that
only matters if it stays that way for some period of time.
Improves `wait_until` by:
* Use `timeout` instead of `iterations`. This allows changing the
timeout/interval parameters independently.
* Make `timeout` and `interval` optional (default 20s and 0.5s). Most
callers don't care.
* Only output status every 1s by default, and add optional
`status_interval` parameter.
* Remove `show_intermediate_error`, this was always emitted anyway.
Most callers have been updated to use the defaults, except where they
had good reason otherwise.
## Problem
We saw unexpected container terminations when running in k8s with with
small CPU resource requests.
The /status and /ready handlers called `maybe_forward`, which always
takes the lock on Service::inner.
If there is a lot of writer lock contention, and the container is
starved of CPU, this increases the likelihood that we will get killed by
the kubelet.
It isn't certain that this was a cause of issues, but it is a potential
source that we can eliminate.
## Summary of changes
- Revise logic to return immediately if the URL is in the non-forwarded
list, rather than calling maybe_forward
## Problem
See https://neondb.slack.com/archives/C04DGM6SMTM/p1732110190129479
We observe the following error in the logs
```
[XX000] ERROR: [NEON_SMGR] [shard 3] Incorrect prefetch read: status=1 response=0x7fafef335138 my=128 receive=128
```
most likely caused by changing `neon.readahead_buffer_size`
## Summary of changes
1. Copy shard state
2. Do not use prefetch_set_unused in readahead_buffer_resize
3. Change prefetch buffer overflow criteria
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
Current compute images for Postgres 14-16 don't build on Debian 12
because of issues with extensions.
This PR fixes that, but for the current setup, it is mostly a no-op
change.
## Summary of changes
- Use `/bin/bash -euo pipefail` as SHELL to fail earlier
- Fix `plv8` build: backport a trivial patch for v8
- Fix `postgis` build: depend `sfgal` version on Debian version instead
of Postgres version
Tested in: https://github.com/neondatabase/neon/pull/9849
#8564
## Problem
The main and backup consumption metric pushes are completely
independent,
resulting in different event time windows and different idempotency
keys.
## Summary of changes
* Merge the push tasks, but keep chunks the same size.
# Problem
The timeout-based batching adds latency to unbatchable workloads.
We can choose a short batching timeout (e.g. 10us) but that requires
high-resolution timers, which tokio doesn't have.
I thoroughly explored options to use OS timers (see
[this](https://github.com/neondatabase/neon/pull/9822) abandoned PR).
In short, it's not an attractive option because any timer implementation
adds non-trivial overheads.
# Solution
The insight is that, in the steady state of a batchable workload, the
time we spend in `get_vectored` will be hundreds of microseconds anyway.
If we prepare the next batch concurrently to `get_vectored`, we will
have a sizeable batch ready once `get_vectored` of the current batch is
done and do not need an explicit timeout.
This can be reasonably described as **pipelining of the protocol
handler**.
# Implementation
We model the sub-protocol handler for pagestream requests
(`handle_pagrequests`) as two futures that form a pipeline:
2. Batching: read requests from the connection and fill the current
batch
3. Execution: `take` the current batch, execute it using `get_vectored`,
and send the response.
The Reading and Batching stage are connected through a new type of
channel called `spsc_fold`.
See the long comment in the `handle_pagerequests_pipelined` for details.
# Changes
- Refactor `handle_pagerequests`
- separate functions for
- reading one protocol message; produces a `BatchedFeMessage` with just
one page request in it
- batching; tried to merge an incoming `BatchedFeMessage` into an
existing `BatchedFeMessage`; returns `None` on success and returns back
the incoming message in case merging isn't possible
- execution of a batched message
- unify the timeline handle acquisition & request span construction; it
now happen in the function that reads the protocol message
- Implement serial and pipelined model
- serial: what we had before any of the batching changes
- read one protocol message
- execute protocol messages
- pipelined: the design described above
- optionality for execution of the pipeline: either via concurrent
futures vs tokio tasks
- Pageserver config
- remove batching timeout field
- add ability to configure pipelining mode
- add ability to limit max batch size for pipelined configurations
(required for the rollout, cf
https://github.com/neondatabase/cloud/issues/20620 )
- ability to configure execution mode
- Tests
- remove `batch_timeout` parametrization
- rename `test_getpage_merge_smoke` to `test_throughput`
- add parametrization to test different max batch sizes and execution
moes
- rename `test_timer_precision` to `test_latency`
- rename the test case file to `test_page_service_batching.py`
- better descriptions of what the tests actually do
## On the holding The `TimelineHandle` in the pending batch
While batching, we hold the `TimelineHandle` in the pending batch.
Therefore, the timeline will not finish shutting down while we're
batching.
This is not a problem in practice because the concurrently ongoing
`get_vectored` call will fail quickly with an error indicating that the
timeline is shutting down.
This results in the Execution stage returning a `QueryError::Shutdown`,
which causes the pipeline / entire page service connection to shut down.
This drops all references to the
`Arc<Mutex<Option<Box<BatchedFeMessage>>>>` object, thereby dropping the
contained `TimelineHandle`s.
- => fixes https://github.com/neondatabase/neon/issues/9850
# Performance
Local run of the benchmarks, results in [this empty
commit](1cf5b1463f)
in the PR branch.
Key take-aways:
* `concurrent-futures` and `tasks` deliver identical `batching_factor`
* tail latency impact unknown, cf
https://github.com/neondatabase/neon/issues/9837
* `concurrent-futures` has higher throughput than `tasks` in all
workloads (=lower `time` metric)
* In unbatchable workloads, `concurrent-futures` has 5% higher
`CPU-per-throughput` than that of `tasks`, and 15% higher than that of
`serial`.
* In batchable-32 workload, `concurrent-futures` has 8% lower
`CPU-per-throughput` than that of `tasks` (comparison to tput of
`serial` is irrelevant)
* in unbatchable workloads, mean and tail latencies of
`concurrent-futures` is practically identical to `serial`, whereas
`tasks` adds 20-30us of overhead
Overall, `concurrent-futures` seems like a slightly more attractive
choice.
# Rollout
This change is disabled-by-default.
Rollout plan:
- https://github.com/neondatabase/cloud/issues/20620
# Refs
- epic: https://github.com/neondatabase/neon/issues/9376
- this sub-task: https://github.com/neondatabase/neon/issues/9377
- the abandoned attempt to improve batching timeout resolution:
https://github.com/neondatabase/neon/pull/9820
- closes https://github.com/neondatabase/neon/issues/9850
- fixes https://github.com/neondatabase/neon/issues/9835
## Problem
It appears that the Azure storage API tends to hang TCP connections more
than S3 does.
Currently we use a 2 minute timeout for all downloads. This is large
because sometimes the objects we download are large. However, waiting 2
minutes when doing something like downloading a manifest on tenant
attach is problematic, because when someone is doing a "create tenant,
create timeline" workflow, that 2 minutes is long enough for them
reasonably to give up creating that timeline.
Rather than propagate oversized timeouts further up the stack, we should
use a different timeout for objects that we expect to be small.
Closes: https://github.com/neondatabase/neon/issues/9836
## Summary of changes
- Add a `small_timeout` configuration attribute to remote storage,
defaulting to 30 seconds (still a very generous period to do something
like download an index)
- Add a DownloadKind parameter to DownloadOpts, so that callers can
indicate whether they expect the object to be small or large.
- In the azure client, use small timeout for HEAD requests, and for GET
requests if DownloadKind::Small is used.
- Use DownloadKind::Small for manifests, indices, and heatmap downloads.
This PR intentionally does not make the equivalent change to the S3
client, to reduce blast radius in case this has unexpected consequences
(we could accomplish the same thing by editing lots of configs, but just
skipping the code is simpler for right now)
## Problem
It was not always possible to judge what exactly some `cloud_admin`
connections were doing because we didn't consistently set
`application_name` everywhere.
## Summary of changes
Unify the way we connect to Postgres:
1. Switch to building configs everywhere
2. Always set `application_name` and make naming consistent
Follow-up for #9919
Part of neondatabase/cloud#20948
## Problem
To add Safekeeper heap profiling in #9778, we need to switch to an
allocator that supports it. Pageserver and proxy already use jemalloc.
Touches #9534.
## Summary of changes
Use jemalloc in Safekeeper.
## Problem
When picking locations for a shard, we should use a ScheduleContext that
includes all the other shards in the tenant, so that we apply proper
anti-affinity between shards. If we don't do this, then it can lead to
unstable scheduling, where we place a shard somewhere that the optimizer
will then immediately move it away from.
We didn't always do this, because it was a bit awkward to accumulate the
context for a tenant rather than just walking tenants.
This was a TODO in `handle_node_availability_transition`:
```
// TODO: populate a ScheduleContext including all shards in the same tenant_id (only matters
// for tenants without secondary locations: if they have a secondary location, then this
// schedule() call is just promoting an existing secondary)
```
This is a precursor to https://github.com/neondatabase/neon/issues/8264,
where the current imperfect scheduling during node evacuation hampers
testing.
## Summary of changes
- Add an iterator type that yields each shard along with a
schedulecontext that includes all the other shards from the same tenant
- Use the iterator to replace hand-crafted logic in optimize_all_plan
(functionally identical)
- Use the iterator in `handle_node_availability_transition` to apply
proper anti-affinity during node evacuation.
Our rust-postgres fork is getting messy. Mostly because proxy wants more
control over the raw protocol than tokio-postgres provides. As such,
it's diverging more and more. Storage and compute also make use of
rust-postgres, but in more normal usage, thus they don't need our crazy
changes.
Idea:
* proxy maintains their subset
* other teams use a minimal patch set against upstream rust-postgres
Reviewing this code will be difficult. To implement it, I
1. Copied tokio-postgres, postgres-protocol and postgres-types from
00940fcdb5
2. Updated their package names with the `2` suffix to make them compile
in the workspace.
3. Updated proxy to use those packages
4. Copied in the code from tokio-postgres-rustls 0.13 (with some patches
applied https://github.com/jbg/tokio-postgres-rustls/pull/32https://github.com/jbg/tokio-postgres-rustls/pull/33)
5. Removed as much dead code as I could find in the vendored libraries
6. Updated the tokio-postgres-rustls code to use our existing channel
binding implementation
Adds a benchmark for logical message WAL ingestion throughput
end-to-end. Logical messages are essentially noops, and thus ignored by
the Pageserver.
Example results from my MacBook, with fsync enabled:
```
postgres_ingest: 14.445 s
safekeeper_ingest: 29.948 s
pageserver_ingest: 30.013 s
pageserver_recover_ingest: 8.633 s
wal_written: 10,340 MB
message_count: 1310720 messages
postgres_throughput: 715 MB/s
safekeeper_throughput: 345 MB/s
pageserver_throughput: 344 MB/s
pageserver_recover_throughput: 1197 MB/s
```
See
https://github.com/neondatabase/neon/issues/9642#issuecomment-2475995205
for running analysis.
Touches #9642.
## Problem
We used `set_path()` to replace the database name in the connection
string. It automatically does url-safe encoding if the path is not
already encoded, but it does it as per the URL standard, which assumes
that tabs can be safely removed from the path without changing the
meaning of the URL. See, e.g.,
https://url.spec.whatwg.org/#concept-basic-url-parser. It also breaks
for DBs with properly %-encoded names, like with `%20`, as they are kept
intact, but actually should be escaped.
Yet, this is not true for Postgres, where it's completely valid to have
trailing tabs in the database name.
I think this is the PR that caused this regression
https://github.com/neondatabase/neon/pull/9717, as it switched from
`postgres::config::Config` back to `set_path()`.
This was fixed a while ago already [1], btw, I just haven't added a test
to catch this regression back then :(
## Summary of changes
This commit changes the code back to use
`postgres/tokio_postgres::Config` everywhere.
While on it, also do some changes around, as I had to touch this code:
1. Bump some logging from `debug` to `info` in the spec apply path. We
do not use `debug` in prod, and it was tricky to understand what was
going on with this bug in prod.
2. Refactor configuration concurrency calculation code so it was
reusable. Yet, still keep `1` in the case of reconfiguration. The
database can be actively used at this moment, so we cannot guarantee
that there will be enough spare connection slots, and the underlying
code won't handle connection errors properly.
3. Simplify the installed extensions code. It was spawning a blocking
task inside async function, which doesn't make much sense. Instead, just
have a main sync function and call it with `spawn_blocking` in the API
code -- the only place we need it to be async.
4. Add regression python test to cover this and related problems in the
future. Also, add more extensive testing of schema dump and DBs and
roles listing API.
[1]:
4d1e48f3b9
[2]:
https://www.postgresql.org/message-id/flat/20151023003445.931.91267%40wrigleys.postgresql.orgResolvesneondatabase/cloud#20869
## Problem
Currently, we rerun only known flaky tests. This approach was chosen to
reduce the number of tests that go unnoticed (by forcing people to take
a look at failed tests and rerun the job manually), but it has some
drawbacks:
- In PRs, people tend to push new changes without checking failed tests
(that's ok)
- In the main, tests are just restarted without checking
(understandable)
- Parametrised tests become flaky one by one, i.e. if `test[1]` is flaky
`, test[2]` is not marked as flaky automatically (which may or may not
be the case).
I suggest rerunning all failed tests to increase the stability of GitHub
jobs and using the Grafana Dashboard with flaky tests for deeper
analysis.
## Summary of changes
- Rerun all failed tests twice at max
## Problem
For the interpreted proto the pageserver is not returning the correct
LSN
in replies to keep alive requests. This is because the interpreted
protocol arm
was not updating `last_rec_lsn`.
## Summary of changes
* Return correct LSN in keep-alive responses
* Fix shard field in wal sender traces
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://releases.rs/docs/1.83.0/).
Also update `cargo-hakari`, `cargo-deny`, `cargo-hack` and
`cargo-nextest` to their latest versions.
Prior update was in #9445.
## Problem
We currently see elevated levels of errors for GetBlob requests. This is
because 404 and 304 are counted as errors for metric reporting.
## Summary of Changes
Bring the implementation in line with the S3 client and treat 404 and
304 responses as ok for metric purposes.
Related: https://github.com/neondatabase/cloud/issues/20666
## Problem
For cancellation, a connection is open during all the cancel checks.
## Summary of changes
Spawn cancellation checks in the background, and close connection
immediately.
Use task_tracker for cancellation checks.
## 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
In a recent refactor, we accidentally dropped the cancel session early
## Summary of changes
Hold the cancel session during proxy passthrough
## 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.
## 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
## 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>
There is currently no cleanup done after a delta layer creation error,
so delta layers can accumulate. The problem gets worse as the operation
gets retried and delta layers accumulate on the disk. Therefore, delete
them from disk (if something has been written to disk).
## Problem
When a tenant is in Attaching state, and waiting for the
`concurrent_tenant_warmup` semaphore, it also listens for the tenant
cancellation token. When that token fires, Tenant::attach drops out.
Meanwhile, Tenant::set_stopping waits forever for the tenant to exit
Attaching state.
Fixes: https://github.com/neondatabase/neon/issues/6423
## Summary of changes
- In the absence of a valid state for the tenant, it is set to Broken in
this path. A more elegant solution will require more refactoring, beyond
this minimal fix.
(cherry picked from commit 93572a3e99)
Before this patch, the select! still retured immediately if `futs` was
empty. Must have tested a stale build in my manual testing of #6388.
(cherry picked from commit 15c0df4de7)
To exercise MAX_SEND_SIZE sending from safekeeper; we've had a bug with WAL
records torn across several XLogData messages. Add failpoint to safekeeper to
slow down sending. Also check for corrupted WAL complains in standby log.
Make the test a bit simpler in passing, e.g. we don't need explicit commits as
autocommit is enabled by default.
https://neondb.slack.com/archives/C05L7D1JAUS/p1703774799114719https://github.com/neondatabase/cloud/issues/9057
Otherwise they are left orphaned when compute_ctl is terminated with a
signal. It was invisible most of the time because normally neon_local or k8s
kills postgres directly and then compute_ctl finishes gracefully. However, in
some tests compute_ctl gets stuck waiting for sync-safekeepers which
intentionally never ends because safekeepers are offline, and we want to stop
compute_ctl without leaving orphanes behind.
This is a quite rough approach which doesn't wait for children termination. A
better way would be to convert compute_ctl to async which would make waiting
easy.
Release 2023-12-19
We need to do a config change that requires restarting the pageservers.
Slip in two metrics-related commits that didn't make this week's regularly release.
Pre-merge `git merge --squash` of
https://github.com/neondatabase/neon/pull/6115
Lowering the tracing level in get_value_reconstruct_data and
get_or_maybe_download from info to debug reduces the overhead
of span creation in non-debug environments.
## Problem
#6112 added some logs and metrics: clean these up a bit:
- Avoid counting startup completions for tenants launched after startup
- exclude no-op cases from timing histograms
- remove a rogue log messages
Error indicating request cancellation OR timeline shutdown was deemed as
a reason to exit the background worker that calculated synthetic size.
Fix it to only be considered for avoiding logging such of such errors.
This conflicted on tenant_shard_id having already replaced tenant_id on
`main`.
```
could not start the compute node: compute is in state "failed": db error: ERROR: could not access file "$libdir/timescaledb-2.10.1": No such file or directory Caused by: ERROR: could not access file "$libdir/timescaledb-2.10.1": No such file or directory
```
Only applicable change was neondatabase/autoscaling#584, setting
pgbouncer auth_dbname=postgres in order to fix superuser connections
from preventing dropping databases.
Only applicable change was neondatabase/autoscaling#571, removing the
postgres_exporter flags `--auto-discover-databases` and
`--exclude-databases=...`
## Problem
Logical replication requires new AUX_FILES_KEY which is definitely
absent in existed database.
We do not have function to check if key exists in our KV storage.
So I have to handle the error in `list_aux_files` method.
But this key is also included in key space range and accessed y
`create_image_layer` method.
## Summary of changes
Check if AUX_FILES_KEY exists before including it in keyspace.
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
Co-authored-by: Shany Pozin <shany@neon.tech>
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
Fixes an issue we observed on staging that happens when the
autoscaler-agent attempts to immediately downscale the VM after binding,
which is typical for pooled computes.
The issue was occurring because the autoscaler-agent was requesting
downscaling before the vm-monitor had gathered sufficient cgroup memory
stats to be confident in approving it. When the vm-monitor returned an
internal error instead of denying downscaling, the autoscaler-agent
retried the connection and immediately hit the same issue (in part
because cgroup stats are collected per-connection, rather than
globally).
There's currently an issue with the vm-monitor on staging that's not
really feasible to debug because the current display impl gives no
context to the errors (just says "failed to downscale").
Logging the full error should help.
For communications with the autoscaler-agent, it's ok to only provide
the outermost cause, because we can cross-reference with the VM logs.
At some point in the future, we may want to change that.
tl;dr it's really hard to avoid throttling from memory.high, and it
counts tmpfs & page cache usage, so it's also hard to make sense of.
In the interest of fixing things quickly with something that should be
*good enough*, this PR switches to instead periodically fetch memory
statistics from the cgroup's memory.stat and use that data to determine
if and when we should upscale.
This PR fixes#5444, which has a lot more detail on the difficulties
we've hit with memory.high. This PR also supersedes #5488.
Before this PR, when we restarted pageserver, we'd see a rush of
`$number_of_tenants` concurrent eviction tasks starting to do imitate
accesses building up in the period of `[init_order allows activations,
$random_access_delay + EvictionPolicyLayerAccessThreshold::period]`.
We simply cannot handle that degree of concurrent IO.
We already solved the problem for compactions by adding a semaphore.
So, this PR shares that semaphore for use by evictions.
Part of https://github.com/neondatabase/neon/issues/5479
Which is again part of https://github.com/neondatabase/neon/issues/4743
Risks / Changes In System Behavior
==================================
* we don't do evictions as timely as we currently do
* we log a bunch of warnings about eviction taking too long
* imitate accesses and compactions compete for the same concurrency
limit, so, they'll slow each other down through this shares semaphore
Changes
=======
- Move the `CONCURRENT_COMPACTIONS` semaphore into `tasks.rs`
- Rename it to `CONCURRENT_BACKGROUND_TASKS`
- Use it also for the eviction imitate accesses:
- Imitate acceses are both per-TIMELINE and per-TENANT
- The per-TENANT is done through coalescing all the per-TIMELINE
tasks via a tokio mutex `eviction_task_tenant_state`.
- We acquire the CONCURRENT_BACKGROUND_TASKS permit early, at the
beginning of the eviction iteration, much before the imitate
acesses start (and they may not even start at all in the given
iteration, as they happen only every $threshold).
- Acquiring early is **sub-optimal** because when the per-timline
tasks coalesce on the `eviction_task_tenant_state` mutex,
they are already holding a CONCURRENT_BACKGROUND_TASKS permit.
- It's also unfair because tenants with many timelines win
the CONCURRENT_BACKGROUND_TASKS more often.
- I don't think there's another way though, without refactoring
more of the imitate accesses logic, e.g, making it all per-tenant.
- Add metrics for queue depth behind the semaphore.
I found these very useful to understand what work is queued in the
system.
- The metrics are tagged by the new `BackgroundLoopKind`.
- On a green slate, I would have used `TaskKind`, but we already had
pre-existing labels whose names didn't map exactly to task kind.
Also the task kind is kind of a lower-level detail, so, I think
it's fine to have a separate enum to identify background work kinds.
Future Work
===========
I guess I could move the eviction tasks from a ticker to "sleep for
$period".
The benefit would be that the semaphore automatically "smears" the
eviction task scheduling over time, so, we only have the rush on restart
but a smeared-out rush afterward.
The downside is that this perverts the meaning of "$period", as we'd
actually not run the eviction at a fixed period. It also means the the
"took to long" warning & metric becomes meaningless.
Then again, that is already the case for the compaction and gc tasks,
which do sleep for `$period` instead of using a ticker.
(cherry picked from commit 9256788273)
## Problem
Folks have re-taged releases for `pg_jsonschema` and `pg_graphql` (to
increase timeouts on their CI), for us, these are a noop changes,
but unfortunately, this will cause our builds to fail due to checksums
mismatch (this might not strike right away because of the build cache).
- 8ba7c7be9d
- aa7509370a
## Summary of changes
- `pg_jsonschema` update checksum
- `pg_graphql` update checksum
When you log more than a few blocks, you need to reserve the space in
advance. We didn't do that, so we got errors. Now we do that, and
shouldn't get errors.
## Problem
See https://neondb.slack.com/archives/C05L7D1JAUS/p1694614585955029https://www.notion.so/neondatabase/Duplicate-key-issue-651627ce843c45188fbdcb2d30fd2178
## Summary of changes
Swap old/new block references
## 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: Heikki Linnakangas <heikki@neon.tech>
The sequence that can lead to a deadlock:
1. DELETE request gets all the way to `tenant.shutdown(progress,
false).await.is_err() ` , while holding TENANTS.read()
2. POST request for tenant creation comes in, calls `tenant_map_insert`,
it does `let mut guard = TENANTS.write().await;`
3. Something that `tenant.shutdown()` needs to wait for needs a
`TENANTS.read().await`.
The only case identified in exhaustive manual scanning of the code base
is this one:
Imitate size access does `get_tenant().await`, which does
`TENANTS.read().await` under the hood.
In the above case (1) waits for (3), (3)'s read-lock request is queued
behind (2)'s write-lock, and (2) waits for (1).
Deadlock.
I made a reproducer/proof-that-above-hypothesis-holds in
https://github.com/neondatabase/neon/pull/5281 , but, it's not ready for
merge yet and we want the fix _now_.
fixes https://github.com/neondatabase/neon/issues/5284
## Problem
We were returning Pending when a connection had a notice/notification
(introduced recently in #5020). When returning pending, the runtime
assumes you will call `cx.waker().wake()` in order to continue
processing.
We weren't doing that, so the connection task would get stuck
## Summary of changes
Don't return pending. Loop instead
## Problem
cargo deny lint broken
Links to the CVEs:
[rustsec.org/advisories/RUSTSEC-2023-0052](https://rustsec.org/advisories/RUSTSEC-2023-0052)
[rustsec.org/advisories/RUSTSEC-2023-0053](https://rustsec.org/advisories/RUSTSEC-2023-0053)
One is fixed, the other one isn't so we allow it (for now), to unbreak
CI. Then later we'll try to get rid of webpki in favour of the rustls
fork.
## Summary of changes
```
+ignore = ["RUSTSEC-2023-0052"]
```
## Problem
When an endpoint is shutting down, it can take a few seconds. Currently
when starting a new compute, this causes an "endpoint is in transition"
error. We need to add delays before retrying to ensure that we allow
time for the endpoint to shutdown properly.
## Summary of changes
Adds a delay before retrying in auth. connect_to_compute already has
this delay
commit
commit 5f8fd640bf
Author: Alek Westover <alek.westover@gmail.com>
Date: Wed Jul 26 08:24:03 2023 -0400
Upload Test Remote Extensions (#4792)
switched to using the release tag instead of `latest`, but,
the `promote-images` job only uploads `latest` to the prod ECR.
The switch to using release tag was good in principle, but,
reverting that part to make the release pipeine work.
Note that a proper fix should abandon use of `:latest` tag
at all: currently, if a `main` pipeline runs concurrently
with a `release` pipeline, the `release` pipeline may end
up using the `main` pipeline's images.
## Problem
If we fail to wake up the compute node, a subsequent connect attempt
will definitely fail. However, kubernetes won't fail the connection
immediately, instead it hangs until we timeout (10s).
## Summary of changes
Refactor the loop to allow fast retries of compute_wake and to skip a
connect attempt.
## Problem
#4598 compute nodes are not accessible some time after wake up due to
kubernetes DNS not being fully propagated.
## Summary of changes
Update connect retry mechanism to support handling IO errors and
sleeping for 100ms
## 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.
```
CREATE EXTENSION embedding;
CREATE TABLE t (val real[]);
INSERT INTO t (val) VALUES ('{0,0,0}'), ('{1,2,3}'), ('{1,1,1}'), (NULL);
CREATE INDEX ON t USING hnsw (val) WITH (maxelements = 10, dims=3, m=3);
INSERT INTO t (val) VALUES (array[1,2,4]);
SELECT * FROM t ORDER BY val <-> array[3,3,3];
val
---------
{1,2,3}
{1,2,4}
{1,1,1}
{0,0,0}
(5 rows)
```
The consumption metrics synthetic size worker does logical size calculation.
Logical size calculation currently does synchronous disk IO.
This blocks the MGMT_REQUEST_RUNTIME's executor threads, starving other futures.
While there's work on the way to move the synchronous disk IO into spawn_blocking,
the quickfix here is to use the BACKGROUND_RUNTIME instead of MGMT_REQUEST_RUNTIME.
Actually it's not just a quickfix. We simply shouldn't be blocking MGMT_REQUEST_RUNTIME
executor threads on CPU or sync disk IO.
That work isn't done yet, as many of the mgmt tasks still _do_ disk IO.
But it's not as intensive as the logical size calculations that we're fixing here.
While we're at it, fix disk-usage-based eviction in a similar way.
It wasn't the culprit here, according to prod logs, but it can theoretically be
a little CPU-intensive.
More context, including graphs from Prod:
https://neondb.slack.com/archives/C03F5SM1N02/p1687541681336949
(cherry picked from commit d6e35222ea)
This commit introduces an SQL-over-HTTP endpoint in the proxy, with a JSON
response structure resembling that of the node-postgres driver. This method,
using HTTP POST, achieves smaller amortized latencies in edge setups due to
fewer round trips and an enhanced open connection reuse by the v8 engine.
This update involves several intricacies:
1. SQL injection protection: We employed the extended query protocol, modifying
the rust-postgres driver to send queries in one roundtrip using a text
protocol rather than binary, bypassing potential issues like those identified
in https://github.com/sfackler/rust-postgres/issues/1030.
2. Postgres type compatibility: As not all postgres types have binary
representations (e.g., acl's in pg_class), we adjusted rust-postgres to
respond with text protocol, simplifying serialization and fixing queries with
text-only types in response.
3. Data type conversion: Considering JSON supports fewer data types than
Postgres, we perform conversions where possible, passing all other types as
strings. Key conversions include:
- postgres int2, int4, float4, float8 -> json number (NaN and Inf remain
text)
- postgres bool, null, text -> json bool, null, string
- postgres array -> json array
- postgres json and jsonb -> json object
4. Alignment with node-postgres: To facilitate integration with js libraries,
we've matched the response structure of node-postgres, returning command tags
and column oids. Command tag capturing was added to the rust-postgres
functionality as part of this change.
## Problem
Compatibility tests don't support Postgres 15 yet, but we're still
trying to upload compatibility snapshot (which we do not collect).
Ref
https://github.com/neondatabase/neon/actions/runs/4991394158/jobs/8940369368#step:4:38129
## Summary of changes
Add `pg_version` parameter to `run-python-test-set` actions and do not
upload compatibility snapshot for Postgres 15
This reverts commit 732acc5.
Reverted PR: #3869
As noted in PR #4094, we do in fact try to insert duplicates to the
layer map, if L0->L1 compaction is interrupted. We do not have a proper
fix for that right now, and we are in a hurry to make a release to
production, so revert the changes related to this to the state that we
have in production currently. We know that we have a bug here, but
better to live with the bug that we've had in production for a long
time, than rush a fix to production without testing it in staging first.
Cc: #4094, #4088
Otherwise they get lost. Normally buffer is empty before proxy pass, but this is
not the case with pipeline mode of out npm driver; fixes connection hangup
introduced by b80fe41af3 for it.
fixes https://github.com/neondatabase/neon/issues/3822
## Describe your changes
We have previously changed the neon-proxy to use RollingUpdate. This
should be enabled in legacy proxy too in order to avoid breaking
connections for the clients and allow for example backups to run even
during deployment. (https://github.com/neondatabase/neon/pull/3683)
## Issue ticket number and link
https://github.com/neondatabase/neon/issues/3333
## Describe your changes
Rebase vendored PostgreSQL onto 14.7 and 15.2
## Issue ticket number and link
#3579
## 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.
```
The version of PostgreSQL that we use is updated to 14.7 for PostgreSQL
14 and 15.2 for PostgreSQL 15.
```
previously we applied the ratelimiting only up to receiving the headers
from s3, or somewhere near it. the commit adds an adapter which carries
the permit until the AsyncRead has been disposed.
fixes#3662.
Calculation of logical size is now async because of layer downloads, so
we shouldn't use spawn_blocking for it. Use of `spawn_blocking`
exhausted resources which are needed by `tokio::io::copy` when copying
from a stream to a file which lead to deadlock.
Fixes: #3657
these are happening in tests because of #3655 but they sure took some
time to appear.
makes the `Compaction failed, retrying in 2s: Cannot run compaction
iteration on inactive tenant` into a globally allowed error, because it
has been seen failing on different test cases.
Small changes, but hopefully this will help with the panic detected in
staging, for which we cannot get the debugging information right now
(end-of-branch before branch-point).
Before only the timelines which have passed the `gc_horizon` were
processed which failed with orphans at the tree_sort phase. Example
input in added `test_branched_empty_timeline_size` test case.
The PR changes iteration to happen through all timelines, and in
addition to that, any learned branch points will be calculated as they
would had been in the original implementation if the ancestor branch had
been over the `gc_horizon`.
This also changes how tenants where all timelines are below `gc_horizon`
are handled. Previously tenant_size 0 was returned, but now they will
have approximately `initdb_lsn` worth of tenant_size.
The PR also adds several new tenant size tests that describe various corner
cases of branching structure and `gc_horizon` setting.
They are currently disabled to not consume time during CI.
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
Co-authored-by: Anastasia Lubennikova <anastasia@neon.tech>
Previously, we were trying to re-assign owned objects of the already
deleted role. This were causing a crash loop in the case when compute
was restarted with a spec that includes delta operation for role
deletion. To avoid such cases, check that role is still present before
calling `reassign_owned_objects`.
Resolvesneondatabase/cloud#3553
This reverts commit 826e89b9ce.
The problem with that commit was that it deletes the TempDir while
there are still EphemeralFile instances open.
At first I thought this could be fixed by simply adding
Handle::current().block_on(task_mgr::shutdown(None, Some(tenant_id), None))
to TenantHarness::drop, but it turned out to be insufficient.
So, reverting the commit until we find a proper solution.
refs https://github.com/neondatabase/neon/issues/3385
Refactors Compute::prepare_and_run. It's split into subroutines
differently, to make it easier to attach tracing spans to the
different stages. The high-level logic for waiting for Postgres to
exit is moved to the caller.
Replace 'env_logger' with 'tracing', and add `#instrument` directives
to different stages fo the startup process. This is a fairly
mechanical change, except for the changes in 'spec.rs'. 'spec.rs'
contained some complicated formatting, where parts of log messages
were printed directly to stdout with `print`s. That was a bit messed
up because the log normally goes to stderr, but those lines were
printed to stdout. In our docker images, stderr and stdout both go to
the same place so you wouldn't notice, but I don't think it was
intentional.
This changes the log format to the default
'tracing_subscriber::format' format. It's different from the Postgres
log format, however, and because both compute_tools and Postgres print
to the same log, it's now a mix of two different formats. I'm not
sure how the Grafana log parsing pipeline can handle that. If it's a
problem, we can build custom formatter to change the compute_tools log
format to be the same as Postgres's, like it was before this commit,
or we can change the Postgres log format to match tracing_formatter's,
or we can start printing compute_tool's log output to a different
destination than Postgres
IMDSv2 has limits, and if we query it on every s3 interaction we are
going to go over those limits. Changes the s3_bucket client
configuration to use:
- ChainCredentialsProvider to handle env variables or imds usage
- LazyCachingCredentialsProvider to actually cache any credentials
Related: https://github.com/awslabs/aws-sdk-rust/issues/629
Possibly related: https://github.com/neondatabase/neon/issues/3118
plv8 can only be built with a fairly new gold linker version. We used to install
it via binutils packages from testing, but it also updates libc and that causes
troubles in the resulting image as different extensions were built against
different libc versions. We could either use libc from debian-testing everywhere
or restrain from using testing packages and install necessary programs manually.
This patch uses the latter approach: gold for plv8 and cmake for h3 are
installed manually.
In a passing declare h3_postgis as a safe extension (previous omission).
`GRANT CREATE ON SCHEMA public` fails if there is no schema `public`.
Disable it in release for now and make a better fix later (it is
needed for v15 support).
* Check for entire range during sasl validation (#2281)
* Gen2 GH runner (#2128)
* Re-add rustup override
* Try s3 bucket
* Set git version
* Use v4 cache key to prevent problems
* Switch to v5 for key
* Add second rustup fix
* Rebase
* Add kaniko steps
* Fix typo and set compress level
* Disable global run default
* Specify shell for step
* Change approach with kaniko
* Try less verbose shell spec
* Add submodule pull
* Add promote step
* Adjust dependency chain
* Try default swap again
* Use env
* Don't override aws key
* Make kaniko build conditional
* Specify runs on
* Try without dependency link
* Try soft fail
* Use image with git
* Try passing to next step
* Fix duplicate
* Try other approach
* Try other approach
* Fix typo
* Try other syntax
* Set env
* Adjust setup
* Try step 1
* Add link
* Try global env
* Fix mistake
* Debug
* Try other syntax
* Try other approach
* Change order
* Move output one step down
* Put output up one level
* Try other syntax
* Skip build
* Try output
* Re-enable build
* Try other syntax
* Skip middle step
* Update check
* Try first step of dockerhub push
* Update needs dependency
* Try explicit dir
* Add missing package
* Try other approach
* Try other approach
* Specify region
* Use with
* Try other approach
* Add debug
* Try other approach
* Set region
* Follow AWS example
* Try github approach
* Skip Qemu
* Try stdin
* Missing steps
* Add missing close
* Add echo debug
* Try v2 endpoint
* Use v1 endpoint
* Try without quotes
* Revert
* Try crane
* Add debug
* Split steps
* Fix duplicate
* Add shell step
* Conform to options
* Add verbose flag
* Try single step
* Try workaround
* First request fails hunch
* Try bullseye image
* Try other approach
* Adjust verbose level
* Try previous step
* Add more debug
* Remove debug step
* Remove rogue indent
* Try with larger image
* Add build tag step
* Update workflow for testing
* Add tag step for test
* Remove unused
* Update dependency chain
* Add ownership fix
* Use matrix for promote
* Force update
* Force build
* Remove unused
* Add new image
* Add missing argument
* Update dockerfile copy
* Update Dockerfile
* Update clone
* Update dockerfile
* Go to correct folder
* Use correct format
* Update dockerfile
* Remove cd
* Debug find where we are
* Add debug on first step
* Changedir to postgres
* Set workdir
* Use v1 approach
* Use other dependency
* Try other approach
* Try other approach
* Update dockerfile
* Update approach
* Update dockerfile
* Update approach
* Update dockerfile
* Update dockerfile
* Add workspace hack
* Update Dockerfile
* Update Dockerfile
* Update Dockerfile
* Change last step
* Cleanup pull in prep for review
* Force build images
* Add condition for latest tagging
* Use pinned version
* Try without name value
* Remove more names
* Shorten names
* Add kaniko comments
* Pin kaniko
* Pin crane and ecr helper
* Up one level
* Switch to pinned tag for rust image
* Force update for test
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@b04468bf-cdf4-41eb-9c94-aff4ca55e4bf.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@Rorys-Mac-Studio.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@4795e9ee-4f32-401f-85f3-f316263b62b8.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@2f8bc4e5-4ec2-4ea2-adb1-65d863c4a558.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@27565b2b-72d5-4742-9898-a26c9033e6f9.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@ecc96c26-c6c4-4664-be6e-34f7c3f89a3c.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@7caff3a5-bf03-4202-bd0e-f1a93c86bdae.fritz.box>
* Add missing step output, revert one deploy step (#2285)
* Add missing step output, revert one deploy step
* Conform to syntax
* Update approach
* Add missing value
* Add missing needs
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
* Error for fatal not git repo (#2286)
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
* Use main, not branch for ref check (#2288)
* Use main, not branch for ref check
* Add more debug
* Count main, not head
* Try new approach
* Conform to syntax
* Update approach
* Get full history
* Skip checkout
* Cleanup debug
* Remove more debug
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
* Fix docker zombie process issue (#2289)
* Fix docker zombie process issue
* Init everywhere
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
* Fix 1.63 clippy lints (#2282)
* split out timeline metrics, track layer map loading and size calculation
* reset rust cache for clippy run to avoid an ICE
additionally remove trailing whitespaces
* Rename pg_control_ffi.h to bindgen_deps.h, for clarity.
The pg_control_ffi.h name implies that it only includes stuff related to
pg_control.h. That's mostly true currently, but really the point of the
file is to include everything that we need to generate Rust definitions
from.
* Make local mypy behave like CI mypy (#2291)
* Fix flaky pageserver restarts in tests (#2261)
* Remove extra type aliases (#2280)
* Update cachepot endpoint (#2290)
* Update cachepot endpoint
* Update dockerfile & remove env
* Update image building process
* Cannot use metadata endpoint for this
* Update workflow
* Conform to kaniko syntax
* Update syntax
* Update approach
* Update dockerfiles
* Force update
* Update dockerfiles
* Update dockerfile
* Cleanup dockerfiles
* Update s3 test location
* Revert s3 experiment
* Add more debug
* Specify aws region
* Remove debug, add prefix
* Remove one more debug
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
* workflows/benchmarking: increase timeout (#2294)
* Rework `init` in pageserver CLI (#2272)
* Do not create initial tenant and timeline (adjust Python tests for that)
* Rework config handling during init, add --update-config to manage local config updates
* Fix: Always build images (#2296)
* Always build images
* Remove unused
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
* Move auto-generated 'bindings' to a separate inner module.
Re-export only things that are used by other modules.
In the future, I'm imagining that we run bindgen twice, for Postgres
v14 and v15. The two sets of bindings would go into separate
'bindings_v14' and 'bindings_v15' modules.
Rearrange postgres_ffi modules.
Move function, to avoid Postgres version dependency in timelines.rs
Move function to generate a logical-message WAL record to postgres_ffi.
* fix cargo test
* Fix walreceiver and safekeeper bugs (#2295)
- There was an issue with zero commit_lsn `reason: LaggingWal { current_commit_lsn: 0/0, new_commit_lsn: 1/6FD90D38, threshold: 10485760 } }`. The problem was in `send_wal.rs`, where we initialized `end_pos = Lsn(0)` and in some cases sent it to the pageserver.
- IDENTIFY_SYSTEM previously returned `flush_lsn` as a physical end of WAL. Now it returns `flush_lsn` (as it was) to walproposer and `commit_lsn` to everyone else including pageserver.
- There was an issue with backoff where connection was cancelled right after initialization: `connected!` -> `safekeeper_handle_db: Connection cancelled` -> `Backoff: waiting 3 seconds`. The problem was in sleeping before establishing the connection. This is fixed by reworking retry logic.
- There was an issue with getting `NoKeepAlives` reason in a loop. The issue is probably the same as the previous.
- There was an issue with filtering safekeepers based on retry attempts, which could filter some safekeepers indefinetely. This is fixed by using retry cooldown duration instead of retry attempts.
- Some `send_wal.rs` connections failed with errors without context. This is fixed by adding a timeline to safekeepers errors.
New retry logic works like this:
- Every candidate has a `next_retry_at` timestamp and is not considered for connection until that moment
- When walreceiver connection is closed, we update `next_retry_at` using exponential backoff, increasing the cooldown on every disconnect.
- When `last_record_lsn` was advanced using the WAL from the safekeeper, we reset the retry cooldown and exponential backoff, allowing walreceiver to reconnect to the same safekeeper instantly.
* on safekeeper registration pass availability zone param (#2292)
Co-authored-by: Kirill Bulatov <kirill@neon.tech>
Co-authored-by: Rory de Zoete <33318916+zoete@users.noreply.github.com>
Co-authored-by: Rory de Zoete <rdezoete@RorysMacStudio.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@b04468bf-cdf4-41eb-9c94-aff4ca55e4bf.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@Rorys-Mac-Studio.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@4795e9ee-4f32-401f-85f3-f316263b62b8.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@2f8bc4e5-4ec2-4ea2-adb1-65d863c4a558.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@27565b2b-72d5-4742-9898-a26c9033e6f9.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@ecc96c26-c6c4-4664-be6e-34f7c3f89a3c.fritz.box>
Co-authored-by: Rory de Zoete <rdezoete@7caff3a5-bf03-4202-bd0e-f1a93c86bdae.fritz.box>
Co-authored-by: Dmitry Rodionov <dmitry@neon.tech>
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
Co-authored-by: bojanserafimov <bojan.serafimov7@gmail.com>
Co-authored-by: Alexander Bayandin <alexander@neon.tech>
Co-authored-by: Anastasia Lubennikova <anastasia@neon.tech>
Co-authored-by: Anton Galitsyn <agalitsyn@users.noreply.github.com>
* github/workflows: Fix git dubious ownership (#2223)
* Move relation size cache from WalIngest to DatadirTimeline (#2094)
* Move relation sie cache to layered timeline
* Fix obtaining current LSN for relation size cache
* Resolve merge conflicts
* Resolve merge conflicts
* Reestore 'lsn' field in DatadirModification
* adjust DatadirModification lsn in ingest_record
* Fix formatting
* Pass lsn to get_relsize
* Fix merge conflict
* Update pageserver/src/pgdatadir_mapping.rs
Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>
* Update pageserver/src/pgdatadir_mapping.rs
Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>
Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>
* refactor: replace lazy-static with once-cell (#2195)
- Replacing all the occurrences of lazy-static with `once-cell::sync::Lazy`
- fixes#1147
Signed-off-by: Ankur Srivastava <best.ankur@gmail.com>
* Add more buckets to pageserver latency metrics (#2225)
* ignore record property warning to fix benchmarks
* increase statement timeout
* use event so it fires only if workload thread successfully finished
* remove debug log
* increase timeout to pass test with real s3
* avoid duplicate parameter, increase timeout
* Major migration script (#2073)
This script can be used to migrate a tenant across breaking storage versions, or (in the future) upgrading postgres versions. See the comment at the top for an overview.
Co-authored-by: Anastasia Lubennikova <anastasia@neon.tech>
* Fix etcd typos
* Fix links to safekeeper protocol docs. (#2188)
safekeeper/README_PROTO.md was moved to docs/safekeeper-protocol.md in
commit 0b14fdb078, as part of reorganizing the docs into 'mdbook' format.
Fixes issue #1475. Thanks to @banks for spotting the outdated references.
In addition to fixing the above issue, this patch also fixes other broken links as a result of 0b14fdb078. See https://github.com/neondatabase/neon/pull/2188#pullrequestreview-1055918480.
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
Co-authored-by: Thang Pham <thang@neon.tech>
* Update CONTRIBUTING.md
* Update CONTRIBUTING.md
* support node id and remote storage params in docker_entrypoint.sh
* Safe truncate (#2218)
* Move relation sie cache to layered timeline
* Fix obtaining current LSN for relation size cache
* Resolve merge conflicts
* Resolve merge conflicts
* Reestore 'lsn' field in DatadirModification
* adjust DatadirModification lsn in ingest_record
* Fix formatting
* Pass lsn to get_relsize
* Fix merge conflict
* Update pageserver/src/pgdatadir_mapping.rs
Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>
* Update pageserver/src/pgdatadir_mapping.rs
Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>
* Check if relation exists before trying to truncat it
refer #1932
* Add test reporducing FSM truncate problem
Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>
* Fix exponential backoff values
* Update back `vendor/postgres` back; it was changed accidentally. (#2251)
Commit 4227cfc96e accidentally reverted vendor/postgres to an older
version. Update it back.
* Add pageserver checkpoint_timeout option.
To flush inmemory layer eventually when no new data arrives, which helps
safekeepers to suspend activity (stop pushing to the broker). Default 10m should
be ok.
* Share exponential backoff code and fix logic for delete task failure (#2252)
* Fix bug when import large (>1GB) relations (#2172)
Resolves#2097
- use timeline modification's `lsn` and timeline's `last_record_lsn` to determine the corresponding LSN to query data in `DatadirModification::get`
- update `test_import_from_pageserver`. Split the test into 2 variants: `small` and `multisegment`.
+ `small` is the old test
+ `multisegment` is to simulate #2097 by using a larger number of inserted rows to create multiple segment files of a relation. `multisegment` is configured to only run with a `release` build
* Fix timeline physical size flaky tests (#2244)
Resolves#2212.
- use `wait_for_last_flush_lsn` in `test_timeline_physical_size_*` tests
## Context
Need to wait for the pageserver to catch up with the compute's last flush LSN because during the timeline physical size API call, it's possible that there are running `LayerFlushThread` threads. These threads flush new layers into disk and hence update the physical size. This results in a mismatch between the physical size reported by the API and the actual physical size on disk.
### Note
The `LayerFlushThread` threads are processed **concurrently**, so it's possible that the above error still persists even with this patch. However, making the tests wait to finish processing all the WALs (not flushing) before calculating the physical size should help reduce the "flakiness" significantly
* postgres_ffi/waldecoder: validate more header fields
* postgres_ffi/waldecoder: remove unused startlsn
* postgres_ffi/waldecoder: introduce explicit `enum State`
Previously it was emulated with a combination of nullable fields.
This change should make the logic more readable.
* disable `test_import_from_pageserver_multisegment` (#2258)
This test failed consistently on `main` now. It's better to temporarily disable it to avoid blocking others' PRs while investigating the root cause for the test failure.
See: #2255, #2256
* get_binaries uses DOCKER_TAG taken from docker image build step (#2260)
* [proxy] Rework wire format of the password hack and some errors (#2236)
The new format has a few benefits: it's shorter, simpler and
human-readable as well. We don't use base64 anymore, since
url encoding got us covered.
We also show a better error in case we couldn't parse the
payload; the users should know it's all about passing the
correct project name.
* test_runner/pg_clients: collect docker logs (#2259)
* get_binaries script fix (#2263)
* get_binaries uses DOCKER_TAG taken from docker image build step
* remove docker tag discovery at all and fix get_binaries for version variable
* Better storage sync logs (#2268)
* Find end of WAL on safekeepers using WalStreamDecoder.
We could make it inside wal_storage.rs, but taking into account that
- wal_storage.rs reading is async
- we don't need s3 here
- error handling is different; error during decoding is normal
I decided to put it separately.
Test
cargo test test_find_end_of_wal_last_crossing_segment
prepared earlier by @yeputons passes now.
Fixes https://github.com/neondatabase/neon/issues/544https://github.com/neondatabase/cloud/issues/2004
Supersedes https://github.com/neondatabase/neon/pull/2066
* Improve walreceiver logic (#2253)
This patch makes walreceiver logic more complicated, but it should work better in most cases. Added `test_wal_lagging` to test scenarios where alive safekeepers can lag behind other alive safekeepers.
- There was a bug which looks like `etcd_info.timeline.commit_lsn > Some(self.local_timeline.get_last_record_lsn())` filtered all safekeepers in some strange cases. I removed this filter, it should probably help with #2237
- Now walreceiver_connection reports status, including commit_lsn. This allows keeping safekeeper connection even when etcd is down.
- Safekeeper connection now fails if pageserver doesn't receive safekeeper messages for some time. Usually safekeeper sends messages at least once per second.
- `LaggingWal` check now uses `commit_lsn` directly from safekeeper. This fixes the issue with often reconnects, when compute generates WAL really fast.
- `NoWalTimeout` is rewritten to trigger only when we know about the new WAL and the connected safekeeper doesn't stream any WAL. This allows setting a small `lagging_wal_timeout` because it will trigger only when we observe that the connected safekeeper has stuck.
* increase timeout in wait_for_upload to avoid spurious failures when testing with real s3
* Bump vendor/postgres to include XLP_FIRST_IS_CONTRECORD fix. (#2274)
* Set up a workflow to run pgbench against captest (#2077)
Signed-off-by: Ankur Srivastava <best.ankur@gmail.com>
Co-authored-by: Alexander Bayandin <alexander@neon.tech>
Co-authored-by: Konstantin Knizhnik <knizhnik@garret.ru>
Co-authored-by: Heikki Linnakangas <heikki@zenith.tech>
Co-authored-by: Ankur Srivastava <ansrivas@users.noreply.github.com>
Co-authored-by: bojanserafimov <bojan.serafimov7@gmail.com>
Co-authored-by: Dmitry Rodionov <dmitry@neon.tech>
Co-authored-by: Anastasia Lubennikova <anastasia@neon.tech>
Co-authored-by: Kirill Bulatov <kirill@neon.tech>
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
Co-authored-by: Thang Pham <thang@neon.tech>
Co-authored-by: Stas Kelvich <stas.kelvich@gmail.com>
Co-authored-by: Arseny Sher <sher-ars@yandex.ru>
Co-authored-by: Egor Suvorov <egor@neon.tech>
Co-authored-by: Andrey Taranik <andrey@cicd.team>
Co-authored-by: Dmitry Ivanov <ivadmi5@gmail.com>
[HOTFIX] Release deploy fix
This PR uses this branch neondatabase/postgres#171 and several required commits from the main to use only locally built compute-tools. This should allow us to rollout safekeepers sync issue fix on prod
# Keep the version the same as in compute/compute-node.Dockerfile
ENVPROTOC_VERSION=25.1
RUN curl -fsSL "https://github.com/protocolbuffers/protobuf/releases/download/v${PROTOC_VERSION}/protoc-${PROTOC_VERSION}-linux-$(uname -m | sed 's/aarch64/aarch_64/g').zip" -o "protoc.zip"\
&& unzip -q protoc.zip -d protoc \
@@ -179,7 +180,7 @@ RUN curl -sL "https://github.com/peak/s5cmd/releases/download/v${S5CMD_VERSION}/
&& mv s5cmd /usr/local/bin/s5cmd
# LLVM
ENVLLVM_VERSION=19
ENVLLVM_VERSION=20
RUN curl -fsSL 'https://apt.llvm.org/llvm-snapshot.gpg.key'| apt-key add - \
In this RFC, we will describe how we will implement per-tenant feature flags.
## PostHog as Feature Flag Service
Before we start, let's talk about how current feature flag services work. PostHog is the feature flag service we are currently using across multiple user-facing components in the company. PostHog has two modes of operation: HTTP evaluation and server-side local evaluation.
Let's assume we have a storage feature flag called gc-compaction and we want to roll it out to scale-tier users with resident size >= 10GB and <= 100GB.
### Define User Profiles
The first step is to synchronize our user profiles to the PostHog service. We can simply assume that each tenant is a user in PostHog. Each user profile has some properties associated with it. In our case, it will be: plan type (free, scale, enterprise, etc); resident size (in bytes); primary pageserver (string); region (string).
### Define Feature Flags
We would create a feature flag called gc-compaction in PostHog with 4 variants: disabled, stage-1, stage-2, fully-enabled. We will flip the feature flags from disabled to fully-enabled stage by stage for some percentage of our users.
### Option 1: HTTP Evaluation Mode
When using PostHog's HTTP evaluation mode, the client will make request to the PostHog service, asking for the value of a feature flag for a specific user.
* Control plane will report the plan type to PostHog each time it attaches a tenant to the storcon or when the user upgrades/downgrades. It calls the PostHog profile API to associate tenant ID with the plan type. Assume we have X active tenants and such attach or plan change event happens each week, that would be 4X profile update requests per month.
* Pageservers will report the resident size and the primary pageserver to the PostHog service. Assume we report resident size every 24 hours, that would be 30X requests per month.
* Each tenant will request the state of the feature flag every 1 hour, that's 720X requests per month.
* The Rust client would be easy to implement as we only need to call the `/decide` API on PostHog.
Using the HTTP evaluation mode we will issue 754X requests a month.
### Option 2: Local Evaluation Mode
When using PostHog's HTTP evaluation mode, the client (usually the server in a browser/server architecture) will poll the feature flag configuration every 30s (default in the Python client) from PostHog. Such configuration contains data like:
<details>
<summary>Example JSON response from the PostHog local evaluation API</summary>
```
[
{
"id": 1,
"name": "Beta Feature",
"key": "person-flag",
"is_simple_flag": True,
"active": True,
"filters": {
"groups": [
{
"properties": [
{
"key": "location",
"operator": "exact",
"value": ["Straße"],
"type": "person",
}
],
"rollout_percentage": 100,
},
{
"properties": [
{
"key": "star",
"operator": "exact",
"value": ["ſun"],
"type": "person",
}
],
"rollout_percentage": 100,
},
],
},
}
]
```
</details>
Note that the API only contains information like "under what condition => rollout percentage". The user is responsible to provide the properties required to the client for local evaluation, and the PostHog service (web UI) cannot know if a feature is enabled for the tenant or not until the client uses the `capture` API to report the result back. To control the rollout percentage, the user ID gets mapped to a float number in `[0, 1)` on a consistent hash ring. All values <= the percentage will get the feature enabled or set to the desired value.
To use the local evaluation mode, the system needs:
* Assume each pageserver will poll PostHog for the local evaluation JSON every 5 minutes (instead of the 30s default as it's too frequent). That's 8640Y per month, Y is the number of pageservers. Local evaluation requests cost 10x more than the normal decide request, so that's 86400Y request units to bill.
* Storcon needs to store the plan type in the database and pass that information to the pageserver when attaching the tenant.
* Storcon also needs to update PostHog with the active tenants, for example, when the tenant gets detached/attached. Assume each active tenant gets detached/attached every week, that would be 4X requests per month.
* We do not need to update bill type or resident size to PostHog as all these are evaluated locally.
* After each local evaluation of the feature flag, we need to call PostHog's capture event API to update the result of the evaluation that the feature is enabled. We can do this when the flag gets changed compared with the last cached state in memory. That would be at least 4X (assume we do deployment every week so the cache gets cleared) and maybe an additional multiplifier of 10 assume we have 10 active features.
In this case, we will issue 86400Y + 40X requests per month.
Assume X = 1,000,000 and Y = 100,
| | HTTP Evaluation | Local Evaluation |
|---|---|---|
| Latency of propagating the conditions/properties for feature flag | 24 hours | available locally |
| Latency of applying the feature flag | 1 hour | 5 minutes |
| Can properties be reported from different services | Yes | No |
| Do we need to sync billing info etc to pageserver | No | Yes |
| Cost | 75400$ / month | 4864$ / month |
# Our Solution
We will use PostHog _only_ as an UI to configure the feature flags. Whether a feature is enabled or not can only be queried through storcon/pageserver instead of using the PostHog UI. (We could report it back to PostHog via `capture_event` but it costs $$$.) This allows us to ramp up the feature flag functionality fast at first. At the same time, it would also give us the option to migrate to our own solution once we want to have more properties and more complex evaluation rules in our system.
* We will create several fake users (tenants) in PostHog that contains all the properties we will use for evaluating a feature flag (i.e., resident size, billing type, pageserver id, etc.)
* We will use PostHog's local evaluation API to poll the configuration of the feature flags and evaluate them locally on each of the pageserver.
* The evaluation result will not be reported back to PostHog.
* Storcon needs to pull some information from cplane database.
* To know if a feature is currently enabled or not, we need to call the storcon/pageserver API; and we won't be able to know if a feature has been enabled on a tenant before easily: we need to look at the Grafana logs.
We only need to pay for the 86400Y local evaluation requests (that would be setting Y=0 in solution 2 => $864/month, and even less if we proxy it through storcon).
## Implementation
* Pageserver: implement a PostHog local evaluation client. The client will be shared across all tenants on the pageserver with a single API: `evaluate(tenant_id, feature_flag, properties) -> json`.
* Storcon: if we need plan type as the evaluation condition, pull it from cplane database.
* Storcon/Pageserver: implement an HTTP API `:tenant_id/feature/:feature` to retrieve the current feature flag status.
* Storcon/Pageserver: a loop to update the feature flag spec on both storcon and pageserver. Pageserver loop will only be activated if storcon does not push the specs to the pageserver.
## Difference from Tenant Config
* Feature flags can be modified by percentage, and the default config for each feature flag can be modified in UI without going through the release process.
* Feature flags are more flexible and won't be persisted anywhere and will be passed as plain JSON over the wire so that do not need to handle backward/forward compatibility as in tenant config.
* The expectation of tenant config is that once we add a flag we cannot remove it (or it will be hard to remove), but feature flags are more flexible.
# Final Implementation
* We added a new crate `posthog_lite_client` that supports local feature evaluations.
* We set up two projects "Storage (staging)" and "Storage (production)" in the PostHog console.
* Each pageserver reports 10 fake tenants to PostHog so that we can get all combinations of regions (and other properties) in the PostHog UI.
* Supported properties: AZ, neon_region, pageserver, tenant_id.
* You may use "Pageserver Feature Flags" dashboard to see the evaluation status.
* The feature flag spec is polled on storcon every 30s (in each of the region) and storcon will propagate the spec to the pageservers.
* The pageserver housekeeping loop updates the tenant-specific properties (e.g., remote size) for evaluation.
Each tenant has a `feature_resolver` object. After you add a feature flag in the PostHog console, you can retrieve it with:
```rust
// Boolean flag
self
.feature_resolver
.evaluate_boolean("flag")
.is_ok()
// Multivariate flag
self
.feature_resolver
.evaluate_multivariate("gc-comapction-strategy")
.ok();
```
The user needs to handle the case where the evaluation result is an error. This can occur in a variety of cases:
* During the pageserver start, the feature flag spec has not been retrieved.
* No condition group is matched.
* The feature flag spec contains an operand/operation not supported by the lite PostHog library.
For boolean flags, the return value is `Result<(), Error>`. `Ok(())` means the flag is evaluated to true. Otherwise,
there is either an error in evaluation or it does not match any groups.
For multivariate flags, the return value is `Result<String, Error>`. `Ok(variant)` indicates the flag is evaluated
to a variant. Otherwise, there is either an error in evaluation or it does not match any groups.
The evaluation logic is documented in the PostHog lite library. It compares the consistent hash of a flag key + tenant_id
with the rollout percentage and determines which tenant to roll out a specific feature.
Users can use the feature flag evaluation API to get the flag evaluation result of a specific tenant for debugging purposes.
This RFC describes an approach to reduce performance degradation due to missing caches after compute node restart, i.e.:
1. Rolling restart of the running instance via 'warm' replica.
2. Auto-prewarm compute caches after unplanned restart or scale-to-zero.
## Motivation
Neon currently implements several features that guarantee high uptime of compute nodes:
1. Storage high-availability (HA), i.e. each tenant shard has a secondary pageserver location, so we can quickly switch over compute to it in case of primary pageserver failure.
2. Fast compute provisioning, i.e. we have a fleet of pre-created empty computes, that are ready to serve workload, so restarting unresponsive compute is very fast.
3. Preemptive NeonVM compute provisioning in case of k8s node unavailability.
This helps us to be well-within the uptime SLO of 99.95% most of the time. Problems begin when we go up to multi-TB workloads and 32-64 CU computes.
During restart, compute loses all caches: LFC, shared buffers, file system cache. Depending on the workload, it can take a lot of time to warm up the caches,
so that performance could be degraded and might be even unacceptable for certain workloads. The latter means that although current approach works well for small to
medium workloads, we still have to do some additional work to avoid performance degradation after restart of large instances.
## Non Goals
- Details of the persistence storage for prewarm data are out of scope, there is a separate RFC for that: <https://github.com/neondatabase/neon/pull/9661>.
- Complete compute/Postgres HA setup and flow. Although it was originally in scope of this RFC, during preliminary research it appeared to be a rabbit hole, so it's worth of a separate RFC.
- Low-level implementation details for Postgres replica-to-primary promotion. There are a lot of things to think and care about: how to start walproposer, [logical replication failover](https://www.postgresql.org/docs/current/logical-replication-failover.html), and so on, but it's worth of at least a separate one-pager design document if not RFC.
## Impacted components
Postgres, compute_ctl, Control plane, Endpoint storage for unlogged storage of compute files.
For the latter, we will need to implement a uniform abstraction layer on top of S3, ABS, etc., but
S3 is used in text interchangeably with 'endpoint storage' for simplicity.
## Proposed implementation
### compute_ctl spec changes and auto-prewarm
We are going to extend the current compute spec with the following attributes
```rust
structComputeSpec{
/// [All existing attributes]
...
/// Whether to do auto-prewarm at start or not.
/// Default to `false`.
publfc_auto_prewarm: bool
/// Interval in seconds between automatic dumps of
/// LFC state into S3. Default `None`, which means 'off'.
publfc_dump_interval_sec: Option<i32>
}
```
When `lfc_dump_interval_sec` is set to `N`, `compute_ctl` will periodically dump the LFC state
and store it in S3, so that it could be used either for auto-prewarm after restart or by replica
during the rolling restart. For enabling periodic dumping, we should consider the following value
`lfc_dump_interval_sec=300` (5 minutes), same as in the upstream's `pg_prewarm.autoprewarm_interval`.
When `lfc_auto_prewarm` is set to `true`, `compute_ctl` will start prewarming the LFC upon restart
iif some of the previous states is present in S3.
### compute_ctl API
1.`POST /store_lfc_state` -- dump LFC state using Postgres SQL interface and store result in S3.
This has to be a blocking call, i.e. it will return only after the state is stored in S3.
If there is any concurrent request in progress, we should return `429 Too Many Requests`,
and let the caller to retry.
2.`GET /dump_lfc_state` -- dump LFC state using Postgres SQL interface and return it as is
in text format suitable for the future restore/prewarm. This API is not strictly needed at
the end state, but could be useful for a faster prototyping of a complete rolling restart flow
with prewarm, as it doesn't require persistent for LFC state storage.
3.`POST /restore_lfc_state` -- restore/prewarm LFC state with request
```yaml
RestoreLFCStateRequest:
oneOf:
- type: object
required:
- lfc_state
properties:
lfc_state:
type: string
description: Raw LFC content dumped with GET `/dump_lfc_state`
- type: object
required:
- lfc_cache_key
properties:
lfc_cache_key:
type: string
description: |
endpoint_id of the source endpoint on the same branch
to use as a 'donor' for LFC content. Compute will look up
LFC content dump in S3 using this key and do prewarm.
```
where `lfc_state` and `lfc_cache_key` are mutually exclusive.
The actual prewarming will happen asynchronously, so the caller need to check the
prewarm status using the compute's standard `GET /status` API.
4. `GET /status` -- extend existing API with following attributes
```rust
struct ComputeStatusResponse {
// [All existing attributes]
...
pub prewarm_state: PrewarmState
}
/// Compute prewarm state. Will be stored in the shared Compute state
/// in compute_ctl
struct PrewarmState {
pub status: PrewarmStatus
/// Total number of pages to prewarm
pub pages_total: i64
/// Number of pages prewarmed so far
pub pages_processed: i64
/// Optional prewarm error
pub error: Option<String>
}
pub enum PrewarmStatus {
/// Prewarming was never requested on this compute
Off,
/// Prewarming was requested, but not started yet
Pending,
/// Prewarming is in progress. The caller should follow
/// `PrewarmState::progress`.
InProgress,
/// Prewarming has been successfully completed
Completed,
/// Prewarming failed. The caller should look at
/// `PrewarmState::error` for the reason.
Failed,
/// It is intended to be used by auto-prewarm if none of
/// the previous LFC states is available in S3.
/// This is a distinct state from the `Failed` because
/// technically it's not a failure and could happen if
/// compute was restart before it dumped anything into S3,
/// or just after the initial rollout of the feature.
Skipped,
}
```
5. `POST /promote` -- this is a **blocking** API call to promote compute replica into primary.
This API should be very similar to the existing `POST /configure` API, i.e. accept the
spec (primary spec, because originally compute was started as replica). It's a distinct
API method because semantics and response codes are different:
- If promotion is done successfully, it will return `200 OK`.
- If compute is already primary, the call will be no-op and `compute_ctl`
will return `412 Precondition Failed`.
- If, for some reason, second request reaches compute that is in progress of promotion,
it will respond with `429 Too Many Requests`.
- If compute hit any permanent failure during promotion `500 Internal Server Error`
will be returned.
### Control plane operations
The complete flow will be present as a sequence diagram in the next section, but here
we just want to list some important steps that have to be done by control plane during
the rolling restart via warm replica, but without much of low-level implementation details.
1. Register the 'intent' of the instance restart, but not yet interrupt any workload at
primary and also accept new connections. This may require some endpoint state machine
changes, e.g. introduction of the `pending_restart` state. Being in this state also
**mustn't prevent any other operations except restart**: suspend, live-reconfiguration
(e.g. due to notify-attach call from the storage controller), deletion.
2. Start new replica compute on the same timeline and start prewarming it. This process
may take quite a while, so the same concurrency considerations as in 1. should be applied
here as well.
3. When warm replica is ready, control plane should:
3.1. Terminate the primary compute. Starting from here, **this is a critical section**,
if anything goes off, the only option is to start the primary normally and proceed
with auto-prewarm.
3.2. Send cache invalidation message to all proxies, notifying them that all new connections
should request and wait for the new connection details. At this stage, proxy has to also
drop any existing connections to the old primary, so they didn't do stale reads.
3.3. Attach warm replica compute to the primary endpoint inside control plane metadata
database.
3.4. Promote replica to primary.
3.5. When everything is done, finalize the endpoint state to be just `active`.
### Complete rolling restart flow
```mermaid
sequenceDiagram
autonumber
participant proxy as Neon proxy
participant cplane as Control plane
participant primary as Compute (primary)
box Compute (replica)
participant ctl as compute_ctl
participant pg as Postgres
end
box Endpoint unlogged storage
participant s3proxy as Endpoint storage service
participant s3 as S3/ABS/etc.
end
cplane ->> primary: POST /store_lfc_state
primary -->> cplane: 200 OK
cplane ->> ctl: POST /restore_lfc_state
activate ctl
ctl -->> cplane: 202 Accepted
activate cplane
cplane ->> ctl: GET /status: poll prewarm status
ctl ->> s3proxy: GET /read_file
s3proxy ->> s3: read file
s3 -->> s3proxy: file content
s3proxy -->> ctl: 200 OK: file content
proxy ->> cplane: GET /proxy_wake_compute
cplane -->> proxy: 200 OK: old primary conninfo
ctl ->> pg: prewarm LFC
activate pg
pg -->> ctl: prewarm is completed
deactivate pg
ctl -->> cplane: 200 OK: prewarm is completed
deactivate ctl
deactivate cplane
cplane -->> cplane: reassign replica compute to endpoint,<br>start terminating the old primary compute
activate cplane
cplane ->> proxy: invalidate caches
proxy ->> cplane: GET /proxy_wake_compute
cplane -x primary: POST /terminate
primary -->> cplane: 200 OK
note over primary: old primary<br>compute terminated
cplane ->> ctl: POST /promote
activate ctl
ctl ->> pg: pg_ctl promote
activate pg
pg -->> ctl: done
deactivate pg
ctl -->> cplane: 200 OK
deactivate ctl
cplane -->> cplane: finalize operation
cplane -->> proxy: 200 OK: new primary conninfo
deactivate cplane
```
### Network bandwidth and prewarm speed
It's currently known that pageserver can sustain about 3000 RPS per shard for a few running computes.
Large tenants are usually split into 8 shards, so the final formula may look like this:
```text
8 shards * 3000 RPS * 8 KB =~ 190 MB/s
```
so depending on the LFC size, prewarming will take at least:
- ~5s for 1 GB
- ~50s for 10 GB
- ~5m for 100 GB
- \>1h for 1 TB
In total, one pageserver is normally capped by 30k RPS, so it obviously can't sustain many computes
doing prewarm at the same time. Later, we may need an additional mechanism for computes to throttle
the prewarming requests gracefully.
### Reliability, failure modes and corner cases
We consider following failures while implementing this RFC:
1. Compute got interrupted/crashed/restarted during prewarm. The caller -- control plane -- should
detect that and start prewarm from the beginning.
2. Control plane promotion request timed out or hit network issues. If it never reached the
compute, control plane should just repeat it. If it did reach the compute, then during
retry control plane can hit `409` as previous request triggered the promotion already.
In this case, control plane need to retry until either `200` or
permanent error `500` is returned.
3. Compute got interrupted/crashed/restarted during promotion. At restart it will ask for
a spec from control plane, and its content should signal compute to start as **primary**,
so it's expected that control plane will continue polling for certain period of time and
will discover that compute is ready to accept connections if restart is fast enough.
4. Any other unexpected failure or timeout during prewarming. This **failure mustn't be fatal**,
control plane has to report failure, terminate replica and keep primary running.
5. Any other unexpected failure or timeout during promotion. Unfortunately, at this moment
we already have the primary node stopped, so the only option is to start primary again
and proceed with auto-prewarm.
6. Any unexpected failure during auto-prewarm. This **failure mustn't be fatal**,
`compute_ctl` has to report the failure, but do not crash the compute.
7. Control plane failed to confirm that old primary has terminated. This can happen, especially
in the future HA setup. In this case, control plane has to ensure that it sent VM deletion
and pod termination requests to k8s, so long-term we do not have two running primaries
on the same timeline.
### Security implications
There are two security implications to consider:
1. Access to `compute_ctl` API. It has to be accessible from the outside of compute, so all
new API methods have to be exposed on the **external** HTTP port and **must** be authenticated
with JWT.
2. Read/write only your own LFC state data in S3. Although it's not really a security concern,
since LFC state is just a mapping of blocks present in LFC at certain moment in time;
it still has to be highly restricted, so that i) only computes on the same timeline can
read S3 state; ii) each compute can only write to the path that contains it's `endpoint_id`.
Both of this must be validated by Endpoint storage service using the JWT token provided by `compute_ctl`.
### Unresolved questions
#### Billing, metrics and monitoring
Currently, we only label computes with `endpoint_id` after attaching them to the endpoint.
In this proposal, this means that temporary replica will remain unlabelled until it's promoted
to primary. We can also hide it from users in the control plane API, but what to do with
billing and monitoring is still unclear.
We can probably mark it as 'billable' and tag with `project_id`, so it will be billed, but
not interfere in any way with the current primary monitoring.
Another thing to consider is how logs and metrics export will switch to the new compute.
It's expected that OpenTelemetry collector will auto-discover the new compute and start
scraping metrics from it.
#### Auto-prewarm
It's still an open question whether we need auto-prewarm at all. The author's gut-feeling is
that yes, we need it, but might be not for all workloads, so it could end up exposed as a
user-controllable knob on the endpoint. There are two arguments for that:
1. Auto-prewarm existing in upstream's `pg_prewarm`, _probably for a reason_.
2. There are still could be 2 flows when we cannot perform the rolling restart via the warm
replica: i) any failure or interruption during promotion; ii) wake up after scale-to-zero.
The latter might be challenged as well, i.e. one can argue that auto-prewarm may and will
compete with user-workload for storage resources. This is correct, but it might as well
reduce the time to get warm LFC and good performance.
#### Low-level details of the replica promotion
There are many things to consider here, but three items just off the top of my head:
1. How to properly start the `walproposer` inside Postgres.
2. What to do with logical replication. Currently, we do not include logical replication slots
inside basebackup, because nobody advances them at replica, so they just prevent the WAL
deletion. Yet, we do need to have them at primary after promotion. Starting with Postgres 17,
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.
