## Problem
The page service logic asserts that a tracing span is present with
tenant/timeline/shard IDs. An initial gRPC page service implementation
thus requires a tracing span.
Touches https://github.com/neondatabase/neon/issues/11728.
## Summary of changes
Adds an `ObservabilityLayer` middleware that generates a tracing span
and decorates it with IDs from the gRPC metadata.
This is a minimal implementation to address the tracing span assertion.
It will be extended with additional observability in later PRs.
Precursor to https://github.com/neondatabase/cloud/issues/28333.
We want per-endpoint configuration for rate limits, which will be
distributed via the `GetEndpointAccessControl` API. This lays some of
the ground work.
1. Allow the endpoint rate limiter to accept a custom leaky bucket
config on check.
2. Remove the unused auth rate limiter, as I don't want to think about
how it fits into this.
3. Refactor the caching of `GetEndpointAccessControl`, as it adds
friction for adding new cached data to the API.
That third one was rather large. I couldn't find any way to split it up.
The core idea is that there's now only 2 cache APIs.
`get_endpoint_access_controls` and `get_role_access_controls`.
I'm pretty sure the behaviour is unchanged, except I did a drive by
change to fix#8989 because it felt harmless. The change in question is
that when a password validation fails, we eagerly expire the role cache
if the role was cached for 5 minutes. This is to allow for edge cases
where a user tries to connect with a reset password, but the cache never
expires the entry due to some redis related quirk (lag, or
misconfiguration, or cplane error)
There were some incompatible changes. Most churn was from switching from
the now-deprecated fcntl:flock() function to
fcntl::Flock::lock(). The new function returns a guard object, while
with the old function, the lock was associated directly with the file
descriptor.
It's good to stay up-to-date in general, but the impetus to do this now
is that in https://github.com/neondatabase/neon/pull/11929, I want to
use some functions that were added only in the latest version of 'nix',
and it's nice to not have to build multiple versions. (Although,
different versions of 'nix' are still pulled in as indirect dependencies
from other packages)
Use the current production config for batching & concurrent IO.
Remove the permutation testing for unit tests from CI.
(The pageserver unit test matrix takes ~10min for debug builds).
Drive-by-fix use of `if cfg!(test)` inside crate `pageserver_api`.
It is ineffective for early-enabling new defaults for pageserver unit
tests only.
The reason is that the `test` cfg is only set for the crate under test
but not its dependencies.
So, `cargo test -p pageserver` will build `pageserver_api` with
`cfg!(test) == false`.
Resort to checking for feature flag `testing` instead, since all our
unit tests are run with `--feature testing`.
refs
- `scattered-lsn` batching has been implemented and rolled out in all
envs, cf https://github.com/neondatabase/neon/issues/10765
- preliminary for https://github.com/neondatabase/neon/pull/10466
- epic https://github.com/neondatabase/neon/issues/9377
- epic https://github.com/neondatabase/neon/issues/9378
- drive-by fix
https://neondb.slack.com/archives/C0277TKAJCA/p1746821515504219
# Problem
Before this PR, `test_pageserver_catchup_while_compute_down` would
occasionally fail due to scary-looking WARN log line
```
WARN ephemeral_file_buffered_writer{...}:flush_attempt{attempt=1}: \
error flushing buffered writer buffer to disk, retrying after backoff err=Operation canceled (os error 125)
```
After lengthy investigation, the conclusion is that this is likely due
to a kernel bug related due to io_uring async workers (io-wq) and
signals.
The main indicator is that the error only ever happens in correlation
with pageserver shtudown when SIGTERM is received.
There is a fix that is merged in 6.14
kernels (`io-wq: backoff when retrying worker creation`).
However, even when I revert that patch, the issue is not reproducible
on 6.14, so, it remains a speculation.
It was ruled out that the ECANCELED is due to the executor thread
exiting before the async worker starts processing the operation.
# Solution
The workaround in this issue is to retry the operation on ECANCELED
once.
Retries are safe because the low-level io_engine operations are
idempotent.
(We don't use O_APPEND and I can't think of another flag that would make
the APIs covered by this patch not idempotent.)
# Testing
With this PR, the warn! log no longer happens on [my reproducer
setup](https://github.com/neondatabase/neon/issues/11446#issuecomment-2843015111).
And the new rate-limited `info!`-level log line informing about the
internal retry shows up instead, as expected.
# Refs
- fixes https://github.com/neondatabase/neon/issues/11446
Add `/lfc/(prewarm|offload)` routes to `compute_ctl` which interact with
endpoint storage.
Add `prewarm_lfc_on_startup` spec option which, if enabled, downloads
LFC prewarm data on compute startup.
Resolves: https://github.com/neondatabase/cloud/issues/26343
Instead of encoding a certain structure for claims, let's allow the
caller to specify what claims be encoded.
Signed-off-by: Tristan Partin <tristan@neon.tech>
I like to run nightly clippy every so often to make our future rust
upgrades easier. Some notable changes:
* Prefer `next_back()` over `last()`. Generic iterators will implement
`last()` to run forward through the iterator until the end.
* Prefer `io::Error::other()`.
* Use implicit returns
One case where I haven't dealt with the issues is the now
[more-sensitive "large enum variant"
lint](https://github.com/rust-lang/rust-clippy/pull/13833). I chose not
to take any decisions around it here, and simply marked them as allow
for now.
Service targeted for storing and retrieving LFC prewarm data.
Can be used for proxying S3 access for Postgres extensions like
pg_mooncake as well.
Requests must include a Bearer JWT token.
Token is validated using a pemfile (should be passed in infra/).
Note: app is not tolerant to extra trailing slashes, see app.rs
`delete_prefix` test for comments.
Resolves: https://github.com/neondatabase/cloud/issues/26342
Unrelated changes: gate a `rename_noreplace` feature and disable it in
`remote_storage` so as `object_storage` can be built with musl
## Problem
We don't have metrics to exactly quantify the end user impact of
on-demand downloads.
Perf tracing is underway (#11140) to supply us with high-resolution
*samples*.
But it will also be useful to have some aggregate per-timeline and
per-instance metrics that definitively contain all observations.
## Summary of changes
This PR consists of independent commits that should be reviewed
independently.
However, for convenience, we're going to merge them together.
- refactor(metrics): measure_remote_op can use async traits
- impr(pageserver metrics): task_kind dimension for
remote_timeline_client latency histo
- implements https://github.com/neondatabase/cloud/issues/26800
- refs
https://github.com/neondatabase/cloud/issues/26193#issuecomment-2769705793
- use the opportunity to rename the metric and add a _global suffix;
checked grafana export, it's only used in two personal dashboards, one
of them mine, the other by Heikki
- log on-demand download latency for expensive-to-query but precise
ground truth
- metric for wall clock time spent waiting for on-demand downloads
## Refs
- refs https://github.com/neondatabase/cloud/issues/26800
- a bunch of minor investigations / incidents into latency outliers
In
-
https://github.com/neondatabase/neon/pull/10993#issuecomment-2690428336
I added infinite retries for buffered writer flush IOs, primarily to
gracefully handle ENOSPC but more generally so that the buffered writer
is not left in a state where reads from the surrounding InMemoryLayer
cause panics.
However, I didn't add cancellation sensitivity, which is concerning
because then there is no way to detach a timeline/tenant that is
encountering the write IO errors.
That’s a legitimate scenario in the case of some edge case bug.
See the #10993 description for details.
This PR
- first makes flush loop infallible, enabled by infinite retries
- then adds sensitivity to `Timeline::cancel` to the flush loop, thereby
making it fallible in one specific way again
- finally fixes the InMemoryLayer/EphemeralFile/BufferedWriter
amalgamate to remain read-available after flush loop is cancelled.
The support for read-availability after cancellation is necessary so
that reads from the InMemoryLayer that are already queued up behind the
RwLock that wraps the BufferedWriter won't panic because of the
`mutable=None` that we leave behind in case the flush loop gets
cancelled.
# Alternatives
One might think that we can only ship the change for read-availability
if flush encounters an error, without the infinite retrying and/or
cancellation sensitivity complexity.
The problem with that is that read-availability sounds good but is
really quite useless, because we cannot ingest new WAL without a
writable InMemoryLayer. Thus, very soon after we transition to read-only
mode, reads from compute are going to wait anyway, but on `wait_lsn`
instead of the RwLock, because ingest isn't progressing.
Thus, having the infinite flush retries still makes more sense because
they're just "slowness" to the user, whereas wait_lsn is hard errors.
# Problem
We leave too few observability breadcrumbs in the case where wait_lsn is
exceptionally slow.
# Changes
- refactor: extract the monitoring logic out of `log_slow` into
`monitor_slow_future`
- add global + per-timeline counter for time spent waiting for wait_lsn
- It is updated while we're still waiting, similar to what we do for
page_service response flush.
- add per-timeline counterpair for started & finished wait_lsn count
- add slow-logging to leave breadcrumbs in logs, not just metrics
For the slow-logging, we need to consider not flooding the logs during a
broker or network outage/blip.
The solution is a "log-streak-level" concurrency limit per timeline.
At any given time, there is at most one slow wait_lsn that is logging
the "still running" and "completed" sequence of logs.
Other concurrent slow wait_lsn's don't log at all.
This leaves at least one breadcrumb in each timeline's logs if some
wait_lsn was exceptionally slow during a given period.
The full degree of slowness can then be determined by looking at the
per-timeline metric.
# Performance
Reran the `bench_log_slow` benchmark, no difference, so, existing call
sites are fine.
We do use a Semaphore, but only try_acquire it _after_ things have
already been determined to be slow. So, no baseline overhead
anticipated.
# Refs
-
https://github.com/neondatabase/cloud/issues/23486#issuecomment-2711587222
We want to export performance traces from the pageserver in OTEL format.
End goal is to see them in Grafana.
To this end, there are two changes here:
1. Update the `tracing-utils` crate to allow for explicitly specifying
the export configuration. Pageserver configuration is loaded from a file
on start-up. This allows us to use the same flow for export configs
there.
2. Update the `utils::logging::init` common entry point to set up OTEL
tracing infrastructure if requested. Note that an entirely different
tracing subscriber is used. This is to avoid interference with the
existing tracing set-up. For now, no service uses this functionality.
PR to plug this into the pageserver is
[here](https://github.com/neondatabase/neon/pull/11140).
Related https://github.com/neondatabase/neon/issues/9873
## Problem
We don't have any logging for Sentry initialization. This makes it hard
to verify that it has been configured correctly.
## Summary of changes
Log some basic info when Sentry has been initialized, but omit the
public key (which allows submitting events). Also log when `SENTRY_DSN`
isn't specified at all, and when it fails to initialize (which is
supposed to panic, but we may as well).
## Problem
The code to generate symbolized pprof heap profiles and flamegraph SVGs
has been upstreamed to the `jemalloc_pprof` crate:
* https://github.com/polarsignals/rust-jemalloc-pprof/pull/22
* https://github.com/polarsignals/rust-jemalloc-pprof/pull/23
## Summary of changes
Use `jemalloc_pprof` to generate symbolized pprof heap profiles and
flamegraph SVGs.
This reintroduces a bunch of internal jemalloc stack frames that we'd
previously strip, e.g. each stack now always ends with
`prof_backtrace_impl` (where jemalloc takes a stack trace for heap
profiling), but that seems ok.
Migrates the remaining crates to edition 2024. We like to stay on the
latest edition if possible. There is no functional changes, however some
code changes had to be done to accommodate the edition's breaking
changes.
Like the previous migration PRs, this is comprised of three commits:
* the first does the edition update and makes `cargo check`/`cargo
clippy` pass. we had to update bindgen to make its output [satisfy the
requirements of edition
2024](https://doc.rust-lang.org/edition-guide/rust-2024/unsafe-extern.html)
* the second commit does a `cargo fmt` for the new style edition.
* the third commit reorders imports as a one-off change. As before, it
is entirely optional.
Part of #10918
## Problem
We recently added slow GetPage request logging. However, this
unintentionally included the flush time when logging (which we already
have separate logging for). It also logs at WARN level, which is a bit
aggressive since we see this fire quite frequently.
Follows https://github.com/neondatabase/neon/pull/10906.
## Summary of changes
* Only log the request execution time, not the flush time.
* Extract a `pagestream_dispatch_batched_message()` helper.
* Rename `warn_slow()` to `log_slow()` and downgrade to INFO.
## Problem
We don't have good observability for "stuck" getpage requests.
Resolves https://github.com/neondatabase/cloud/issues/23808.
## Summary of changes
Log a periodic warning (every 30 seconds) if GetPage request execution
is slow to complete, to aid in debugging stuck GetPage requests.
This does not cover response flushing (we have separate logging for
that), nor reading the request from the socket and batching it (expected
to be insignificant and not straightforward to handle with the current
protocol).
This costs 95 nanoseconds on the happy path when awaiting a
`tokio::task::yield_now()`:
```
warn_slow/enabled=false time: [45.716 ns 46.116 ns 46.687 ns]
warn_slow/enabled=true time: [141.53 ns 141.83 ns 142.18 ns]
```
## Problem
We lack an API for warming up attached locations based on the heatmap
contents.
This is problematic in two places:
1. If we manually migrate and cut over while the secondary is still cold
2. When we re-attach a previously offloaded tenant
## Summary of changes
https://github.com/neondatabase/neon/pull/10597 made heatmap generation
additive
across migrations, so we won't clobber it a after a cold migration. This
allows us to implement:
1. An endpoint for downloading all missing heatmap layers on the
pageserver:
`/v1/tenant/:tenant_shard_id/timeline/:timeline_id/download_heatmap_layers`.
Only one such operation per timeline is allowed at any given time. The
granularity is tenant shard.
2. An endpoint to the storage controller to trigger the downloads on the
pageserver:
`/v1/tenant/:tenant_shard_id/timeline/:timeline_id/download_heatmap_layers`.
This works both at
tenant and tenant shard level. If an unsharded tenant id is provided,
the operation is started on
all shards, otherwise only the specified shard.
3. A storcon cli command. Again, tenant and tenant-shard level
granularities are supported.
Cplane will call into storcon and trigger the downloads for all shards.
When we want to rescue a migration, we will use storcon cli targeting
the specific tenant shard.
Related: https://github.com/neondatabase/neon/issues/10541
Preparations for a successor of #10440:
* move `pull_timeline` to `safekeeper_api` and add it to
`SafekeeperClient`. we want to do `pull_timeline` on any creations that
we couldn't do initially.
* Add a `SafekeeperGeneration` type instead of relying on a type alias.
we want to maintain a safekeeper specific generation number now in the
storcon database. A separate type is important to make it impossible to
mix it up with the tenant's pageserver specific generation number. We
absolutely want to avoid that for correctness reasons. If someone mixes
up a safekeeper and pageserver id (both use the `NodeId` type), that's
bad but there is no wrong generations flying around.
part of #9011
# Summary
In
- https://github.com/neondatabase/neon/pull/10813
we added slow flush logging but it didn't log the TCP send & recv queue
length.
This PR adds that data to the log message.
I believe the implementation to be safe & correct right now, but it's
brittle and thus this PR should be reverted or improved upon once the
investigation is over.
Refs:
- stacked atop https://github.com/neondatabase/neon/pull/10813
- context:
https://neondb.slack.com/archives/C08DE6Q9C3B/p1739464533762049?thread_ts=1739462628.361019&cid=C08DE6Q9C3B
- improves https://github.com/neondatabase/neon/issues/10668
- part of https://github.com/neondatabase/cloud/issues/23515
# How It Works
The trouble is two-fold:
1. getting to the raw socket file descriptor through the many Rust types
that wrap it and
2. integrating with the `measure()` function
Rust wraps it in types to model file descriptor lifetimes and ownership,
and usually one can get access using `as_raw_fd()`.
However, we `split()` the stream and the resulting
[`tokio::io::WriteHalf`](https://docs.rs/tokio/latest/tokio/io/struct.WriteHalf.html)
.
Check the PR commit history for my attempts to do it.
My solution is to get the socket fd before we wrap it in our protocol
types, and to store that fd in the new `PostgresBackend::socket_fd`
field.
I believe it's safe because the lifetime of `PostgresBackend::socket_fd`
value == the lifetime of the `TcpStream` that wrap and store in
`PostgresBackend::framed`.
Specifically, the only place that close()s the socket is the `impl Drop
for TcpStream`.
I think the protocol stack calls `TcpStream::shutdown()`, but, that
doesn't `close()` the file descriptor underneath.
Regarding integration with the `measure()` function, the trouble is that
`flush_fut` is currently a generic `Future` type. So, we just pass in
the `socket_fd` as a separate argument.
A clean implementation would convert the `pgb_writer.flush()` to a named
future that provides an accessor for the socket fd while not being
polled.
I tried (see PR history), but failed to break through the `WriteHalf`.
# Testing
Tested locally by running
```
./target/debug/pagebench get-page-latest-lsn --num-clients=1000 --queue-depth=1000
```
in one terminal, waiting a bit, then
```
pkill -STOP pagebench
```
then wait for slow logs to show up in `pageserver.log`.
Pick one of the slow log message's port pairs, e.g., `127.0.0.1:39500`,
and then checking sockstat output
```
ss -ntp | grep '127.0.0.1:39500'
```
to ensure that send & recv queue size match those in the log message.
Avoids compiling the crate and its dependencies into binaries that don't
need them. Shrinks the compute_ctl binary from about 31MB to 28MB in the
release-line-debug-size-lto profile.
## Problem
The compaction loop currently runs periodically, which can cause it to
wait for up to 20 seconds before starting L0 compaction by default.
Also, when we later separate the semaphores for L0 compaction and image
compaction, we want to give up waiting for the image compaction
semaphore if L0 compaction is needed on any timeline.
Touches #10694.
## Summary of changes
Notify the compaction loop when an L0 flush (on any timeline) exceeds
`compaction_threshold`.
Also do some opportunistic cleanups in the area.
## Problem
Following #10641, let's add a few critical errors.
Resolves#10094.
## Summary of changes
Adds the following critical errors:
* WAL sender read/decode failure.
* WAL record ingestion failure.
* WAL redo failure.
* Missing key during compaction.
We don't add an error for missing keys during GetPage requests, since
we've seen a handful of these in production recently, and the cause is
still unclear (most likely a benign race).
## Problem
We don't currently have good alerts for critical errors, e.g. data
loss/corruption.
Touches #10094.
## Summary of changes
Add a `critical!` macro and corresponding
`libmetrics_tracing_event_count{level="critical"}` metric. This will:
* Emit an `ERROR` log message with prefix `"CRITICAL:"` and a backtrace.
* Increment `libmetrics_tracing_event_count{level="critical"}`, and
indirectly `level="error"`.
* Trigger a pageable alert (via the metric above).
* In debug builds, panic the process.
I'll add uses of the macro separately.
## Refs
- Epic: https://github.com/neondatabase/neon/issues/9378
Co-authored-by: Vlad Lazar <vlad@neon.tech>
Co-authored-by: Christian Schwarz <christian@neon.tech>
## Problem
The read path does its IOs sequentially.
This means that if N values need to be read to reconstruct a page,
we will do N IOs and getpage latency is `O(N*IoLatency)`.
## Solution
With this PR we gain the ability to issue IO concurrently within one
layer visit **and** to move on to the next layer without waiting for IOs
from the previous visit to complete.
This is an evolved version of the work done at the Lisbon hackathon,
cf https://github.com/neondatabase/neon/pull/9002.
## Design
### `will_init` now sourced from disk btree index keys
On the algorithmic level, the only change is that the
`get_values_reconstruct_data`
now sources `will_init` from the disk btree index key (which is
PS-page_cache'd), instead
of from the `Value`, which is only available after the IO completes.
### Concurrent IOs, Submission & Completion
To separate IO submission from waiting for its completion, while
simultaneously
feature-gating the change, we introduce the notion of an `IoConcurrency`
struct
through which IO futures are "spawned".
An IO is an opaque future, and waiting for completions is handled
through
`tokio::sync::oneshot` channels.
The oneshot Receiver's take the place of the `img` and `records` fields
inside `VectoredValueReconstructState`.
When we're done visiting all the layers and submitting all the IOs along
the way
we concurrently `collect_pending_ios` for each value, which means
for each value there is a future that awaits all the oneshot receivers
and then calls into walredo to reconstruct the page image.
Walredo is now invoked concurrently for each value instead of
sequentially.
Walredo itself remains unchanged.
The spawned IO futures are driven to completion by a sidecar tokio task
that
is separate from the task that performs all the layer visiting and
spawning of IOs.
That tasks receives the IO futures via an unbounded mpsc channel and
drives them to completion inside a `FuturedUnordered`.
(The behavior from before this PR is available through
`IoConcurrency::Sequential`,
which awaits the IO futures in place, without "spawning" or "submitting"
them
anywhere.)
#### Alternatives Explored
A few words on the rationale behind having a sidecar *task* and what
alternatives were considered.
One option is to queue up all IO futures in a FuturesUnordered that is
polled
the first time when we `collect_pending_ios`.
Firstly, the IO futures are opaque, compiler-generated futures that need
to be polled at least once to submit their IO. "At least once" because
tokio-epoll-uring may not be able to submit the IO to the kernel on
first
poll right away.
Second, there are deadlocks if we don't drive the IO futures to
completion
independently of the spawning task.
The reason is that both the IO futures and the spawning task may hold
some
_and_ try to acquire _more_ shared limited resources.
For example, both spawning task and IO future may try to acquire
* a VirtualFile file descriptor cache slot async mutex (observed during
impl)
* a tokio-epoll-uring submission slot (observed during impl)
* a PageCache slot (currently this is not the case but we may move more
code into the IO futures in the future)
Another option is to spawn a short-lived `tokio::task` for each IO
future.
We implemented and benchmarked it during development, but found little
throughput improvement and moderate mean & tail latency degradation.
Concerns about pressure on the tokio scheduler made us discard this
variant.
The sidecar task could be obsoleted if the IOs were not arbitrary code
but a well-defined struct.
However,
1. the opaque futures approach taken in this PR allows leaving the
existing
code unchanged, which
2. allows us to implement the `IoConcurrency::Sequential` mode for
feature-gating
the change.
Once the new mode sidecar task implementation is rolled out everywhere,
and `::Sequential` removed, we can think about a descriptive submission
& completion interface.
The problems around deadlocks pointed out earlier will need to be solved
then.
For example, we could eliminate VirtualFile file descriptor cache and
tokio-epoll-uring slots.
The latter has been drafted in
https://github.com/neondatabase/tokio-epoll-uring/pull/63.
See the lengthy doc comment on `spawn_io()` for more details.
### Error handling
There are two error classes during reconstruct data retrieval:
* traversal errors: index lookup, move to next layer, and the like
* value read IO errors
A traversal error fails the entire get_vectored request, as before this
PR.
A value read error only fails that value.
In any case, we preserve the existing behavior that once
`get_vectored` returns, all IOs are done. Panics and failing
to poll `get_vectored` to completion will leave the IOs dangling,
which is safe but shouldn't happen, and so, a rate-limited
log statement will be emitted at warning level.
There is a doc comment on `collect_pending_ios` giving more code-level
details and rationale.
### Feature Gating
The new behavior is opt-in via pageserver config.
The `Sequential` mode is the default.
The only significant change in `Sequential` mode compared to before
this PR is the buffering of results in the `oneshot`s.
## Code-Level Changes
Prep work:
* Make `GateGuard` clonable.
Core Feature:
* Traversal code: track `will_init` in `BlobMeta` and source it from
the Delta/Image/InMemory layer index, instead of determining `will_init`
after we've read the value. This avoids having to read the value to
determine whether traversal can stop.
* Introduce `IoConcurrency` & its sidecar task.
* `IoConcurrency` is the clonable handle.
* It connects to the sidecar task via an `mpsc`.
* Plumb through `IoConcurrency` from high level code to the
individual layer implementations' `get_values_reconstruct_data`.
We piggy-back on the `ValuesReconstructState` for this.
* The sidecar task should be long-lived, so, `IoConcurrency` needs
to be rooted up "high" in the call stack.
* Roots as of this PR:
* `page_service`: outside of pagestream loop
* `create_image_layers`: when it is called
* `basebackup`(only auxfiles + replorigin + SLRU segments)
* Code with no roots that uses `IoConcurrency::sequential`
* any `Timeline::get` call
* `collect_keyspace` is a good example
* follow-up: https://github.com/neondatabase/neon/issues/10460
* `TimelineAdaptor` code used by the compaction simulator, unused in
practive
* `ingest_xlog_dbase_create`
* Transform Delta/Image/InMemoryLayer to
* do their values IO in a distinct `async {}` block
* extend the residence of the Delta/Image layer until the IO is done
* buffer their results in a `oneshot` channel instead of straight
in `ValuesReconstructState`
* the `oneshot` channel is wrapped in `OnDiskValueIo` /
`OnDiskValueIoWaiter`
types that aid in expressiveness and are used to keep track of
in-flight IOs so we can print warnings if we leave them dangling.
* Change `ValuesReconstructState` to hold the receiving end of the
`oneshot` channel aka `OnDiskValueIoWaiter`.
* Change `get_vectored_impl` to `collect_pending_ios` and issue walredo
concurrently, in a `FuturesUnordered`.
Testing / Benchmarking:
* Support queue-depth in pagebench for manual benchmarkinng.
* Add test suite support for setting concurrency mode ps config
field via a) an env var and b) via NeonEnvBuilder.
* Hacky helper to have sidecar-based IoConcurrency in tests.
This will be cleaned up later.
More benchmarking will happen post-merge in nightly benchmarks, plus in
staging/pre-prod.
Some intermediate helpers for manual benchmarking have been preserved in
https://github.com/neondatabase/neon/pull/10466 and will be landed in
later PRs.
(L0 layer stack generator!)
Drive-By:
* test suite actually didn't enable batching by default because
`config.compatibility_neon_binpath` is always Truthy in our CI
environment
=> https://neondb.slack.com/archives/C059ZC138NR/p1737490501941309
* initial logical size calculation wasn't always polled to completion,
which was
surfaced through the added WARN logs emitted when dropping a
`ValuesReconstructState` that still has inflight IOs.
* remove the timing histograms
`pageserver_getpage_get_reconstruct_data_seconds`
and `pageserver_getpage_reconstruct_seconds` because with planning,
value read
IO, and walredo happening concurrently, one can no longer attribute
latency
to any one of them; we'll revisit this when Vlad's work on
tracing/sampling
through RequestContext lands.
* remove code related to `get_cached_lsn()`.
The logic around this has been dead at runtime for a long time,
ever since the removal of the materialized page cache in #8105.
## Testing
Unit tests use the sidecar task by default and run both modes in CI.
Python regression tests and benchmarks also use the sidecar task by
default.
We'll test more in staging and possibly preprod.
# Future Work
Please refer to the parent epic for the full plan.
The next step will be to fold the plumbing of IoConcurrency
into RequestContext so that the function signatures get cleaned up.
Once `Sequential` isn't used anymore, we can take the next
big leap which is replacing the opaque IOs with structs
that have well-defined semantics.
---------
Co-authored-by: Christian Schwarz <christian@neon.tech>
# Refs
- extracted from https://github.com/neondatabase/neon/pull/9353
# Problem
Before this PR, when task_mgr shutdown is signalled, e.g. during
pageserver shutdown or Tenant shutdown, initial logical size calculation
stops polling and drops the future that represents the calculation.
This is against the current policy that we poll all futures to
completion.
This became apparent during development of concurrent IO which warns if
we drop a `Timeline::get_vectored` future that still has in-flight IOs.
We may revise the policy in the future, but, right now initial logical
size calculation is the only part of the codebase that doesn't adhere to
the policy, so let's fix it.
## Code Changes
- make sensitive exclusively to `Timeline::cancel`
- This should be sufficient for all cases of shutdowns; the sensitivity
to task_mgr shutdown is unnecessary.
- this broke the various cancel tests in `test_timeline_size.py`, e.g.,
`test_timeline_initial_logical_size_calculation_cancellation`
- the tests would time out because the await point was not sensitive to
cancellation
- to fix this, refactor `pausable_failpoint` so that it accepts a
cancellation token
- side note: we _really_ should write our own failpoint library; maybe
after we get heap-allocated RequestContext, we can plumb failpoints
through there.
## Problem
gc-compaction needs the partitioning data to decide the job split. This
refactor allows concurrent access/computing the partitioning.
## Summary of changes
Make `partitioning` an ArcSwap so that others can access the
partitioning while we compute it. Fully eliminate the `repartition is
called concurrently` warning when gc-compaction is going on.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
Taking continuous profiles every 20 seconds is likely too expensive (in
dollar terms). Let's try 60-second profiles. We can now interrupt
running profiles via `?force=true`, so this should be fine.
With a new beta build of the rust compiler, it's good to check out the
new lints. Either to find false positives, or find flaws in our code.
Additionally, it helps reduce the effort required to update to 1.85 in 6
weeks.
## Problem
It's only possible to take one CPU profile at a time. With Grafana
continuous profiling, a (low-frequency) CPU profile will always be
running, making it hard to take an ad hoc CPU profile at the same time.
Resolves#10072.
## Summary of changes
Add a `force` parameter for `/profile/cpu` which will end and return an
already running CPU profile, starting a new one for the current caller.
## Problem
Unlike CPU profiles, the `/profile/heap` endpoint can't automatically
generate SVG flamegraphs. This requires the user to install and use
`pprof` tooling, which is unnecessary and annoying.
Resolves#10203.
## Summary of changes
Add `format=svg` for the `/profile/heap` route, and generate an SVG
flamegraph using the `inferno` crate, similarly to what `pprof-rs`
already does for CPU profiles.
# Problem
Before this PR, there were cases where send() in state
SenderWaitsForReceiverToConsume would never be woken up
by the receiver, because it never registered with `wake_sender`.
Example Scenario 1: we stop polling a send() future A that was waiting
for the receiver to consume. We drop A and create a new send() future B.
B would return Poll::Pending and never regsister a waker.
Example Scenario 2: a send() future A transitions from HasData
to SenderWaitsForReceiverToConsume. This registers the context X
with `wake_sender`. But before the Receiver consumes the data,
we poll A from a different context Y.
The state is still SenderWaitsForReceiverToConsume, but we wouldn't
register the new context with `wake_sender`.
When the Receiver comes around to consume and `wake_sender.notify()`s,
it wakes the old context X instead of Y.
# Fix
Register the waker in the case where we're polled in
state `SenderWaitsForReceiverToConsume`.
# Relation to #10309
I found this bug while investigating #10309.
There was never proof that this bug here is the root cause for #10309.
In the meantime we found a more probably hypothesis
for the root cause than what is being fixed here.
Regardless, let's walk through my thought process about
how it might have been relevant:
There (in page_service), Scenario 1 does not apply because
we poll the send() future to completion.
Scenario 2 (`tokio::join!`) also does not apply with the
current `tokio::join!()` impl, because it will just poll each
future every time, each with the same context.
Although if we ever used something like a FuturesUnordered anywhere,
that will be using a different context, so, in that case,
the bug might materialize.
Regarding tokio & spurious poll in general:
@conradludgate is not aware of any spurious wakeup cases in current
tokio,
but within a `tokio::join!()`, any wake meant for one future will poll
all
the futures, so that can appear as a spurious wake up to the N-1 futures
of the `tokio::join!()`.
## Problem
Jemalloc heap profiles aren't symbolized. This is inconvenient, and
doesn't work with Grafana Cloud Profiles.
Resolves#9964.
## Summary of changes
Symbolize the heap profiles in-process, and strip unnecessary cruft.
This uses about 100 MB additional memory to cache the DWARF information,
but I believe this is already the case with CPU profiles, which use the
same library for symbolization. With cached DWARF information, the
symbolization CPU overhead is negligible.
Example profiles:
*
[pageserver.pb.gz](https://github.com/user-attachments/files/18141395/pageserver.pb.gz)
*
[safekeeper.pb.gz](https://github.com/user-attachments/files/18141396/safekeeper.pb.gz)
## Problem
Cplane and storage controller tenant config changes are not additive.
Any change overrides all existing tenant configs. This would be fine if
both did client side patching, but that's not the case.
Once this merges, we must update cplane to use the PATCH endpoint.
## Summary of changes
### High Level
Allow for patching of tenant configuration with a `PATCH
/v1/tenant/config` endpoint.
It takes the same data as it's PUT counterpart. For example the payload
below will update `gc_period` and unset `compaction_period`. All other
fields are left in their original state.
```
{
"tenant_id": "1234",
"gc_period": "10s",
"compaction_period": null
}
```
### Low Level
* PS and storcon gain `PATCH /v1/tenant/config` endpoints. PS endpoint
is only used for cplane managed instances.
* `storcon_cli` is updated to have separate commands for
`set-tenant-config` and `patch-tenant-config`
Related https://github.com/neondatabase/cloud/issues/21043
We've seen cases where stray keys end up on the wrong shard. This
shouldn't happen. Add debug assertions to prevent this. In release
builds, we should be lenient in order to handle changing key ownership
policies.
Touches #9914.
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 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 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
https://github.com/neondatabase/neon/pull/9746 lifted decoding and
interpretation of WAL to the safekeeper.
This reduced the ingested amount on the pageservers by around 10x for a
tenant with 8 shards, but doubled
the ingested amount for single sharded tenants.
Also, https://github.com/neondatabase/neon/pull/9746 uses bincode which
doesn't support schema evolution.
Technically the schema can be evolved, but it's very cumbersome.
## Summary of changes
This patch set addresses both problems by adding protobuf support for
the interpreted wal records and adding compression support. Compressed
protobuf reduced the ingested amount by 100x on the 32 shards
`test_sharded_ingest` case (compared to non-interpreted proto). For the
1 shard case the reduction is 5x.
Sister change to `rust-postgres` is
[here](https://github.com/neondatabase/rust-postgres/pull/33).
## Links
Related: https://github.com/neondatabase/neon/issues/9336
Epic: https://github.com/neondatabase/neon/issues/9329
## Problem
For any given tenant shard, pageservers receive all of the tenant's WAL
from the safekeeper.
This soft-blocks us from using larger shard counts due to bandwidth
concerns and CPU overhead of filtering
out the records.
## Summary of changes
This PR lifts the decoding and interpretation of WAL from the pageserver
into the safekeeper.
A customised PG replication protocol is used where instead of sending
raw WAL, the safekeeper sends
filtered, interpreted records. The receiver drives the protocol
selection, so, on the pageserver side, usage
of the new protocol is gated by a new pageserver config:
`wal_receiver_protocol`.
More granularly the changes are:
1. Optionally inject the protocol and shard identity into the arguments
used for starting replication
2. On the safekeeper side, implement a new wal sending primitive which
decodes and interprets records
before sending them over
3. On the pageserver side, implement the ingestion of this new
replication message type. It's very similar
to what we already have for raw wal (minus decoding and interpreting).
## Notes
* This PR currently uses my [branch of
rust-postgres](https://github.com/neondatabase/rust-postgres/tree/vlad/interpreted-wal-record-replication-support)
which includes the deserialization logic for the new replication message
type. PR for that is open
[here](https://github.com/neondatabase/rust-postgres/pull/32).
* This PR contains changes for both pageservers and safekeepers. It's
safe to merge because the new protocol is disabled by default on the
pageserver side. We can gradually start enabling it in subsequent
releases.
* CI tests are running on https://github.com/neondatabase/neon/pull/9747
## Links
Related: https://github.com/neondatabase/neon/issues/9336
Epic: https://github.com/neondatabase/neon/issues/9329
The notifications need to be sent whenever the waiters heap changes, per
the comment in `update_status`. But if 'advance' is called when there
are no waiters, or the new LSN is lower than the waiters so that no one
needs to be woken up, there's no need to send notifications. This saves
some CPU cycles in the common case that there are no waiters.
## Problem
We don't have a convenient way to gather CPU profiles from a running
binary, e.g. during production incidents or end-to-end benchmarks, nor
during microbenchmarks (particularly on macOS).
We would also like to have continuous profiling in production, likely
using [Grafana Cloud
Profiles](https://grafana.com/products/cloud/profiles-for-continuous-profiling/).
We may choose to use either eBPF profiles or pprof profiles for this
(pending testing and discussion with SREs), but pprof profiles appear
useful regardless for the reasons listed above. 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 #9534.
Touches https://github.com/neondatabase/cloud/issues/14888.
## Summary of changes
Adds a HTTP route `/profile/cpu` that takes a CPU profile and returns
it. Defaults to a 5-second pprof Protobuf profile for use with e.g.
`pprof` or Grafana Alloy, but can also emit an SVG flamegraph. Query
parameters:
* `format`: output format (`pprof` or `svg`)
* `frequency`: sampling frequency in microseconds (default 100)
* `seconds`: number of seconds to profile (default 5)
Also integrates pprof profiles into Criterion benchmarks, such that
flamegraph reports can be taken with `cargo bench ... --profile-duration
<seconds>`. Output under `target/criterion/*/profile/flamegraph.svg`.
Example profiles:
* pprof profile (use [`pprof`](https://github.com/google/pprof)):
[profile.pb.gz](https://github.com/user-attachments/files/17756788/profile.pb.gz)
* Web interface: `pprof -http :6060 profile.pb.gz`
* Interactive flamegraph:
[profile.svg.gz](https://github.com/user-attachments/files/17756782/profile.svg.gz)
part of https://github.com/neondatabase/neon/issues/9114, we want to be
able to run partial gc-compaction in tests. In the future, we can also
expand this functionality to legacy compaction, so that we can trigger
compaction for a specific key range.
## Summary of changes
* Support passing compaction key range through pageserver routes.
* Refactor input parameters of compact related function to take the new
`CompactOptions`.
* Add tests for partial compaction. Note that the test may or may not
trigger compaction based on GC horizon. We need to improve the test case
to ensure things always get below the gc_horizon and the gc-compaction
can be triggered.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
