## Problem
We wish to have high level WAL decoding logic in `wal_decoder::decoder`
module.
## Summary of Changes
For this we need the `Value` and `NeonWalRecord` types accessible there, so:
1. Move `Value` and `NeonWalRecord` to `pageserver::value` and
`pageserver::record` respectively.
2. Get rid of `pageserver::repository` (follow up from (1))
3. Move PG specific WAL record types to `postgres_ffi::walrecord`. In
theory they could live in `wal_decoder`, but it would create a circular
dependency between `wal_decoder` and `postgres_ffi`. Long term it makes
sense for those types to be PG version specific, so that will work out nicely.
4. Move higher level WAL record types (to be ingested by pageserver)
into `wal_decoder::models`
Related: https://github.com/neondatabase/neon/issues/9335
Epic: https://github.com/neondatabase/neon/issues/9329
Follow-up of #9234 to give hyper 1.0 the version-free name, and the
legacy version of hyper the one with the version number inside. As we
move away from hyper 0.14, we can remove the `hyper0` name piece by
piece.
Part of #9255
## Problem
- In https://github.com/neondatabase/neon/pull/8784, the validate
controller API is modified to check generations directly in the
database. It batches tenants into separate queries to avoid generating a
huge statement, but
- While updating this, I realized that "control_plane_client" is a kind
of confusing name for the client code now that it primarily talks to the
storage controller (the case of talking to the control plane will go
away in a few months).
## Summary of changes
- Big rename to "ControllerUpcallClient" -- this reflects the storage
controller's api naming, where the paths used by the pageserver are in
`/upcall/`
- When sending validate requests, break them up into chunks so that we
avoid possible edge cases of generating any HTTP requests that require
database I/O across many thousands of tenants.
This PR mixes a functional change with a refactor, but the commits are
cleanly separated -- only the last commit is a functional change.
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
Part of [Epic: Bypass PageCache for user data
blocks](https://github.com/neondatabase/neon/issues/7386).
# Problem
`InMemoryLayer` still uses the `PageCache` for all data stored in the
`VirtualFile` that underlies the `EphemeralFile`.
# Background
Before this PR, `EphemeralFile` is a fancy and (code-bloated) buffered
writer around a `VirtualFile` that supports `blob_io`.
The `InMemoryLayerInner::index` stores offsets into the `EphemeralFile`.
At those offset, we find a varint length followed by the serialized
`Value`.
Vectored reads (`get_values_reconstruct_data`) are not in fact vectored
- each `Value` that needs to be read is read sequentially.
The `will_init` bit of information which we use to early-exit the
`get_values_reconstruct_data` for a given key is stored in the
serialized `Value`, meaning we have to read & deserialize the `Value`
from the `EphemeralFile`.
The L0 flushing **also** needs to re-determine the `will_init` bit of
information, by deserializing each value during L0 flush.
# Changes
1. Store the value length and `will_init` information in the
`InMemoryLayer::index`. The `EphemeralFile` thus only needs to store the
values.
2. For `get_values_reconstruct_data`:
- Use the in-memory `index` figures out which values need to be read.
Having the `will_init` stored in the index enables us to do that.
- View the EphemeralFile as a byte array of "DIO chunks", each 512 bytes
in size (adjustable constant). A "DIO chunk" is the minimal unit that we
can read under direct IO.
- Figure out which chunks need to be read to retrieve the serialized
bytes for thes values we need to read.
- Coalesce chunk reads such that each DIO chunk is only read once to
serve all value reads that need data from that chunk.
- Merge adjacent chunk reads into larger
`EphemeralFile::read_exact_at_eof_ok` of up to 128k (adjustable
constant).
3. The new `EphemeralFile::read_exact_at_eof_ok` fills the IO buffer
from the underlying VirtualFile and/or its in-memory buffer.
4. The L0 flush code is changed to use the `index` directly, `blob_io`
5. We can remove the `ephemeral_file::page_caching` construct now.
The `get_values_reconstruct_data` changes seem like a bit overkill but
they are necessary so we issue the equivalent amount of read system
calls compared to before this PR where it was highly likely that even if
the first PageCache access was a miss, remaining reads within the same
`get_values_reconstruct_data` call from the same `EphemeralFile` page
were a hit.
The "DIO chunk" stuff is truly unnecessary for page cache bypass, but,
since we're working on [direct
IO](https://github.com/neondatabase/neon/issues/8130) and
https://github.com/neondatabase/neon/issues/8719 specifically, we need
to do _something_ like this anyways in the near future.
# Alternative Design
The original plan was to use the `vectored_blob_io` code it relies on
the invariant of Delta&Image layers that `index order == values order`.
Further, `vectored_blob_io` code's strategy for merging IOs is limited
to adjacent reads. However, with direct IO, there is another level of
merging that should be done, specifically, if multiple reads map to the
same "DIO chunk" (=alignment-requirement-sized and -aligned region of
the file), then it's "free" to read the chunk into an IO buffer and
serve the two reads from that buffer.
=> https://github.com/neondatabase/neon/issues/8719
# Testing / Performance
Correctness of the IO merging code is ensured by unit tests.
Additionally, minimal tests are added for the `EphemeralFile`
implementation and the bit-packed `InMemoryLayerIndexValue`.
Performance testing results are presented below.
All pref testing done on my M2 MacBook Pro, running a Linux VM.
It's a release build without `--features testing`.
We see definitive improvement in ingest performance microbenchmark and
an ad-hoc microbenchmark for getpage against InMemoryLayer.
```
baseline: commit 7c74112b2a origin/main
HEAD: ef1c55c52e
```
<details>
```
cargo bench --bench bench_ingest -- 'ingest 128MB/100b seq, no delta'
baseline
ingest-small-values/ingest 128MB/100b seq, no delta
time: [483.50 ms 498.73 ms 522.53 ms]
thrpt: [244.96 MiB/s 256.65 MiB/s 264.73 MiB/s]
HEAD
ingest-small-values/ingest 128MB/100b seq, no delta
time: [479.22 ms 482.92 ms 487.35 ms]
thrpt: [262.64 MiB/s 265.06 MiB/s 267.10 MiB/s]
```
</details>
We don't have a micro-benchmark for InMemoryLayer and it's quite
cumbersome to add one. So, I did manual testing in `neon_local`.
<details>
```
./target/release/neon_local stop
rm -rf .neon
./target/release/neon_local init
./target/release/neon_local start
./target/release/neon_local tenant create --set-default
./target/release/neon_local endpoint create foo
./target/release/neon_local endpoint start foo
psql 'postgresql://cloud_admin@127.0.0.1:55432/postgres'
psql (13.16 (Debian 13.16-0+deb11u1), server 15.7)
CREATE TABLE wal_test (
id SERIAL PRIMARY KEY,
data TEXT
);
DO $$
DECLARE
i INTEGER := 1;
BEGIN
WHILE i <= 500000 LOOP
INSERT INTO wal_test (data) VALUES ('data');
i := i + 1;
END LOOP;
END $$;
-- => result is one L0 from initdb and one 137M-sized ephemeral-2
DO $$
DECLARE
i INTEGER := 1;
random_id INTEGER;
random_record wal_test%ROWTYPE;
start_time TIMESTAMP := clock_timestamp();
selects_completed INTEGER := 0;
min_id INTEGER := 1; -- Minimum ID value
max_id INTEGER := 100000; -- Maximum ID value, based on your insert range
iters INTEGER := 100000000; -- Number of iterations to run
BEGIN
WHILE i <= iters LOOP
-- Generate a random ID within the known range
random_id := min_id + floor(random() * (max_id - min_id + 1))::int;
-- Select the row with the generated random ID
SELECT * INTO random_record
FROM wal_test
WHERE id = random_id;
-- Increment the select counter
selects_completed := selects_completed + 1;
-- Check if a second has passed
IF EXTRACT(EPOCH FROM clock_timestamp() - start_time) >= 1 THEN
-- Print the number of selects completed in the last second
RAISE NOTICE 'Selects completed in last second: %', selects_completed;
-- Reset counters for the next second
selects_completed := 0;
start_time := clock_timestamp();
END IF;
-- Increment the loop counter
i := i + 1;
END LOOP;
END $$;
./target/release/neon_local stop
baseline: commit 7c74112b2a origin/main
NOTICE: Selects completed in last second: 1864
NOTICE: Selects completed in last second: 1850
NOTICE: Selects completed in last second: 1851
NOTICE: Selects completed in last second: 1918
NOTICE: Selects completed in last second: 1911
NOTICE: Selects completed in last second: 1879
NOTICE: Selects completed in last second: 1858
NOTICE: Selects completed in last second: 1827
NOTICE: Selects completed in last second: 1933
ours
NOTICE: Selects completed in last second: 1915
NOTICE: Selects completed in last second: 1928
NOTICE: Selects completed in last second: 1913
NOTICE: Selects completed in last second: 1932
NOTICE: Selects completed in last second: 1846
NOTICE: Selects completed in last second: 1955
NOTICE: Selects completed in last second: 1991
NOTICE: Selects completed in last second: 1973
```
NB: the ephemeral file sizes differ by ca 1MiB, ours being 1MiB smaller.
</details>
# Rollout
This PR changes the code in-place and is not gated by a feature flag.
After #8655, we needed to mark some tests to shut down immediately. To
aid these tests, try the new pattern of `flush_ep_to_pageserver`
followed by a non-compacting checkpoint. This moves the general graceful
shutdown problem of having too much to flush at shutdown into the test.
Also, add logging for how long the graceful shutdown took, if we got to
complete it for faster log eyeballing.
Fixes: #8712
Cc: #8715, #8708
## Problem
The default Postgres version is set to 15 in code, while we use 16 in
most of the other places (and Postgres 17 is coming)
## Summary of changes
- Run `benchmarks` job with Postgres 16 (instead of Postgres 14)
- Set `DEFAULT_PG_VERSION` to 16 in all places
- Remove deprecated `--pg-version` pytest argument
- Update `test_metadata_bincode_serde_ensure_roundtrip` for Postgres 16
# Motivation
The working theory for hung systemd during PS deploy
(https://github.com/neondatabase/cloud/issues/11387) is that leftover
walredo processes trigger a race condition.
In https://github.com/neondatabase/neon/pull/8150 I arranged that a
clean Tenant shutdown does actually kill its walredo processes.
But many prod machines don't manage to shut down all their tenants until
the 10s systemd timeout hits and, presumably, triggers the race
condition in systemd / the Linux kernel that causes the frozen systemd
# Solution
This PR bolts on a rather ugly mechanism to shut down tenant managers
out of order 8s after we've received the SIGTERM from systemd.
# Changes
- add a global registry of `Weak<WalRedoManager>`
- add a special thread spawned during `shutdown_pageserver` that sleeps
for 8s, then shuts down all redo managers in the registry and prevents
new redo managers from being created
- propagate the new failure mode of tenant spawning throughout the code
base
- make sure shut down tenant manager results in
PageReconstructError::Cancelled so that if Timeline::get calls come in
after the shutdown, they do the right thing
Since the introduction of sharding, the protocol handling loop in
`handle_pagerequests` cannot know anymore which concrete
`Tenant`/`Timeline` object any of the incoming `PagestreamFeMessage`
resolves to.
In fact, one message might resolve to one `Tenant`/`Timeline` while
the next one may resolve to another one.
To avoid going to tenant manager, we added the `shard_timelines` which
acted as an ever-growing cache that held timeline gate guards open for
the lifetime of the connection.
The consequence of holding the gate guards open was that we had to be
sensitive to every cached `Timeline::cancel` on each interaction with
the network connection, so that Timeline shutdown would not have to wait
for network connection interaction.
We can do better than that, meaning more efficiency & better
abstraction.
I proposed a sketch for it in
* https://github.com/neondatabase/neon/pull/8286
and this PR implements an evolution of that sketch.
The main idea is is that `mod page_service` shall be solely concerned
with the following:
1. receiving requests by speaking the protocol / pagestream subprotocol
2. dispatching the request to a corresponding method on the correct
shard/`Timeline` object
3. sending response by speaking the protocol / pagestream subprotocol.
The cancellation sensitivity responsibilities are clear cut:
* while in `page_service` code, sensitivity to page_service cancellation
is sufficient
* while in `Timeline` code, sensitivity to `Timeline::cancel` is
sufficient
To enforce these responsibilities, we introduce the notion of a
`timeline::handle::Handle` to a `Timeline` object that is checked out
from a `timeline::handle::Cache` for **each request**.
The `Handle` derefs to `Timeline` and is supposed to be used for a
single async method invocation on `Timeline`.
See the lengthy doc comment in `mod handle` for details of the design.
## Motivation & Context
We want to move away from `task_mgr` towards explicit tracking of child
tasks.
This PR is extracted from https://github.com/neondatabase/neon/pull/8339
where I refactor `PageRequestHandler` to not depend on task_mgr anymore.
## Changes
This PR refactors all global tasks but `PageRequestHandler` to use some
combination of `JoinHandle`/`JoinSet` + `CancellationToken`.
The `task_mgr::spawn(.., shutdown_process_on_error)` functionality is
preserved through the new `exit_on_panic_or_error` wrapper.
Some global tasks were not using it before, but as of this PR, they are.
The rationale is that all global tasks are relevant for correct
operation of the overall Neon system in one way or another.
## Future Work
After #8339, we can make `task_mgr::spawn` require a `TenantId` instead
of an `Option<TenantId>` which concludes this step of cleanup work and
will help discourage future usage of task_mgr for global tasks.
`trace_read_requests` is a per `Tenant`-object option.
But the `handle_pagerequests` loop doesn't know which
`Tenant` object (i.e., which shard) the request is for.
The remaining use of the `Tenant` object is to check `tenant.cancel`.
That check is incorrect [if the pageserver hosts multiple
shards](https://github.com/neondatabase/neon/issues/7427#issuecomment-2220577518).
I'll fix that in a future PR where I completely eliminate the holding
of `Tenant/Timeline` objects across requests.
See [my code RFC](https://github.com/neondatabase/neon/pull/8286) for
the
high level idea.
Note that we can always bring the tracing functionality if we need it.
But since it's actually about logging the `page_service` wire bytes,
it should be a `page_service`-level config option, not per-Tenant.
And for enabling tracing on a single connection, we can implement
a `set pageserver_trace_connection;` option.
part of https://github.com/neondatabase/neon/issues/7418
# Motivation
(reproducing #7418)
When we do an `InMemoryLayer::write_to_disk`, there is a tremendous
amount of random read I/O, as deltas from the ephemeral file (written in
LSN order) are written out to the delta layer in key order.
In benchmarks (https://github.com/neondatabase/neon/pull/7409) we can
see that this delta layer writing phase is substantially more expensive
than the initial ingest of data, and that within the delta layer write a
significant amount of the CPU time is spent traversing the page cache.
# High-Level Changes
Add a new mode for L0 flush that works as follows:
* Read the full ephemeral file into memory -- layers are much smaller
than total memory, so this is afforable
* Do all the random reads directly from this in memory buffer instead of
using blob IO/page cache/disk reads.
* Add a semaphore to limit how many timelines may concurrently do this
(limit peak memory).
* Make the semaphore configurable via PS config.
# Implementation Details
The new `BlobReaderRef::Slice` is a temporary hack until we can ditch
`blob_io` for `InMemoryLayer` => Plan for this is laid out in
https://github.com/neondatabase/neon/issues/8183
# Correctness
The correctness of this change is quite obvious to me: we do what we did
before (`blob_io`) but read from memory instead of going to disk.
The highest bug potential is in doing owned-buffers IO. I refactored the
API a bit in preliminary PR
https://github.com/neondatabase/neon/pull/8186 to make it less
error-prone, but still, careful review is requested.
# Performance
I manually measured single-client ingest performance from `pgbench -i
...`.
Full report:
https://neondatabase.notion.site/2024-06-28-benchmarking-l0-flush-performance-e98cff3807f94cb38f2054d8c818fe84?pvs=4
tl;dr:
* no speed improvements during ingest, but
* significantly lower pressure on PS PageCache (eviction rate drops to
1/3)
* (that's why I'm working on this)
* noticable but modestly lower CPU time
This is good enough for merging this PR because the changes require
opt-in.
We'll do more testing in staging & pre-prod.
# Stability / Monitoring
**memory consumption**: there's no _hard_ limit on max `InMemoryLayer`
size (aka "checkpoint distance") , hence there's no hard limit on the
memory allocation we do for flushing. In practice, we a) [log a
warning](23827c6b0d/pageserver/src/tenant/timeline.rs (L5741-L5743))
when we flush oversized layers, so we'd know which tenant is to blame
and b) if we were to put a hard limit in place, we would have to decide
what to do if there is an InMemoryLayer that exceeds the limit.
It seems like a better option to guarantee a max size for frozen layer,
dependent on `checkpoint_distance`. Then limit concurrency based on
that.
**metrics**: we do have the
[flush_time_histo](23827c6b0d/pageserver/src/tenant/timeline.rs (L3725-L3726)),
but that includes the wait time for the semaphore. We could add a
separate metric for the time spent after acquiring the semaphore, so one
can infer the wait time. Seems unnecessary at this point, though.
## Problem
In https://github.com/neondatabase/neon/pull/5299, the new config-v1
tenant config file was added to hold the LocationConf type. We left the
old config file in place for forward compat, and because running without
generations (therefore without LocationConf) as still useful before the
storage controller was ready for prime-time.
Closes: https://github.com/neondatabase/neon/issues/5388
## Summary of changes
- Remove code for reading and writing the legacy config file
- Remove Generation::Broken: it was unused.
- Treat missing config file on disk as an error loading a tenant, rather
than defaulting it. We can now remove LocationConf::default, and thereby
guarantee that we never construct a tenant with a None generation.
- Update some comments + add some assertions to clarify that
Generation::None is only used in layer metadata, not in the state of a
running tenant.
- Update docker compose test to create tenants with a generation
## Problem
These APIs have be unused for some time. They were superseded by
/location_conf: the equivalent of ignoring a tenant is now to put it in
secondary mode.
## Summary of changes
- Remove APIs
- Remove tests & helpers that used them
- Remove error variants that are no longer needed.
## Problem
This is historical baggage from when the pageserver could be run with
local disk only: we had a bunch of places where we had to treat remote
storage as optional.
Closes: https://github.com/neondatabase/neon/issues/6890
## Changes
- Remove Option<> around remote storage (in
https://github.com/neondatabase/neon/pull/7722 we made remote storage
clearly mandatory)
- Remove code for deleting old metadata files: they're all gone now.
- Remove other references to metadata files when loading directories, as
none exist.
I checked last 14 days of logs for "found legacy metadata", there are no
instances.
Extracted from https://github.com/neondatabase/neon/pull/7375. We assume
everything >= 0x80 are metadata keys. AUX file keys are part of the
metadata keys, and we use `0x90` as the prefix for AUX file keys.
The AUX file encoding is described in the code comment. We use xxhash128
as the hash algorithm. It seems to be portable according to the
introduction,
> xxHash is an Extremely fast Hash algorithm, processing at RAM speed
limits. Code is highly portable, and produces hashes identical across
all platforms (little / big endian).
...though whether the Rust version follows the same convention is
unknown and might need manual review of the library. Anyways, we can
always change the hash algorithm before rolling it out in
staging/end-user, and I made a quick decision to use xxhash here because
it generates 128b hash + portable. We can save the discussion of which
hash algorithm to use later.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
## Problem
Tenant deletion had a couple of TODOs where we weren't using proper
cancellation tokens that would have aborted the deletions during process
shutdown.
## Summary of changes
- Refactor enough that deletion/shutdown code has access to the
TenantManager's cancellation toke
- Use that cancellation token in tenant deletion instead of dummy
tokens.
Switched the order; doing https://github.com/neondatabase/neon/pull/6139
first then can remove uninit marker after.
## Problem
Previously, existence of a timeline directory was treated as evidence of
the timeline's logical existence. That is no longer the case since we
treat remote storage as the source of truth on each startup: we can
therefore do without this mark file.
The mark file had also been used as a pseudo-lock to guard against
concurrent creations of the same TimelineId -- now that persistence is
no longer required, this is a bit unwieldy.
In #6139 the `Tenant::timelines_creating` was added to protect against
concurrent creations on the same TimelineId, making the uninit mark file
entirely redundant.
## Summary of changes
- Code that writes & reads mark file is removed
- Some nearby `pub` definitions are amended to `pub(crate)`
- `test_duplicate_creation` is added to demonstrate that mutual
exclusion of creations still works.
PR adds a simple at most 1Hz refreshed informational API for querying
pageserver utilization. In this first phase, no actual background
calculation is performed. Instead, the worst possible score is always
returned. The returned bytes information is however correct.
Cc: #6835
Cc: #5331
- The `Attaching tenant` log message omitted some useful information
like the generation and mode
- info-level messages about writing configuration files were
unnecessarily verbose
- During process shutdown, we don't emit logs about the various phases:
this is very cheap to log since we do it once per process lifetime, and
is helpful when figuring out where something got stuck during a hang.
Problem
-------
Before this PR, there was no concurrency limit on initial logical size
computations.
While logical size computations are lazy in theory, in practice
(production), they happen in a short timeframe after restart.
This means that on a PS with 20k tenants, we'd have up to 20k concurrent
initial logical size calculation requests.
This is self-inflicted needless overload.
This hasn't been a problem so far because the `.await` points on the
logical size calculation path never return `Pending`, hence we have a
natural concurrency limit of the number of executor threads.
But, as soon as we return `Pending` somewhere in the logical size
calculation path, other concurrent tasks get scheduled by tokio.
If these other tasks are also logical size calculations, they eventually
pound on the same bottleneck.
For example, in #5479, we want to switch the VirtualFile descriptor
cache to a `tokio::sync::RwLock`, which makes us return `Pending`, and
without measures like this patch, after PS restart, VirtualFile
descriptor cache thrashes heavily for 2 hours until all the logical size
calculations have been computed and the degree of concurrency /
concurrent VirtualFile operations is down to regular levels.
See the *Experiment* section below for details.
<!-- Experiments (see below) show that plain #5479 causes heavy
thrashing of the VirtualFile descriptor cache.
The high degree of concurrency is too much for
In the case of #5479 the VirtualFile descriptor cache size starts
thrashing heavily.
-->
Background
----------
Before this PR, initial logical size calculation was spawned lazily on
first call to `Timeline::get_current_logical_size()`.
In practice (prod), the lazy calculation is triggered by
`WalReceiverConnectionHandler` if the timeline is active according to
storage broker, or by the first iteration of consumption metrics worker
after restart (`MetricsCollection`).
The spawns by walreceiver are high-priority because logical size is
needed by Safekeepers (via walreceiver `PageserverFeedback`) to enforce
the project logical size limit.
The spawns by metrics collection are not on the user-critical path and
hence low-priority. [^consumption_metrics_slo]
[^consumption_metrics_slo]: We can't delay metrics collection
indefintely because there are TBD internal SLOs tied to metrics
collection happening in a timeline manner
(https://github.com/neondatabase/cloud/issues/7408). But let's ignore
that in this issue.
The ratio of walreceiver-initiated spawns vs
consumption-metrics-initiated spawns can be reconstructed from logs
(`spawning logical size computation from context of task kind {:?}"`).
PR #5995 and #6018 adds metrics for this.
First investigation of the ratio lead to the discovery that walreceiver
spawns 75% of init logical size computations.
That's because of two bugs:
- In Safekeepers: https://github.com/neondatabase/neon/issues/5993
- In interaction between Pageservers and Safekeepers:
https://github.com/neondatabase/neon/issues/5962
The safekeeper bug is likely primarily responsible but we don't have the
data yet. The metrics will hopefully provide some insights.
When assessing production-readiness of this PR, please assume that
neither of these bugs are fixed yet.
Changes In This PR
------------------
With this PR, initial logical size calculation is reworked as follows:
First, all initial logical size calculation task_mgr tasks are started
early, as part of timeline activation, and run a retry loop with long
back-off until success. This removes the lazy computation; it was
needless complexity because in practice, we compute all logical sizes
anyways, because consumption metrics collects it.
Second, within the initial logical size calculation task, each attempt
queues behind the background loop concurrency limiter semaphore. This
fixes the performance issue that we pointed out in the "Problem" section
earlier.
Third, there is a twist to queuing behind the background loop
concurrency limiter semaphore. Logical size is needed by Safekeepers
(via walreceiver `PageserverFeedback`) to enforce the project logical
size limit. However, we currently do open walreceiver connections even
before we have an exact logical size. That's bad, and I'll build on top
of this PR to fix that
(https://github.com/neondatabase/neon/issues/5963). But, for the
purposes of this PR, we don't want to introduce a regression, i.e., we
don't want to provide an exact value later than before this PR. The
solution is to introduce a priority-boosting mechanism
(`GetLogicalSizePriority`), allowing callers of
`Timeline::get_current_logical_size` to specify how urgently they need
an exact value. The effect of specifying high urgency is that the
initial logical size calculation task for the timeline will skip the
concurrency limiting semaphore. This should yield effectively the same
behavior as we had before this PR with lazy spawning.
Last, the priority-boosting mechanism obsoletes the `init_order`'s grace
period for initial logical size calculations. It's a separate commit to
reduce the churn during review. We can drop that commit if people think
it's too much churn, and commit it later once we know this PR here
worked as intended.
Experiment With #5479
---------------------
I validated this PR combined with #5479 to assess whether we're making
forward progress towards asyncification.
The setup is an `i3en.3xlarge` instance with 20k tenants, each with one
timeline that has 9 layers.
All tenants are inactive, i.e., not known to SKs nor storage broker.
This means all initial logical size calculations are spawned by
consumption metrics `MetricsCollection` task kind.
The consumption metrics worker starts requesting logical sizes at low
priority immediately after restart. This is achieved by deleting the
consumption metrics cache file on disk before starting
PS.[^consumption_metrics_cache_file]
[^consumption_metrics_cache_file] Consumption metrics worker persists
its interval across restarts to achieve persistent reporting intervals
across PS restarts; delete the state file on disk to get predictable
(and I believe worst-case in terms of concurrency during PS restart)
behavior.
Before this patch, all of these timelines would all do their initial
logical size calculation in parallel, leading to extreme thrashing in
page cache and virtual file cache.
With this patch, the virtual file cache thrashing is reduced
significantly (from 80k `open`-system-calls/second to ~500
`open`-system-calls/second during loading).
### Critique
The obvious critique with above experiment is that there's no skipping
of the semaphore, i.e., the priority-boosting aspect of this PR is not
exercised.
If even just 1% of our 20k tenants in the setup were active in
SK/storage_broker, then 200 logical size calculations would skip the
limiting semaphore immediately after restart and run concurrently.
Further critique: given the two bugs wrt timeline inactive vs active
state that were mentioned in the Background section, we could have 75%
of our 20k tenants being (falsely) active on restart.
So... (next section)
This Doesn't Make Us Ready For Async VirtualFile
------------------------------------------------
This PR is a step towards asynchronous `VirtualFile`, aka, #5479 or even
#4744.
But it doesn't yet enable us to ship #5479.
The reason is that this PR doesn't limit the amount of high-priority
logical size computations.
If there are many high-priority logical size calculations requested,
we'll fall over like we did if #5479 is applied without this PR.
And currently, at very least due to the bugs mentioned in the Background
section, we run thousands of high-priority logical size calculations on
PS startup in prod.
So, at a minimum, we need to fix these bugs.
Then we can ship #5479 and #4744, and things will likely be fine under
normal operation.
But in high-traffic situations, overload problems will still be more
likely to happen, e.g., VirtualFile cache descriptor thrashing.
The solution candidates for that are orthogonal to this PR though:
* global concurrency limiting
* per-tenant rate limiting => #5899
* load shedding
* scaling bottleneck resources (fd cache size (neondatabase/cloud#8351),
page cache size(neondatabase/cloud#8351), spread load across more PSes,
etc)
Conclusion
----------
Even with the remarks from in the previous section, we should merge this
PR because:
1. it's an improvement over the status quo (esp. if the aforementioned
bugs wrt timeline active / inactive are fixed)
2. it prepares the way for
https://github.com/neondatabase/neon/pull/6010
3. it gets us close to shipping #5479 and #4744
## Problem
When shutting down a Tenant, it isn't just important to cause any
background tasks to stop. It's also important to wait until they have
stopped before declaring shutdown complete, in cases where we may re-use
the tenant's local storage for something else, such as running in
secondary mode, or creating a new tenant with the same ID.
## Summary of changes
A `Gate` class is added, inspired by
[seastar::gate](https://docs.seastar.io/master/classseastar_1_1gate.html).
For types that have an important lifetime that corresponds to some
physical resource, use of a Gate as well as a CancellationToken provides
a robust pattern for async requests & shutdown:
- Requests must always acquire the gate as long as they are using the
object
- Shutdown must set the cancellation token, and then `close()` the gate
to wait for requests in progress before returning.
This is not for memory safety: it's for expressing the difference
between "Arc<Tenant> exists", and "This tenant's files on disk are
eligible to be read/written".
- Both Tenant and Timeline get a Gate & CancellationToken.
- The Timeline gate is held during eviction of layers, and during
page_service requests.
- Existing cancellation support in page_service is refined to use the
timeline-scope cancellation token instead of a process-scope
cancellation token. This replaces the use of `task_mgr::associate_with`:
tasks no longer change their tenant/timelineidentity after being
spawned.
The Tenant's Gate is not yet used, but will be important for
Tenant-scoped operations in secondary mode, where we must ensure that
our secondary-mode downloads for a tenant are gated wrt the activity of
an attached Tenant.
This is part of a broader move away from using the global-state driven
`task_mgr` shutdown tokens:
- less global state where we rely on implicit knowledge of what task a
given function is running in, and more explicit references to the
cancellation token that a particular function/type will respect, making
shutdown easier to reason about.
- eventually avoid the big global TASKS mutex.
---------
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
- finally add an `#[instrument]` to Timeline::create_image_layers,
making it easier to see that something is happening because we create
image layers
- format some macro context code
- add a warning not to create new validation functions a la parse do not
validate
Split off from #5198.
## Problem
Currently it's unclear how much of the `initial_tenant_load` period is
in S3 objects, and therefore how impactful it is to make changes to
remote operations during startup.
## Summary of changes
- `Tenant::load` is refactored to load remote indices in parallel and to
wait for all these remote downloads to finish before it proceeds to
construct any `Timeline` objects.
- `pageserver_startup_duration_seconds` gets a new `phase` value of
`initial_tenant_load_remote` which counts the time from startup to when
the last tenant finishes loading remote content.
- `test_pageserver_restart` is extended to validate this phase. The
previous version of the test was relying on order of dict entries, which
stopped working when adding a phase, so this is refactored a bit.
- `test_pageserver_restart` used to explicitly create a branch, now it
uses the default initial_timeline. This avoids startup getting held up
waiting for logical sizes, when one of the branches is not in use.
## Problem
These changes are part of building seamless tenant migration, as
described in the RFC:
- https://github.com/neondatabase/neon/pull/5029
## Summary of changes
- A new configuration type `LocationConf` supersedes `TenantConfOpt` for
storing a tenant's configuration in the pageserver repo dir. It contains
`TenantConfOpt`, as well as a new `mode` attribute that describes what
kind of location this is (secondary, attached, attachment mode etc). It
is written to a file called `config-v1` instead of `config` -- this
prepares us for neatly making any other profound changes to the format
of the file in future. Forward compat for existing pageserver code is
achieved by writing out both old and new style files. Backward compat is
achieved by checking for the old-style file if the new one isn't found.
- The `TenantMap` type changes, to hold `TenantSlot` instead of just
`Tenant`. The `Tenant` type continues to be used for attached tenants
only. Tenants in other states (such as secondaries) are represented by a
different variant of `TenantSlot`.
- Where `Tenant` & `Timeline` used to hold an Arc<Mutex<TenantConfOpt>>,
they now hold a reference to a AttachedTenantConf, which includes the
extra information from LocationConf. This enables them to know the
current attachment mode.
- The attachment mode is used as an advisory input to decide whether to
do compaction and GC (AttachedStale is meant to avoid doing uploads,
AttachedMulti is meant to avoid doing deletions).
- A new HTTP API is added at `PUT /tenants/<tenant_id>/location_config`
to drive new location configuration. This provides a superset of the
functionality of attach/detach/load/ignore:
- Attaching a tenant is just configuring it in an attached state
- Detaching a tenant is configuring it to a detached state
- Loading a tenant is just the same as attaching it
- Ignoring a tenant is the same as configuring it into Secondary with
warm=false (i.e. retain the files on disk but do nothing else).
Caveats:
- AttachedMulti tenants don't do compaction in this PR, but they do in
the follow on #5397
- Concurrent updates to the `location_config` API are not handled
elegantly in this PR, a better mechanism is added in the follow on
https://github.com/neondatabase/neon/pull/5367
- Secondary mode is just a placeholder in this PR: the code to upload
heatmaps and do downloads on secondary locations will be added in a
later PR (but that shouldn't change any external interfaces)
Closes: https://github.com/neondatabase/neon/issues/5379
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
Fixes#4689 by replacing all of `std::Path` , `std::PathBuf` with
`camino::Utf8Path`, `camino::Utf8PathBuf` in
- pageserver
- safekeeper
- control_plane
- libs/remote_storage
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
## Problem
Pageservers must not delete objects or advertise updates to
remote_consistent_lsn without checking that they hold the latest
generation for the tenant in question (see [the RFC](
https://github.com/neondatabase/neon/blob/main/docs/rfcs/025-generation-numbers.md))
In this PR:
- A new "deletion queue" subsystem is introduced, through which
deletions flow
- `RemoteTimelineClient` is modified to send deletions through the
deletion queue:
- For GC & compaction, deletions flow through the full generation
verifying process
- For timeline deletions, deletions take a fast path that bypasses
generation verification
- The `last_uploaded_consistent_lsn` value in `UploadQueue` is replaced
with a mechanism that maintains a "projected" lsn (equivalent to the
previous property), and a "visible" LSN (which is the one that we may
share with safekeepers).
- Until `control_plane_api` is set, all deletions skip generation
validation
- Tests are introduced for the new functionality in
`test_pageserver_generations.py`
Once this lands, if a pageserver is configured with the
`control_plane_api` configuration added in
https://github.com/neondatabase/neon/pull/5163, it becomes safe to
attach a tenant to multiple pageservers concurrently.
---------
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
- #5050
Closes: https://github.com/neondatabase/neon/issues/5136
## Summary of changes
- A new configuration property `control_plane_api` controls other
functionality in this PR: if it is unset (default) then everything still
works as it does today.
- If `control_plane_api` is set, then on startup we call out to control
plane `/re-attach` endpoint to discover our attachments and their
generations. If an attachment is missing from the response we implicitly
detach the tenant.
- Calls to pageserver `/attach` API may include a `generation`
parameter. If `control_plane_api` is set, then this parameter is
mandatory.
- RemoteTimelineClient's loading of index_part.json is generation-aware,
and will try to load the index_part with the most recent generation <=
its own generation.
- The `neon_local` testing environment now includes a new binary
`attachment_service` which implements the endpoints that the pageserver
requires to operate. This is on by default if running `cargo neon` by
hand. In `test_runner/` tests, it is off by default: existing tests
continue to run with in the legacy generation-less mode.
Caveats:
- The re-attachment during startup assumes that we are only re-attaching
tenants that have previously been attached, and not totally new tenants
-- this relies on the control plane's attachment logic to keep retrying
so that we should eventually see the attach API call. That's important
because the `/re-attach` API doesn't tell us which timelines we should
attach -- we still use local disk state for that. Ref:
https://github.com/neondatabase/neon/issues/5173
- Testing: generations are only enabled for one integration test right
now (test_pageserver_restart), as a smoke test that all the machinery
basically works. Writing fuller tests that stress tenant migration will
come later, and involve extending our test fixtures to deal with
multiple pageservers.
- I'm not in love with "attachment_service" as a name for the neon_local
component, but it's not very important because we can easily rename
these test bits whenever we want.
- Limited observability when in re-attach on startup: when I add
generation validation for deletions in a later PR, I want to wrap up the
control plane API calls in some small client class that will expose
metrics for things like errors calling the control plane API, which will
act as a strong red signal that something is not right.
Co-authored-by: Christian Schwarz <christian@neon.tech>
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
- move them to pageserver which is the only dependant on the crate fail
- "move" the exported macro to the new module
- support at init time the same failpoints as runtime
Found while debugging test failures and making tests more repeatable by
allowing "exit" from pageserver start via environment variables. Made
those changes to `test_gc_cutoff.py`.
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
Originated from test failure where we got SlowDown error from s3.
The patch generalizes `download_retry` to not be download specific.
Resulting `retry` function is moved to utils crate. `download_retries`
is now a thin wrapper around this `retry` function.
To ensure that all needed retries are in place test code now uses
`test_remote_failures=1` setting.
Ref https://neondb.slack.com/archives/C059ZC138NR/p1691743624353009
## Problem
Currently to know how long pageserver startup took requires inspecting
logs.
## Summary of changes
`pageserver_startup_duration_ms` metric is added, with label `phase` for
different phases of startup.
These are broken down by phase, where the phases correspond to the
existing wait points in the code:
- Start of doing I/O
- When tenant load is done
- When initial size calculation is done
- When background jobs start
- Then "complete" when everything is done.
`pageserver_startup_is_loading` is a 0/1 gauge that indicates whether we are in the initial load of tenants.
`pageserver_tenant_activation_seconds` is a histogram of time in seconds taken to activate a tenant.
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
During deploys of 2023-08-03 we logged too much on shutdown. Fix the
logging by timing each top level shutdown step, and possibly warn on it
taking more than a rough threshold, based on how long I think it
possibly should be taking. Also remove all shutdown logging from
background tasks since there is already "shutdown is taking a long time"
logging.
Co-authored-by: John Spray <john@neon.tech>
## Problem
Currently we delete local files first, so if pageserver restarts after
local files deletion then remote deletion is not continued. This can be
solved with inversion of these steps.
But even if these steps are inverted when index_part.json is deleted
there is no way to distinguish between "this timeline is good, we just
didnt upload it to remote" and "this timeline is deleted we should
continue with removal of local state". So to solve it we use another
mark file. After index part is deleted presence of this mark file
indentifies that it was a deletion intention.
Alternative approach that was discussed was to delete all except
metadata first, and then delete metadata and index part. In this case we
still do not support local only configs making them rather unsafe
(deletion in them is already unsafe, but this direction solidifies this
direction instead of fixing it). Another downside is that if we crash
after local metadata gets removed we may leave dangling index part on
the remote which in theory shouldnt be a big deal because the file is
small.
It is not a big change to choose another approach at this point.
## Summary of changes
Timeline deletion sequence:
1. Set deleted_at in remote index part.
2. Create local mark file.
3. Delete local files except metadata (it is simpler this way, to be
able to reuse timeline initialization code that expects metadata)
4. Delete remote layers
5. Delete index part
6. Delete meta, timeline directory.
7. Delete mark file.
This works for local only configuration without remote storage.
Sequence is resumable from any point.
resolves#4453
resolves https://github.com/neondatabase/neon/pull/4552 (the issue was
created with async cancellation in mind, but we can still have issues
with retries if metadata is deleted among the first by remove_dir_all
(which doesnt have any ordering guarantees))
---------
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
Co-authored-by: Christian Schwarz <christian@neon.tech>
Initial logical size calculation could still hinder our fast startup
efforts in #4397. See #4183. In deployment of 2023-06-06
about a 200 initial logical sizes were calculated on hosts which
took the longest to complete initial load (12s).
Implements the three step/tier initialization ordering described in
#4397:
1. load local tenants
2. do initial logical sizes per walreceivers for 10s
3. background tasks
Ordering is controlled by:
- waiting on `utils::completion::Barrier`s on background tasks
- having one attempt for each Timeline to do initial logical size
calculation
- `pageserver/src/bin/pageserver.rs` releasing background jobs after
timeout or completion of initial logical size calculation
The timeout is there just to safeguard in case a legitimate non-broken
timeline initial logical size calculation goes long. The timeout is
configurable, by default 10s, which I think would be fine for production
systems. In the test cases I've been looking at, it seems that these
steps are completed as fast as possible.
Co-authored-by: Christian Schwarz <christian@neon.tech>
We have 2 ways of tenant shutdown, we should have just one.
Changes are mostly mechanical simple refactorings.
Added `warn!` on the "shutdown all remaining tasks" should trigger test
failures in the between time of not having solved the "tenant/timeline
owns all spawned tasks" issue.
Cc: #4327.
This is preliminary work for/from #4220 (async
`Layer::get_value_reconstruct_data`).
The motivation is to avoid locking `Tenant::timelines` in places that
can't be `async`, because in #4333 we want to convert Tenant::timelines
from `std::sync::Mutex` to `tokio::sync::Mutex`.
But, the changes here are useful in general because they clean up &
document tenant state transitions.
That also paves the way for #4350, which is an alternative to #4333 that
refactors the pageserver code so that we can keep the
`Tenant::timelines` mutex sync.
This patch consists of the following core insights and changes:
* spawn_load and spawn_attach own the tenant state until they're done
* once load()/attach() calls are done ...
* if they failed, transition them to Broken directly (we know that
there's no background activity because we didn't call activate yet)
* if they succeed, call activate. We can make it infallible. How? Later.
* set_broken() and set_stopping() are changed to wait for spawn_load() /
spawn_attach() to finish.
* This sounds scary because it might hinder detach or shutdown, but
actually, concurrent attach+detach, or attach+shutdown, or
load+shutdown, or attach+shutdown were just racy before this PR.
So, with this change, they're not anymore.
In the future, we can add a `CancellationToken` stored in Tenant to
cancel `load` and `attach` faster, i.e., make `spawn_load` /
`spawn_attach` transition them to Broken state sooner.
See the doc comments on TenantState for the state transitions that are
now possible.
It might seem scary, but actually, this patch reduces the possible state
transitions.
We introduce a new state `TenantState::Activating` to avoid grabbing the
`Tenant::timelines` lock inside the `send_modify` closure.
These were the humble beginnings of this PR (see Motivation section),
and I think it's still the right thing to have this `Activating` state,
even if we decide against async `Tenant::timelines` mutex. The reason is
that `send_modify` locks internally, and by moving locking of
Tenant::timelines out of the closure, the internal locking of
`send_modify` becomes a leaf of the lock graph, and so, we eliminate
deadlock risk.
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
(This is prep work to make `Timeline::activate` infallible.)
This patch removes the global storage_broker client instance from the
pageserver codebase.
Instead, pageserver startup instantiates it and passes it down to the
`Timeline::activate` function, which in turn passes it to the
WalReceiver, which is the entity that actually uses it.
Patch series:
- #4316
- #4317
- #4318
- #4319
This patch adds a pageserver-global background loop that evicts layers
in response to a shortage of available bytes in the $repo/tenants
directory's filesystem.
The loop runs periodically at a configurable `period`.
Each loop iteration uses `statvfs` to determine filesystem-level space
usage. It compares the returned usage data against two different types
of thresholds. The iteration tries to evict layers until app-internal
accounting says we should be below the thresholds. We cross-check this
internal accounting with the real world by making another `statvfs` at
the end of the iteration. We're good if that second statvfs shows that
we're _actually_ below the configured thresholds. If we're still above
one or more thresholds, we emit a warning log message, leaving it to the
operator to investigate further.
There are two thresholds:
- `max_usage_pct` is the relative available space, expressed in percent
of the total filesystem space. If the actual usage is higher, the
threshold is exceeded.
- `min_avail_bytes` is the absolute available space in bytes. If the
actual usage is lower, the threshold is exceeded.
The iteration evicts layers in LRU fashion with a reservation of up to
`tenant_min_resident_size` bytes of the most recent layers per tenant.
The layers not part of the per-tenant reservation are evicted
least-recently-used first until we're below all thresholds. The
`tenant_min_resident_size` can be overridden per tenant as
`min_resident_size_override` (bytes).
In addition to the loop, there is also an HTTP endpoint to perform one
loop iteration synchronous to the request. The endpoint takes an
absolute number of bytes that the iteration needs to evict before
pressure is relieved. The tests use this endpoint, which is a great
simplification over setting up loopback-mounts in the tests, which would
be required to test the statvfs part of the implementation. We will rely
on manual testing in staging to test the statvfs parts.
The HTTP endpoint is also handy in emergencies where an operator wants
the pageserver to evict a given amount of space _now. Hence, it's
arguments documented in openapi_spec.yml. The response type isn't
documented though because we don't consider it stable. The endpoint
should _not_ be used by Console but it could be used by on-call.
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
Co-authored-by: Dmitry Rodionov <dmitry@neon.tech>
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
Walreceiver is a per-timeline abstraction. Move it there to reflect
the hierarchy of abstractions and task_mgr tasks.
The code that sets up the global storage_broker client
is not timeline-scoped. So, break it out into a separate module.
The motivation for this change is to prepare the code base for replacing
the task_mgr global task registry with a more ownership-oriented
approach to manage task lifetimes.
I removed TaskStateUpdate::Init because, after doing the changes,
rustc warned that it was never constructed.
A quick search through the commit history shows that this
has always been true since
commit fb68d01449
Author: Dmitry Rodionov <dmitry@neon.tech>
Date: Mon Sep 26 23:57:02 2022 +0300
Preserve task result in TaskHandle by keeping join handle around (#2521)
So, the warning is not an indication of some accidental code removal.
This is PR: https://github.com/neondatabase/neon/pull/3456
Motivation
==========
Layer Eviction Needs Context
----------------------------
Before we start implementing layer eviction, we need to collect some
access statistics per layer file or maybe even page.
Part of these statistics should be the initiator of a page read request
to answer the question of whether it was page_service vs. one of the
background loops, and if the latter, which of them?
Further, it would be nice to learn more about what activity in the pageserver
initiated an on-demand download of a layer file.
We will use this information to test out layer eviction policies.
Read more about the current plan for layer eviction here:
https://github.com/neondatabase/neon/issues/2476#issuecomment-1370822104
task_mgr problems + cancellation + tenant/timeline lifecycle
------------------------------------------------------------
Apart from layer eviction, we have long-standing problems with task_mgr,
task cancellation, and various races around tenant / timeline lifecycle
transitions.
One approach to solve these is to abandon task_mgr in favor of a
mechanism similar to Golang's context.Context, albeit extended to
support waiting for completion, and specialized to the needs in the
pageserver.
Heikki solves all of the above at once in PR
https://github.com/neondatabase/neon/pull/3228 , which is not yet
merged at the time of writing.
What Is This Patch About
========================
This patch addresses the immediate needs of layer eviction by
introducing a `RequestContext` structure that is plumbed through the
pageserver - all the way from the various entrypoints (page_service,
management API, tenant background loops) down to
Timeline::{get,get_reconstruct_data}.
The struct carries a description of the kind of activity that initiated
the call. We re-use task_mgr::TaskKind for this.
Also, it carries the desired on-demand download behavior of the entrypoint.
Timeline::get_reconstruct_data can then log the TaskKind that initiated
the on-demand download.
I developed this patch by git-checking-out Heikki's big RequestContext
PR https://github.com/neondatabase/neon/pull/3228 , then deleting all
the functionality that we do not need to address the needs for layer
eviction.
After that, I added a few things on top:
1. The concept of attached_child and detached_child in preparation for
cancellation signalling through RequestContext, which will be added in
a future patch.
2. A kill switch to turn DownloadBehavior::Error into a warning.
3. Renamed WalReceiverConnection to WalReceiverConnectionPoller and
added an additional TaskKind WalReceiverConnectionHandler.These were
necessary to create proper detached_child-type RequestContexts for the
various tasks that walreceiver starts.
How To Review This Patch
========================
Start your review with the module-level comment in context.rs.
It explains the idea of RequestContext, what parts of it are implemented
in this patch, and the future plans for RequestContext.
Then review the various `task_mgr::spawn` call sites. At each of them,
we should be creating a new detached_child RequestContext.
Then review the (few) RequestContext::attached_child call sites and
ensure that the spawned tasks do not outlive the task that spawns them.
If they do, these call sites should use detached_child() instead.
Then review the todo_child() call sites and judge whether it's worth the
trouble of plumbing through a parent context from the caller(s).
Lastly, go through the bulk of mechanical changes that simply forwards
the &ctx.
