Compare commits

..

122 Commits

Author SHA1 Message Date
Jan Christian Grünhage
313e01098a no merge: intentionally break benchmarks for verifying CI 2025-04-25 11:29:33 +02:00
Alexander Bayandin
5e989a3148 CI(build-tools): bump packages in build-tools image (#11697)
## Problem

`cargo-deny` 0.16.2 spits a bunch of warnings like:
```
warning[index-failure]: unable to check for yanked crates
```

The issue is fixed for the latest version of `cargo-deny` (0.18.2). And
while we're here, let's bump all the packages we have in `build-tools`
image

## Summary of changes
- bump cargo-hakari to 0.9.36
- bump cargo-deny to 0.18.2
- bump cargo-hack to 0.6.36
- bump cargo-nextest to 0.9.94
- bump diesel_cli to 2.2.9
- bump s5cmd to 2.3.0
- bump mold to 2.37.1
- bump python to 3.11.12
2025-04-24 14:13:04 +00:00
Alexey Kondratov
985056be37 feat(compute): Introduce Postgres downtime metrics (#11346)
## Problem

Currently, we only report the timestamp of the last moment we think
Postgres was active. The problem is that if Postgres gets completely
unresponsive, we still report some old timestamp, and it's impossible to
distinguish situations 'Postgres is effectively down' and 'Postgres is
running, but no client activity'.

## Summary of changes

Refactor the `compute_ctl`'s compute monitor so that it was easier to
track the connection errors and failed activity checks, and report
- `now() - last_successful_check` as current downtime on any failure
- cumulative Postgres downtime during the whole compute lifetime

After adding a test, I also noticed that the compute monitor may not
reconnect even though queries fail with `connection closed` or `error
communicating with the server: Connection reset by peer (os error 54)`,
but for some reason we do not catch it with `client.is_closed()`, so I
added an explicit reconnect in case of any failures.

Discussion:
https://neondb.slack.com/archives/C03TN5G758R/p1742489426966639
2025-04-24 13:51:09 +00:00
Christian Schwarz
9c6ff3aa2b refactor(BufferedWriter): flush task owns the VirtualFile & abstraction for cleanup on drop (#11549)
Main change:

- `BufferedWriter` owns the `W`; no more `Arc<W>`
- We introduce auto-delete-on-drop wrappers for `VirtualFile`.
  - `TempVirtualFile` for write-only users
- `TempVirtualFileCoOwnedByEphemeralFileAndBufferedWriter` for
EphemeralFile which requires read access to the immutable prefix of the
file (see doc comments for details)
- Users of `BufferedWriter` hand it such a wrapped `VirtualFile`.
- The wrapped `VirtualFile` moves to the background flush task.
- On `BufferedWriter` shutdown, ownership moves back.
- Callers remove the wrapper (`disarm_into_inner()`) after doing final
touches, e.g., flushing index blocks and summary for delta/image layer
writers.

If the BufferedWriter isn't shut down properly via
`BufferedWriter::shutdown`, or if there is an error during final
touches, the wrapper type ensures that the file gets unlinked.

We store a GateGuard inside the wrapper to ensure that the Timeline is
still alive when unlinking on drop.

Rust doesn't have async drop yet, so, the unlinking happens using a
synchronous syscall.
NB we don't fsync the surrounding directory.
This is how it's been before this PR; I believe it is correct because
all of these files are temporary paths that get cleaned up on timeline
load.
Again, timeline load does not need to fsync because the next timeline
load will unlink again if the file reappears.

The auto-delete-on-drop can happen after a higher-level mechanism
retries.
Therefore, we switch all users to monotonically increasing, never-reused
temp file disambiguators.

The aspects pointed out in the last two paragraphs will receive further
cleanup in follow-up task
- https://github.com/neondatabase/neon/issues/11692

Drive-by changes:
- It turns out we can remove the two-pronged code in the layer file
download code.
No need to make this a separate PR because all of production already
uses `tokio-epoll-uring` with the buffered writer for many weeks.


Refs
- epic https://github.com/neondatabase/neon/issues/9868
- alternative to https://github.com/neondatabase/neon/pull/11544
2025-04-24 13:07:57 +00:00
Folke Behrens
9d472c79ce Fix what's currently flagged by cargo deny (#11693)
* Replace yanked papaya version
* Remove unused allowed license: OpenSSL
* Remove Zlib license from general allow list since it's listed in the
exceptions section per crate
* Drop clarification for ring since they have separate LICENSE files now
* List the tower-otel repo as allowed source while we sort out the OTel
deps
2025-04-24 13:02:31 +00:00
Arpad Müller
b43203928f Switch tenant snapshot subcommand to remote_storage (#11685)
Switches the tenant snapshot subcommand of the storage scrubber to
`remote_storage`. As this is the last piece of the storage scrubber
still using the S3 SDK, this finishes the project started in #7547.

This allows us to do tenant snapshots on Azure as well.

Builds on #11671
Fixes #8830
2025-04-24 12:22:07 +00:00
Arpad Müller
c35d489539 versioning API for remote_storage (#11671)
Adds a versioning API to remote_storage. We want to use it in the
scrubber, both for tenant snapshot as well as for metadata checks.

for #8830
and for #11588
2025-04-24 11:41:48 +00:00
Vlad Lazar
3a50d95b6d storage_controller: coordinate imports across shards in the storage controller (#11345)
## Problem

Pageservers notify control plane directly when a shard import has
completed.
Control plane has to download the status of each shard from S3 and
figure out if everything is truly done,
before proceeding with branch activation.

Issues with this approach are:
* We can't control shard split behaviour on the storage controller side.
It's unsafe to split
during import.
* Control plane needs to know about shards and implement logic to check
all timelines are indeed ready.

## Summary of changes

In short, storage controller coordinates imports, and, only when
everything is done, notifies control plane.

Big rocks:
1. Store timeline imports in the storage controller database. Each
import stores the status of its shards in the database.
We hook into the timeline creation call as our entry point for this.
2. Pageservers get a new upcall endpoint to notify the storage
controller of shard import updates.
3. Storage controller handles these updates by updating persisted state.
If an update finalizes the import,
then poll pageservers until timeline activation, and, then, notify the
control plane that the import is complete.

Cplane side change with new endpoint is in
https://github.com/neondatabase/cloud/pull/26166

Closes https://github.com/neondatabase/neon/issues/11566
2025-04-24 11:26:06 +00:00
Arpad Müller
d43b8e73ae Update sentry to 0.37 (#11686)
Update the sentry crate to 0.37. This deduplicates the `webpki-roots`
crate in our crate graph, and brings another dependency onto newer
rustls `0.23.18`.
2025-04-24 11:20:41 +00:00
devin-ai-integration[bot]
1808dad269 Add --dev CLI flag to pageserver and safekeeper binaries (#11526)
# Add --dev CLI flag to pageserver and safekeeper binaries

This PR adds the `--dev` CLI flag to both the pageserver and safekeeper
binaries without implementing any functionality yet. This is a precursor
to PR #11517, which will implement the full functionality to require
authentication by default unless the `--dev` flag is specified.

## Changes
- Add `dev_mode` config field to pageserver binary
- Add `--dev` CLI flag to safekeeper binary

This PR is needed for forward compatibility tests to work properly, when
we try to merge #11517

Link to Devin run:
https://app.devin.ai/sessions/ad8231b4e2be430398072b6fc4e85d46
Requested by: John Spray (john@neon.tech)

---------

Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: John Spray <john@neon.tech>
2025-04-24 10:45:40 +00:00
Folke Behrens
7ba8519b43 proxy: Update zerocopy to 0.8 (#11681)
Also add some macros that might result in more efficient code.
2025-04-24 09:39:08 +00:00
Christian Schwarz
f8100d66d5 ci: extend 'Wait for extension build to finish' timeout (#11689)
Refs
- https://neondb.slack.com/archives/C059ZC138NR/p1745427571307149
2025-04-24 08:15:08 +00:00
Christian Schwarz
51cdb570eb bench_ingest: general overhaul & add parametrization over virtual_file_io_mode (#11667)
Changes:
- clean up existing parametrization & criterion `BenchmarkId`
- additional parametrization over `virtual_file_io_mode`
- switch to `multi_thread` to be closer to production ([Slack
thread](https://neondb.slack.com/archives/C033RQ5SPDH/p1745339543093159))

Refs
- epic https://github.com/neondatabase/neon/issues/9868
- extracted from https://github.com/neondatabase/neon/pull/11558
2025-04-24 07:38:18 +00:00
devin-ai-integration[bot]
8e09ecf2ab Fix KeyError in physical replication benchmark test (#11675)
# Fix KeyError in physical replication benchmark test

This PR fixes the failing physical replication benchmark test that was
encountering a KeyError: 'endpoints'.

The issue was in accessing `project["project"]["endpoints"][0]["id"]`
when it should be `project["endpoints"][0]["id"]`, consistent with how
endpoints are accessed elsewhere in the codebase.

Fixed the issue in both test functions:
- test_ro_replica_lag
- test_replication_start_stop

Link to Devin run:
https://app.devin.ai/sessions/be3fe9a9ee5942e4b12e74a7055f541b
Requested by: Peter Bendel

Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: peterbendel@neon.tech <peterbendel@neon.tech>
2025-04-23 14:51:08 +00:00
Mikhail Kot
c3534cea39 Rename object_storage->endpoint_storage (#11678)
1. Rename service to avoid ambiguity as discussed in Slack
2. Ignore endpoint_id in read paths as requested in
https://github.com/neondatabase/cloud/issues/26346#issuecomment-2806758224
2025-04-23 14:03:19 +00:00
Folke Behrens
21d3d60cef proxy/pglb: Add in-process connection support (#11677)
Define a `Connection` and a `Stream` type that resemble simple QUIC
connections
and (multiplexed) streams.
2025-04-23 12:18:30 +00:00
Tristan Partin
b00db536bb Add CPU architecture to the remote extensions object key (#11590)
ARM computes are incoming and we need to account for that in remote
extensions. Previously, we just blindly assumed that all computes were
x86_64.

Note that we use the Go architecture naming convention instead of the
Rust one directly to do our best and be consistent across the stack.

Part-of: https://github.com/neondatabase/cloud/issues/23148

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-22 22:47:22 +00:00
Arpad Müller
149cbd1e0a Support single and two safekeeper scenarios (#11483)
In tests and when one safekeeper is down in small regions, we need to
contend with one or two safekeepers. Before, we gave an error in
`safekeepers_for_new_timeline`. Now we just silently allow the timeline
to be created on one or two safekeepers.

Part of #9011
2025-04-22 21:27:01 +00:00
Alexander Lakhin
7b949daf13 fix(test): allow reconcile errors in test_storage_controller_heartbeats (#11665)
## Problem

test_storage_controller_heartbeats is flaky because of unallowed
reconciler errors (#11625)

## Summary of changes

Allow reconcile errors as in other tests in test_storage_controller.py.
2025-04-22 18:13:16 +00:00
Konstantin Knizhnik
132b6154bb Unlogged build debug compare local v2 (#11554)
## Problem

Init fork is used in DEBUG_COMPARE_LOCAL to determine unlogged relation
or unlogged build.
But it is created only after the relation is initialized and so can be
swapped out, producing `Page is evicted with zero LSN` error.

## Summary of changes

Create init fork together with main fork for unlogged relations in
DEBUG_COMPARE_LOCAL mode.

---------

Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
2025-04-22 18:07:45 +00:00
Alex Chi Z.
ad3519ebcb fix(pageserver): report synthetic size = 1 if all tls offloaded (#11648)
## Problem

A quick workaround for https://github.com/neondatabase/neon/issues/11631

## Summary of changes

Report synthetic size == 1 if all timelines are offloaded.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-22 14:28:22 +00:00
Dmitrii Kovalkov
6173c0f44c safekeeper: add enable_tls_wal_service_api (#11520)
## Problem
Safekeeper doesn't use TLS in wal service
- Closes: https://github.com/neondatabase/cloud/issues/27302

## Summary of changes
- Add `enable_tls_wal_service_api` option to safekeeper's cmd arguments
- Propagate `tls_server_config` to `wal_service` if the option is
enabled
- Create `BACKGROUND_RUNTIME` for small background tasks and offload SSL
certificate reloader to it.

No integration tests for now because support from compute side is
required: https://github.com/neondatabase/cloud/issues/25823
2025-04-22 13:19:03 +00:00
a-masterov
fd916abf25 Remove NOTICE messages, which can make the pg_repack regression test fail. (#11659)
## Problem
The pg_repack test can be flaky due to unpredictable `NOTICE` messages
about waiting for some processes.
E.g., 
```
 INFO: repacking table "public.issue3_2"
+NOTICE: Waiting for 1 transactions to finish. First PID: 427
```
## Summary of changes
The `client_min_messages` set to `warning` for the regression tests.
2025-04-22 11:43:45 +00:00
Alexander Bayandin
cd2e1fbc7c CI(benchmarks): upload perf results for passed tests (#11649)
## Problem

We run benchmarks in batches (five parallel jobs on different runners).
If any test in a batch fails, we won’t upload any results for that
batch, even for the tests that passed.

## Summary of changes
- Move the results upload to a separate step in the run-python-test-set
action, and execute this step even if tests fail.
2025-04-22 09:41:28 +00:00
Tristan Partin
5df4a747e6 Update pgbouncer in compute images to 1.24.1 (#11651)
Fixes CVE-2025-2291.

Link:
https://www.postgresql.org/about/news/pgbouncer-1241-released-fixes-cve-2025-2291-3059/

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-21 17:49:17 +00:00
Vlad Lazar
cbf442292b pageserver: handle empty get vectored queries (#11652)
## Problem

If all batched requests are excluded from the query by
`Timeine::get_rel_page_at_lsn_batched` (e.g. because they are past the
end of the relation), the read path would panic since it doesn't expect
empty queries. This is a change in behaviour that was introduced with
the scattered query implementation.

## Summary of Changes

Handle empty queries explicitly.
2025-04-21 17:45:16 +00:00
Heikki Linnakangas
4d0c1e8b78 refactor: Extract some code in pagebench getpage command to function (#11563)
This makes it easier to add a different client implementation alongside
the current one. I started working on a new gRPC-based protocol to
replace the libpq protocol, which will introduce a new function like
`client_libpq`, but for the new protocol.

It's a little more readable with less indentation anyway.
2025-04-19 08:38:03 +00:00
JC Grünhage
3158442a59 fix(ci): set token for fast-forward failure comments and allow merging with state unstable (#11647)
## Problem

https://github.com/neondatabase/neon/actions/runs/14538136318/job/40790985693?pr=11645
failed, even though the relevant parts of the CI had passed and
auto-merge determined the PR is ready to merge. After that, commenting
failed.

## Summary of changes
- set GH_TOKEN for commenting after fast-forward failure
- allow merging with mergeable_state unstable
2025-04-18 17:49:34 +00:00
JC Grünhage
f006879fb7 fix(ci): make regex to find rc branches less strict (#11646)
## Problem

https://github.com/neondatabase/neon/actions/runs/14537161022/job/40787763965
failed to find the correct RC PR run, preventing artifact re-use. This
broke in https://github.com/neondatabase/neon/pull/11547.

There's a hotfix release containing this in
https://github.com/neondatabase/neon/pull/11645.

## Summary of changes
Make the regex for finding the RC PR run less strict, it was needlessly
precise.
2025-04-18 16:39:18 +00:00
Dmitrii Kovalkov
a0d844dfed pageserver + safekeeper: pass ssl ca certs to broker client (#11635)
## Problem
Pageservers and safakeepers do not pass CA certificates to broker
client, so the client do not trust locally issued certificates.
- Part of https://github.com/neondatabase/cloud/issues/27492

## Summary of changes
- Change `ssl_ca_certs` type in PS/SK's config to `Pem` which may be
converted to both `reqwest` and `tonic` certificates.
- Pass CA certificates to storage broker client in PS and SK
2025-04-18 06:27:23 +00:00
Alex Chi Z.
5073e46df4 feat(pageserver): use rfc3339 time and print ratio in gc-compact stats (#11638)
## Problem

follow-up on https://github.com/neondatabase/neon/pull/11601

## Summary of changes

- serialize the start/end time using rfc3339 time string
- compute the size ratio of the compaction

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-18 05:28:01 +00:00
Alexander Bayandin
182bd95a4e CI(regress-tests): run tests on large-metal (#11634)
## Problem

Regression tests are more flaky on virtualised (`qemu-x64-*`) runners

See https://neondb.slack.com/archives/C069Z2199DL/p1744891865307769
Ref https://github.com/neondatabase/neon/issues/11627

## Summary of changes
- Switch `regress-tests` to metal-only large runners to mitigate flaky
behaviour
2025-04-18 01:25:38 +00:00
Anastasia Lubennikova
ce7795a67d compute: use project_id, endpoint_id as tag (#11556)
for compute audit logs

part of https://github.com/neondatabase/cloud/issues/21955
2025-04-17 23:32:38 +00:00
Suhas Thalanki
134d01c771 remove pg_anon.patch (#11636)
This PR removes `pg_anon.patch` as the `anon` v1 extension has been
removed and the patch is not being used anywhere
2025-04-17 22:08:16 +00:00
Arpad Müller
c1e4befd56 Additional fixes and improvements to storcon safekeeper timelines (#11477)
This delivers some additional fixes and improvements to storcon managed
safekeeper timelines:

* use `i32::MAX` for the generation number of timeline deletion
* start the generation for new timelines at 1 instead of 0: this ensures
that the other components actually are generation enabled
* fix database operations we use for metrics
* use join in list_pending_ops to prevent the classical ORM issue where
one does many db queries
* use enums in `test_storcon_create_delete_sk_down`. we are adding a
second parameter, and having two bool parameters is weird.
* extend `test_storcon_create_delete_sk_down` with a test of whole
tenant deletion. this hasn't been tested before.
* remove some redundant logging contexts
* Don't require mutable access to the service lock for scheduling
pending ops in memory. In order to pull this off, create reconcilers
eagerly. The advantage is that we don't need mutable access to the
service lock that way any more.

Part of #9011

---------

Co-authored-by: Arseny Sher <sher-ars@yandex.ru>
2025-04-17 20:25:30 +00:00
a-masterov
6c2e5c044c random operations test (#10986)
## Problem
We need to test the stability of Neon.

## Summary of changes
The test runs random operations on a Neon project. It performs via the
Public API calls the following operations: `create a branch`, `delete a
branch`, `add a read-only endpoint`, `delete a read-only endpoint`,
`restore a branch to a random position in the past`. All the branches
and endpoints are loaded with `pgbench`.

---------

Co-authored-by: Peter Bendel <peterbendel@neon.tech>
Co-authored-by: Alexander Bayandin <alexander@neon.tech>
2025-04-17 19:59:35 +00:00
Alex Chi Z.
748539b222 fix(pageserver): lower L0 compaction threshold (#11617)
## Problem

We saw OOMs due to L0 compaction happening simultaneously for all shards
of the same tenant right after the shard split.

## Summary of changes

Lower the threshold so that we compact fewer files.

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-17 19:51:28 +00:00
Alex Chi Z.
ad0c5fdae7 fix(test): allow stale generation warnings in storcon (#11624)
## Problem

https://github.com/neondatabase/neon/pull/11531 did not fully fix the
problem because the warning is part of the storcon instead of
pageserver.

## Summary of changes

Allow stale generation error in storcon.

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-17 16:12:24 +00:00
Christian Schwarz
2b041964b3 cover direct IO + concurrent IO in unit, regression & perf tests (#11585)
This mirrors the production config.

Thread that discusses the merits of this:
- https://neondb.slack.com/archives/C033RQ5SPDH/p1744742010740569

# Refs
- context
https://neondb.slack.com/archives/C04BLQ4LW7K/p1744724844844589?thread_ts=1744705831.014169&cid=C04BLQ4LW7K
- prep for https://github.com/neondatabase/neon/pull/11558 which adds
new io mode `direct-rw`

# Impact on CI turnaround time

Spot-checking impact on CI timings

- Baseline: [some recent main
commit](https://github.com/neondatabase/neon/actions/runs/14471549758/job/40587837475)
- Comparison: [this
commit](https://github.com/neondatabase/neon/actions/runs/14471945087/job/40589613274)
in this PR here

Impact on CI turnaround time

- Regression tests:
  - x64: very minor, sometimes better; likely in the noise
  - arm64: substantial  30min => 40min
- Benchmarks (x86 only I think): very minor; noise seems higher than
regress tests

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
Co-authored-by: Alex Chi Z. <4198311+skyzh@users.noreply.github.com>
Co-authored-by: Peter Bendel <peterbendel@neon.tech>
Co-authored-by: Alex Chi Z <chi@neon.tech>
2025-04-17 15:53:10 +00:00
John Spray
d4c059a884 tests: use endpoint http wrapper to get auth (#11628)
## Problem

`test_compute_startup_simple` and `test_compute_ondemand_slru_startup`
are failing.

This test implicitly asserts that the metrics.json endpoint succeeds and
returns all expected metrics, but doesn't make it easy to see what went
wrong if it doesn't (e.g. in this failure
https://neon-github-public-dev.s3.amazonaws.com/reports/main/14513210240/index.html#suites/13d8e764c394daadbad415a08454c04e/b0f92a86b2ed309f/)

In this case, it was failing because of a missing auth token, because it
was using `requests` directly instead of using the endpoint http client
type.

## Summary of changes

- Use endpoint http wrapper to get raise_for_status & auth token
2025-04-17 15:03:23 +00:00
Folke Behrens
2c56c46d48 compute: Set max log level for local proxy sql_over_http mod to WARN (#11629)
neondatabase/cloud#27738
2025-04-17 14:38:19 +00:00
Tristan Partin
d1728a6bcd Remove old compatibility hack for remote extensions (#11620)
Control plane has long since been updated to send the right value.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-17 14:08:42 +00:00
John Spray
0a27973584 pageserver: rename Tenant to TenantShard (#11589)
## Problem

`Tenant` isn't really a whole tenant: it's just one shard of a tenant.

## Summary of changes

- Automated rename of Tenant to TenantShard
- Followup commit to change references in comments
2025-04-17 13:29:16 +00:00
Alexander Bayandin
07c2411f6b tests: remove mentions of ALLOW_*_COMPATIBILITY_BREAKAGE (#11618)
## Problem

There are mentions of `ALLOW_BACKWARD_COMPATIBILITY_BREAKAGE` and
`ALLOW_FORWARD_COMPATIBILITY_BREAKAGE`, but in reality, this mechanism
doesn't work, so let's remove it to avoid confusion.

The idea behind it was to allow some breaking changes by adding a
special label to a PR that would `xfail` the test. However, in practice,
this means we would need to carry this label through all subsequent PRs
until the release (and artifact regeneration). This approach isn't
really viable, as it increases the risk of missing a compatibility break
in another PR.

## Summary of changes
- Remove mentions and handling of
`ALLOW_BACKWARD_COMPATIBILITY_BREAKAGE` /
`ALLOW_FORWARD_COMPATIBILITY_BREAKAGE`
2025-04-17 10:03:21 +00:00
Alexander Bayandin
5819938c93 CI(pg-clients): fix workflow permissions (#11623)
## Problem

`pg-clients` can't start:

```
The workflow is not valid. .github/workflows/pg-clients.yml (Line: 44, Col: 3): Error calling workflow 'neondatabase/neon/.github/workflows/build-build-tools-image.yml@aa19f10e7e958fbe0e0641f2e8c5952ce3be44b3'. The nested job 'check-image' is requesting 'packages: read', but is only allowed 'packages: none'. .github/workflows/pg-clients.yml (Line: 44, Col: 3): Error calling workflow 'neondatabase/neon/.github/workflows/build-build-tools-image.yml@aa19f10e7e958fbe0e0641f2e8c5952ce3be44b3'. The nested job 'build-image' is requesting 'packages: write', but is only allowed 'packages: none'.
```

## Summary of changes
- Grant required `packages: write` permissions to the workflow
2025-04-17 08:54:23 +00:00
Konstantin Knizhnik
b7548de814 Disable autovacuum and increase limit for WS approximation (#11583)
## Problem

Test lfc working set approximation becomes flaky after recent changes in
prefetch.
May be it is caused by updating HLL in `lfc_write`, may be by some other
reasons.

## Summary of changes

1. Disable autovacuum in this test (as possible source of extra page
accesses).
2. Increase upper boundary for WS approximation from 12 to 20.

---------

Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
2025-04-17 05:07:45 +00:00
Tristan Partin
9794f386f4 Make Postgres 17 the default version (#11619)
This is mostly a documentation update, but a few updates with regard to
neon_local, pageserver, and tests.

17 is our default for users in production, so dropping references to 16
makes sense.

Signed-off-by: Tristan Partin <tristan@neon.tech>

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-16 23:23:37 +00:00
Tristan Partin
79083de61c Remove forward compatibility hacks related to compute_ctl auth (#11621)
These various hacks were needed for the forward compatibility tests.
Enough time has passed since the merge that these are no longer needed.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-16 23:14:24 +00:00
Folke Behrens
ec9079f483 Allow unwrap() in tests when clippy::unwrap_used is denied (#11616)
## Problem

The proxy denies using `unwrap()`s in regular code, but we want to use
it in test code
and so have to allow it for each test block.

## Summary of changes

Set `allow-unwrap-in-tests = true` in clippy.toml and remove all
exceptions.
2025-04-16 20:05:21 +00:00
Ivan Efremov
b9b25e13a0 feat(proxy): Return prefixed errors to testodrome (#11561)
Testodrome measures uptime based on the failed requests and errors. In
case of testodrome request we send back error based on the service. This
will help us distinguish error types in testodrome and rely on the
uptime SLI.
2025-04-16 19:03:23 +00:00
Alex Chi Z.
cf2e695f49 feat(pageserver): gc-compaction meta statistics (#11601)
## Problem

We currently only have gc-compaction statistics for each single
sub-compaction job.

## Summary of changes

Add meta statistics across all sub-compaction jobs scheduled.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-16 18:51:48 +00:00
Conrad Ludgate
fc233794f6 fix(proxy): make sure that sql-over-http is TLS aware (#11612)
I noticed that while auth-broker -> local-proxy is TLS aware, and TCP
proxy -> postgres is TLS aware, HTTP proxy -> postgres is not 😅
2025-04-16 18:37:17 +00:00
Tristan Partin
c002236145 Remove compute_ctl authorization bypass if testing feature was enable (#11596)
We want to exercise the authorization middleware in our regression
tests.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-16 17:54:51 +00:00
Erik Grinaker
4af0b9b387 pageserver: don't recompress images in ImageLayerInner::filter() (#11592)
## Problem

During shard ancestor compaction, we currently recompress all page
images as we move them into a new layer file. This is expensive and
unnecessary.

Resolves #11562.
Requires #11607.

## Summary of changes

Pass through compressed page images in `ImageLayerInner::filter()`.
2025-04-16 17:10:15 +00:00
Vlad Lazar
0e00faf528 tests: stability fixes for test_migration_to_cold_secondary (#11606)
1. Compute may generate WAL on shutdown. The test assumes that after
shutdown,
no further ingest happens. Tweak the compute shutdown to make the
assumption true.
2. Assertion of local layer count post cold migration is not right since
we may have downloaded
layers due to ingest. Remove it.

Closes https://github.com/neondatabase/neon/issues/11587
2025-04-16 16:31:23 +00:00
Anastasia Lubennikova
7747a9619f compute: fix copy-paste typo for neon GUC parameters check (#11610)
fix for commit
[5063151](5063151271)
2025-04-16 15:55:11 +00:00
Erik Grinaker
46100717ad pageserver: add VectoredBlob::raw_with_header (#11607)
## Problem

To avoid recompressing page images during layer filtering, we need
access to the raw header and data from vectored reads such that we can
pass them through to the target layer.

Touches #11562.

## Summary of changes

Adds `VectoredBlob::raw_with_header()` to return a raw view of the
header+data, and updates `read()` to track it.

Also adds `blob_io::Header` with header metadata and decode logic, to
reuse for tests and assertions. This isn't yet widely used.
2025-04-16 15:38:10 +00:00
Erik Grinaker
00eeff9b8d pageserver: add compaction_shard_ancestor to disable shard ancestor compaction (#11608)
## Problem

Splits of large tenants (several TB) can cause a huge amount of shard
ancestor compaction work, which can overload Pageservers.

Touches https://github.com/neondatabase/cloud/issues/22532.

## Summary of changes

Add a setting `compaction_shard_ancestor` (default `true`) to disable
shard ancestor compaction on a per-tenant basis.
2025-04-16 14:41:02 +00:00
Matthias van de Meent
2a46426157 Update neon GUCs with new default settings (#11595)
Staging and prod both have these settings configured like this, so let's
update this so we can eventually drop the overrides in prod.
2025-04-16 13:42:22 +00:00
Tristan Partin
edc11253b6 Fix neon_local public key parsing when create compute JWKS (#11602)
Finally figured out the right incantation. I had had this in my original
go, but due to some refactoring and apparently missed testing, I
committed a mistake. The reason this doesn't currently break anything is
that we bypass the authorization middleware when the "testing" cargo
feature is enabled.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-16 12:51:48 +00:00
Heikki Linnakangas
b4e26a6284 Set last-written LSN as part of smgr_end_unlogged_build() (#11584)
This way, the callers don't need to do it, reducing the footprint of
changes we've had to made to various index AM's build functions.
2025-04-16 12:34:18 +00:00
Vlad Lazar
96b46365e4 tests: attach final metrics to allure report (#11604)
## Problem

Metrics are saved in https://github.com/neondatabase/neon/pull/11559,
but the file is not matched by the attachment regex.

## Summary of changes

Make attachment regex match the metrics file.
2025-04-16 10:26:47 +00:00
Alex Chi Z.
aa19f10e7e fix(test): allow shutdown warning in preempt tests (#11600)
## Problem

test_gc_compaction_preempt is still flaky

## Summary of changes

- allow shutdown warning logs

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-15 21:50:28 +00:00
Konstantin Knizhnik
35170656fe Allocate WalProposerConn using TopMemoryAllocator (#11577)
## Problem

See https://neondb.slack.com/archives/C04DGM6SMTM/p1744659631698609
`WalProposerConn` is allocated using current memory context which life
time is not long enough.

## Summary of changes

Allocate `WalProposerConn`  using `TopMemoryContext`.

Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
2025-04-15 19:13:12 +00:00
Tristan Partin
cd9ad75797 Remove compute_ctl authorization bypass on localhost (#11597)
For whatever reason, this never worked in production computes anyway.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-15 19:12:34 +00:00
Tristan Partin
eadb05f78e Teach neon_local to pass the Authorization header to compute_ctl (#11490)
This allows us to remove hacks in the compute_ctl authorization
middleware which allowed for bypasses of auth checks.

Fixes: https://github.com/neondatabase/neon/issues/11316

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-15 17:27:49 +00:00
Fedor Dikarev
c5115518e9 remove temp file from repo (#11586)
## Problem
In https://github.com/neondatabase/neon/pull/11409 we added temp file to
the repo.

## Summary of changes
Remove temp file from the repo.
2025-04-15 15:29:15 +00:00
Alex Chi Z.
931f8c4300 fix(pageserver): check if cancelled before waiting logical size (2/2) (#11575)
## Problem

close https://github.com/neondatabase/neon/issues/11486, proceeding
https://github.com/neondatabase/neon/pull/11531

## Summary of changes

This patch fixes the rest 50% of instability of
`test_create_churn_during_restart`. During tenant warmup, we'll request
logical size; however, if the startup gets cancelled, we won't be able
to spawn the initial logical size calculation task that sets the
`cancel_wait_for_background_loop_concurrency_limit_semaphore`.

Therefore, we check `cancelled` before proceeding to get
`cancel_wait_for_background_loop_concurrency_limit_semaphore`. There
will still be a race if the timeline shutdown happens after L5710 and
before L5711, but it should be enough to reduce the flakiness of the
test.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-15 15:16:16 +00:00
Alexander Bayandin
0f7c2cc382 CI(release): add time to RC PR branch names (#11547)
## Problem

We can't have more than one open release PR created on the same day (due
to non-unique enough branch names).

## Summary of changes
- Add time (hours and minutes) to RC PR branch names
- Also make sure we use UTC for releases
2025-04-15 15:08:05 +00:00
Erik Grinaker
983d56502b pageserver: reduce shard ancestor rewrite threshold to 30% (#11582)
## Problem

When doing power-of-two shard splits (i.e. 4 → 8 → 16), we end up
rewriting all layers since half of the pages will be local due to
striping. This causes a lot of resource usage when splitting large
tenants.

## Summary of changes

Drop the threshold of local/total pages to 30%, to reduce the amount of
layer rewrites after splits.
2025-04-15 14:26:29 +00:00
Erik Grinaker
bcef542d5b pageserver: don't rewrite invisible layers during ancestor compaction (#11580)
## Problem

Shard ancestor compaction can be very expensive following shard splits
of large tenants. We currently rewrite garbage layers after shard splits
as well, which can be a significant amount of data.

Touches https://github.com/neondatabase/cloud/issues/22532.

## Summary of changes

Don't rewrite invisible layers after shard splits.
2025-04-15 14:25:58 +00:00
a-masterov
e31455d936 Add the tests for the extensions pg_jsonschema and pg_session_jwt (#11323)
## Problem
`pg_jsonschema` and `pg_session_jwt` are not yet covered by tests
## Summary of changes
Added the tests for these extensions.
2025-04-15 14:06:01 +00:00
Alex Chi Z.
a4ea7d6194 fix(pageserver): gc-compaction verification false failure (#11564)
## Problem

https://github.com/neondatabase/neon/pull/11515 introduced a bug that
some key history cannot be verified.

If a key only exists above the horizon, the verification will fail for
its first occurrence because the history does not exist at that point.

As gc-compaction skips a key range whenever an error occurs, it might be
doing some wasted work in staging/prod now. But I'm not planning a
hotfix this week as the bug doesn't affect correctness/performance.

## Summary of changes

Allow keys with only above horizon history in the verification.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-15 13:58:32 +00:00
Alexander Bayandin
19bea5fd0c CI: do not wait for tests to trigger deploy job (#11548)
## Problem

There is too much delay between merging a PR into `main` and deploying
the changes to staging

## Summary of changes
- Trigger `deploy` job without waiting for `build-and-test-locally` job
2025-04-15 11:23:41 +00:00
a-masterov
5be94e28c4 Update the documentation of the cloud regress test (#11539)
## Problem
The information in the README.md contained errors, and some information
was missing.
## Summary of changes
Found errors are fixed, and new information is added.

---------

Co-authored-by: Alexander Bayandin <alexander@neon.tech>
2025-04-15 11:00:25 +00:00
Alexander Bayandin
63a106021a CI(allure-report-generate): Install allure to /tmp (#11579)
## Problem

The `/__w/neon/neon` directory is mounted from host to container and
persists between runs.
Sometimes the next workflow run fails to delete it:

```
Deleting the contents of '/__w/neon/neon'
Error: File was unable to be removed Error: EACCES: permission denied, rmdir '/__w/neon/neon/allure-2.32.2/bin'
```

## Summary of changes
- Download and install allure to `/tmp` which exists in container only

Ref https://github.com/neondatabase/cloud/issues/27186
2025-04-15 09:29:36 +00:00
Fedor Dikarev
9a6ace9bde introduce new runners: unit-perf and use them for benchmark jobs (#11409)
## Problem
Benchmarks results are inconsistent on existing small-metal runners

## Summary of changes
Introduce new `unit-perf` runners, and lets run benchmark on them.

The new hardware has slower, but consistent, CPU frequency - if run with
default governor schedutil.
Thus we needed to adjust some testcases' timeouts and add some retry
steps where hard-coded timeouts couldn't be increased without changing
the system under test.
-
[wait_for_last_record_lsn](6592d69a67/test_runner/fixtures/pageserver/utils.py (L193))
1000s -> 2000s
-
[test_branch_creation_many](https://github.com/neondatabase/neon/pull/11409/files#diff-2ebfe76f89004d563c7e53e3ca82462e1d85e92e6d5588e8e8f598bbe119e927)
1000s
-
[test_ingest_insert_bulk](https://github.com/neondatabase/neon/pull/11409/files#diff-e90e685be4a87053bc264a68740969e6a8872c8897b8b748d0e8c5f683a68d9f)
- with back throttling disabled compute becomes unresponsive for more
than 60 seconds (PG hard-coded client authentication connection timeout)
-
[test_sharded_ingest](https://github.com/neondatabase/neon/pull/11409/files#diff-e8d870165bd44acb9a6d8350f8640b301c1385a4108430b8d6d659b697e4a3f1)
600s -> 1200s

Right now there are only 2 runners of that class, and if we decide to go
with them, we have to check how much that type of runners we need, so
jobs not stuck with waiting for that type of runners available.

However we now decided to run those runners with governor performance
instead of schedutil.
This achieves almost same performance as previous runners but still
achieves consistent results for same commit

Related issue to activate performance governor on these runners
https://github.com/neondatabase/runner/pull/138

## Verification that it helps

### analyze runtimes on new runner for same commit

Table of runtimes for the same commit on different runners in
[run](https://github.com/neondatabase/neon/actions/runs/14417589789)

| Run | Benchmarks (1) | Benchmarks (2) |Benchmarks (3) |Benchmarks (4)
| Benchmarks (5) |
|--------|--------|---------|---------|---------|---------|
| 1 | 1950.37s | 6374.55s |  3646.15s |  4149.48s |  2330.22s | 
| 2 | - | 6369.27s |  3666.65s |  4162.42s |  2329.23s | 
| Delta % |  - |  0,07 %  | 0,5 %   |   0,3 % | 0,04 %   |
| with governor performance | 1519.57s |  4131.62s |  - | -  |  - |
| second run gov. perf. | 1513.62s |  4134.67s |  - | -  |  - |
| Delta % |  0,3 % |  0,07 %  |  -  |  - | -   |
| speedup gov. performance | 22 % |  35 % |  - | -  |  - |
| current desktop class hetzner runners (main) | 1487.10s | 3699.67s | -
| - | - |
| slower than desktop class | 2 % |  12 % |  - | -  |  - |


In summary, the runtimes for the same commit on this hardware varies
less than 1 %.

---------

Co-authored-by: BodoBolero <peterbendel@neon.tech>
2025-04-15 08:21:44 +00:00
Erik Grinaker
8c77ccfc01 pageserver: log total progress during shard ancestor compaction (#11565)
## Problem

Shard ancestor compaction doesn't currently log any global progress
information, only for the current batch.

## Summary of changes

Log the number of layers checked for eligibility this iteration, and the
total number of layers to check. This will indicate how far along the
total shard ancestor compaction has gotten for this iteration.
2025-04-15 07:25:09 +00:00
Tristan Partin
cbd2fc2395 Clean up logs and error messages in compute_ctl authorize middleware (#11576)
Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-15 01:21:18 +00:00
Tristan Partin
028a191040 Continue with s/spec/config changes (#11574)
Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-14 21:18:21 +00:00
Vlad Lazar
8cce27bedb pageserver: add a randomized read path test (#11519)
## Problem

Every time we make changes to the read path to fix a bug or add a
feature,
we end up adding another incomprehensible test.

## Summary of changes

Add some generic infrastructure for generating a layer map from a type
spec
and use that for a read path test. The test is randomized but uses a
fixed seed
by default. A fuzzing mode is available for confidence building.

See [Notion
page](https://www.notion.so/neondatabase/Read-Path-Unit-Testing-Fuzzing-1d1f189e0047806c8e5cd37781b0a350?pvs=4)
for a diagram of the layer map
used.

Just for fun I tried removing [this
commit](9990199cb4)
from https://github.com/neondatabase/neon/pull/11494
and it caught the bug in the normal mode (no fuzzing required).
2025-04-14 15:31:32 +00:00
Vlad Lazar
90b706cd96 tests: save pageserver metrics at the end of the test (#11559)
## Problem

Sometimes it's useful to see the pageserver metrics after a test in
order to debug stuff.
For example, for https://github.com/neondatabase/neon/issues/11465 I'd
like to know
what the remote storage latencies are from the client.

## Summary of changes

When stopping the env, record the pageserver metrics into a file in the
pageserver's workdir.
2025-04-14 15:13:20 +00:00
Alex Chi Z.
057ce115de fix(test): allow stale generation errors (1/2) (#11531)
## Problem

Part of https://github.com/neondatabase/neon/issues/11486

## Summary of changes

50% of the test instability of `test_create_churn_during_restart` are
due to error message gets changed. Allow the new error message.

Still need to fix other errors due to failure to acquire semaphore in
this or the next patch.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-14 14:51:17 +00:00
Vlad Lazar
e85607eed8 tests: remove config tweak allowing old versions to start with a batching config (#11560)
## Problem

Pageservers now ignore unknown config fields, so this config tweaking is
no longer needed.

## Summary of changes

Get rid of the hack.

Closes https://github.com/neondatabase/neon/issues/11524
2025-04-14 14:42:35 +00:00
Tristan Partin
437071888e Fix logging in nightly physical replication benchmarks (#11541)
Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-14 13:57:33 +00:00
Vlad Lazar
148b3701cf pageserver: add metrics for get page batch breaking reasons (#11545)
## Problem

https://github.com/neondatabase/neon/pull/11494 changes the batching
logic, but we don't have a way to evaluate it.

## Summary of changes

This PR introduces a global and per timeline metric which tracks the
reason for
which a batch was broken.
2025-04-14 13:24:47 +00:00
Christian Schwarz
daebe50e19 refactor: plumb gate and cancellation down to to blob_io::BlobWriter (#11543)
In #10063 we will switch BlobWriter to use the owned buffers IO buffered
writer, which implements double-buffering by virtue of a background task
that performs the flushing.

That task's lifecylce must be contained within the Timeline lifecycle,
so, it must hold the timeline gate open and respect Timeline::cancel.

This PR does the noisy plumbing to reduce the #10063 diff.

Refs
- extracted from https://github.com/neondatabase/neon/pull/10063
- epic https://github.com/neondatabase/neon/issues/9868
2025-04-14 11:51:01 +00:00
Arpad Müller
e0ee6fbeff Remove deprecated --compute-hook-url storcon param (#11551)
We have already migrated the storage controller to
`--control-plane-url`, added in #11173. The new param was added to
support also safekeeper specific endpoints. See the docs changes in
#11195 for further details.

Part of #11163
2025-04-14 10:36:40 +00:00
Konstantin Knizhnik
307fa2ceb7 Remove unused n_synced variable from HandleSafekeeperResponse (#11553)
## Problem

clang produce warning about unused variable `n_synced` in
HandleSafekeeperResponse

## Summary of changes

Remove local variable.

Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
2025-04-14 09:45:13 +00:00
Vlad Lazar
a338984dc7 pageserver: support keys at different LSNs in one get page batch (#11494)
## Problem

Get page batching stops when we encounter requests at different LSNs.
We are leaving batching factor on the table.

## Summary of changes

The goal is to support keys with different LSNs in a single batch and
still serve them with a single vectored get.
Important restriction: the same key at different LSNs is not supported
in one batch. Returning different key
versions is a much more intrusive change.

Firstly, the read path is changed to support "scattered" queries. This
is a conceptually simple step from
https://github.com/neondatabase/neon/pull/11463. Instead of initializing
the fringe for one keyspace,
we do it for multiple at different LSNs and let the logic already
present into the fringe handle selection.

Secondly, page service code is updated to support batching at different
LSNs. Eeach request parsed from the wire determines its effective
request LSN and keeps it in mem for the batcher toinspect. The batcher
allows keys at
different LSNs in one batch as long one key is not requested at
different LSNs.

I'd suggest doing the first pass commit by commit to get a feel for the
changes.

## Results

I used the batching test from [Christian's
PR](https://github.com/neondatabase/neon/pull/11391) which increases the
change of batch breaks. Looking at the logs I think the new code is at
the max batching factor for the workload (we
only break batches due to them being oversized or because the executor
is idle).

```
Main:
Reasons for stopping batching: {'LSN changed': 22843, 'of batch size': 33417}
test_throughput[release-pg16-50-pipelining_config0-30-100-128-batchable {'max_batch_size': 32, 'execution': 'concurrent-futures', 'mode': 'pipelined'}].perfmetric.batching_factor: 14.6662

My branch:
Reasons for stopping batching: {'of batch size': 37024}
test_throughput[release-pg16-50-pipelining_config0-30-100-128-batchable {'max_batch_size': 32, 'execution': 'concurrent-futures', 'mode': 'pipelined'}].perfmetric.batching_factor: 19.8333
```

Related: https://github.com/neondatabase/neon/issues/10765
2025-04-14 09:05:29 +00:00
Konstantin Knizhnik
8936a7abd8 Increase limit for worker processes for isolation test (#11504)
## Problem

See https://github.com/neondatabase/neon/issues/10652

Neon extension launches 2 BGW which reduce limit for parallel workers
and so affecting parallel_deadlock isolation test.

## Summary of changes

Increase `max_worker_processes` from default 8 to 16 for isolation test.

---------

Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
2025-04-12 18:09:12 +00:00
Conrad Ludgate
946e971df8 feat(proxy): add batching to cancellation queue processing (#10607)
Add batching to the redis queue, which allows us to clear it out quicker
should it slow down temporarily.
2025-04-12 09:16:22 +00:00
Dmitrii Kovalkov
d109bf8c1d neon_local: use ed25519 to gen local ssl certs (#11542)
## Problem
neon_local uses rsa to generate local SSL certs, which is slow
Follow-up on:
- https://github.com/neondatabase/neon/pull/11025#discussion_r1989453785
- https://github.com/neondatabase/neon/pull/11538

## Summary of changes
- Change key from rsa to ed25519 in neon_local
2025-04-11 17:49:15 +00:00
Alex Chi Z.
4f7b2cdd4f feat(pageserver): gc-compaction result verification (#11515)
## Problem

Part of #9114 

There was a debug-mode verification mode that verifies at every
retain_lsn. However, the code was tangled within the actual history
generation itself and it's hard to reason about correctness. This patch
adds a separate post-verification of the gc-compaction result that redos
logs at every retain_lsn and every record above the GC horizon. This
ensures that all key history we produce with gc-compaction is readable,
and if there're read errors after gc-compaction, it can only be
read-path errors instead of gc-compaction bugs.

## Summary of changes

* Add gc_compaction_verification flag, default to true.
* Implement a post-verification process.

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-11 15:50:29 +00:00
Alex Chi Z.
66f56ddaec fix(pageserver): allow shutdown errors for gc compaction tests (#11530)
## Problem

`test_pageserver_compaction_preempt` is flaky.

## Summary of changes

Allow the shutdown errors.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-11 15:20:51 +00:00
Erik Grinaker
fd16caa7d0 pageserver: yield for L0 during ancestor compaction (#11536)
## Problem

Shard ancestor compaction does not yield for L0 compaction, potentially
starving it.

close https://github.com/neondatabase/neon/issues/11125

## Summary of changes

* Yield for L0 during shard ancestor compaction.
* Return `CompactionOutcome::Pending` when limited by `rewrite_max`, for
eager rescheduling.
2025-04-11 15:09:28 +00:00
Tristan Partin
ff5a527167 Consolidate compute_ctl configuration structures (#11514)
Previously, the structure of the spec file was just the compute spec.
However, the response from the control plane get spec request included
the compute spec and the compute_ctl config. This divergence was
hindering other work such as adding regression tests for compute_ctl
HTTP authorization.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-11 15:06:29 +00:00
Arpad Müller
c66444ea15 Add timeline_import http endpoint (#11484)
The added `timleine_import` endpoint allows us to migrate safekeeper
timelines from control plane managed to storcon managed.
 
Part of #9011
2025-04-11 14:10:27 +00:00
Arpad Müller
88f01c1ca1 Introduce WalIngestError (#11506)
Introduces a `WalIngestError` struct together with a
`WalIngestErrorKind` enum, to be used for walingest related failures and
errors.

* the enum captures backtraces, so we don't regress in comparison to
`anyhow::Error`s (backtraces might be a bit shorter if we use one of the
`anyhow::Error` wrappers)
* it explicitly lists most/all of the potential cases that can occur.

I've originally been inspired to do this in #11496, but it's a
longer-term TODO.
2025-04-11 14:08:46 +00:00
Erik Grinaker
a6937a3281 pageserver: improve shard ancestor compaction logging (#11535)
## Problem

Shard ancestor compaction always logs "starting shard ancestor
compaction", even if there is no work to do. This is very spammy (every
20 seconds for every shard). It also has limited progress logging.

## Summary of changes

* Only log "starting shard ancestor compaction" when there's work to do.
* Include details about the amount of work.
* Log progress messages for each layer, and when waiting for uploads.
* Log when compaction is completed, with elapsed duration and whether
there is more work for a later iteration.
2025-04-11 12:14:08 +00:00
Erik Grinaker
3c8565a194 test_runner: propagate config via attach_hook for test fix (#11529)
## Problem

The `pagebench` benchmarks set up an initial dataset by creating a
template tenant, copying the remote storage to a bunch of new tenants,
and attaching them to Pageservers.

In #11420, we found that
`test_pageserver_characterize_throughput_with_n_tenants` had degraded
performance because it set a custom tenant config in Pageservers that
was then replaced with the default tenant config by the storage
controller.

The initial fix was to register the tenants directly in the storage
controller, but this created the tenants with generation 1. This broke
`test_basebackup_with_high_slru_count`, where the template tenant was at
generation 2, leading to all layer files at generation 2 being ignored.

Resolves #11485.
Touches #11381.

## Summary of changes

This patch addresses both test issues by modifying `attach_hook` to also
take a custom tenant config. This allows attaching tenants to
Pageservers from pre-existing remote storage, specifying both the
generation and tenant config when registering them in the storage
controller.
2025-04-11 11:31:12 +00:00
Christian Schwarz
979fa0682b tests: update batching perf test workload to include scattered LSNs (#11391)
The batching perf test workload is currently read-only sequential scans.
However, realistic workloads have concurrent writes (to other pages)
going on.

This PR simulates concurrent writes to other pages by emitting logical
replication messages.

These degrade the achieved batching factor, for the reason see
- https://github.com/neondatabase/neon/issues/10765

PR 
- https://github.com/neondatabase/neon/pull/11494

will fix this problem and get batching factor back up.

---------

Co-authored-by: Vlad Lazar <vlad@neon.tech>
2025-04-11 09:55:49 +00:00
Christian Schwarz
8884865bca tests: make test_pageserver_getpage_throttle less flaky (#11482)
# Refs

- fixes https://github.com/neondatabase/neon/issues/11395

# Problem

Since 2025-03-10, we have observed increased flakiness of
`test_pageserver_getpage_throttle`.

The test is timing-dependent by nature, and was hitting the

```
 assert duration_secs >= 10 * actual_smgr_query_seconds, (
        "smgr metrics should not include throttle wait time"
    )
```

quite frequently.

# Analysis

These failures are not reproducible.

In this PR's history is a commit that reran the test 100 times without
requiring a single retry.

In https://github.com/neondatabase/neon/issues/11395 there is a link to
a query to the test results database.
It shows that the flakiness was not constant, but rather episodic:
2025-03-{10,11,12,13} 2025-03-{19,20,21} 2025-03-31 and 2025-04-01.

To me, this suggests variability in available CPU.

# Solution

The point of the offending assertion is to ensure that most of the
request latency is spent on throttling, because testing of the
throttling mechanism is the point of the test.
The `10` magic number means at most 10% of mean latency may be spent on
request processing.

Ideally we would control the passage of time (virtual clock source) to
make this test deterministic.

But I don't see that happening in our regression test setup.

So, this PR de-flakes the test as follows:
- allot up to 66% of mean latency for request processing
- increase duration from 10s to 20s, hoping to get better protection
from momentary CPU spikes in noisy neighbor tests or VMs on the runner
host

As a drive-by, switch to `pytest.approx` and remove one self-test
assertion I can't make sense of anymore.
2025-04-11 09:38:05 +00:00
Dmitrii Kovalkov
4c4e33bc2e storage: add http/https server and cert resover metrics (#11450)
## Problem
We need to export some metrics about certs/connections to configure
alerts and make sure that all HTTP requests are gone before turning
https-only mode on.
- Closes: https://github.com/neondatabase/cloud/issues/25526

## Summary of changes
- Add started connection and connection error metrics to http/https
Server.
- Add certificate expiration time and reload metrics to
ReloadingCertificateResolver.
2025-04-11 06:11:35 +00:00
Tristan Partin
342607473a Make Endpoint::respec_deep() infinitely deep (#11527)
Because it wasn't recursive, there was a limit to the depth of updates.
This work is necessary because as we teach neon_local and compute_ctl
that the content in --spec-path should match a similar structure we get
from the control plane, the spec object itself will no longer be
toplevel. It will be under the "spec" key.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-10 19:55:51 +00:00
John Spray
9c37bfc90a pageserver/tests: make image_layer_rewrite write less data (#11525)
## Problem

This test is slow to execute, particularly if you're on a slow
environment like vscode in a browser. Might have got much slower when we
switched to direct IO?

## Summary of changes

- Reduce the scale of the test by 10x, since there was nothing special
about the original size.
2025-04-10 17:03:22 +00:00
John Spray
52dee408dc storage controller: improve safety of shard splits coinciding with controller restarts (#11412)
## Problem

The graceful leadership transfer process involves calling step_down on
the old controller, but this was not waiting for shard splits to
complete, and the new controller could therefore end up trying to abort
a shard split while it was still going on.

We mitigated this already in #11256 by avoiding the case where shard
split completion would update the database incorrectly, but this was a
fragile fix because it assumes that is the only problematic part of the
split running concurrently.

Precursors:
- #11290 
- #11256

Closes: #11254 

## Summary of changes

- Hold the reconciler gate from shard splits, so that step_down will
wait for them. Splits should always be fairly prompt, so it is okay to
wait here.
- Defense in depth: if step_down times out (hardcoded 10 second limit),
then fully terminate the controller process rather than letting it
continue running, potentially doing split-brainy things. This makes
sense because the new controller will always declare itself leader
unilaterally if step_down fails, so leaving an old controller running is
not beneficial.
- Tests: extend
`test_storage_controller_leadership_transfer_during_split` to separately
exercise the case of a split holding up step_down, and the case where
the overall timeout on step_down is hit and the controller terminates.
2025-04-10 16:55:37 +00:00
Anastasia Lubennikova
5487a20b72 compute: Set log_parameter=off for audit logging. (#11500)
Log -> Base,
pgaudit.log = 'ddl', pgaudit.log_parameter='off'

Hipaa -> Extended.
pgaudit.log = 'all, -misc', pgaudit.log_parameter='off'
    
add new level Full:
pgaudit.log='all', pgaudit.log_parameter='on'

Keep old parameter names for compatibility,
until cplane side changes are implemented and released.

closes https://github.com/neondatabase/cloud/issues/27202
2025-04-10 15:28:28 +00:00
Alex Chi Z.
f06d721a98 test(pageserver): ensure gc-compaction does not fire critical errors (#11513)
## Problem

Part of https://github.com/neondatabase/neon/issues/10395

## Summary of changes

Add a test case to ensure gc-compaction doesn't fire any critical errors
if the key history is invalid due to partial GC.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-10 14:53:37 +00:00
Christian Schwarz
2e35f23085 tests: remove ignored fair field (#11521)
Pageserver has been ignoring field
`tenant_config.timeline_get_throttle.fair`
for many monhts, since we removed it from the config struct in
neondatabase/neon#8539.

Refs
- epic https://github.com/neondatabase/cloud/issues/27320
2025-04-10 14:24:30 +00:00
Anastasia Lubennikova
5063151271 compute: Add more neon ids to compute (#11366)
Pass more neon ids to compute_ctl.
Expose them to postgres as neon extension GUCs:
neon.project_id, neon.branch_id, neon.endpoint_id.


This is the compute side PR, not yet supported by cplane.
2025-04-10 13:04:18 +00:00
Erik Grinaker
0122d97f95 test_runner: only use last gen in test_location_conf_churn (#11511)
## Problem

`test_location_conf_churn` performs random location updates on
Pageservers. While doing this, it could instruct the compute to connect
to a stale generation and execute queries. This is invalid, and will
fail if a newer generation has removed layer files used by the stale
generation.

Resolves #11348.

## Summary of changes

Only connect to the latest generation when executing queries.
2025-04-10 10:07:16 +00:00
Arseny Sher
fae7528adb walproposer: make it aware of membership (#11407)
## Problem

Walproposer should get elected and commit WAL on safekeepers specified
by the membership configuration.

## Summary of changes

- Add to wp `members_safekeepers` and `new_members_safekeepers` arrays
mapping configuration members to connection slots. Establish this
mapping (by node id) when safekeeper sends greeting, giving its id and
when mconf becomes known / changes.
- Add to TermsCollected, VotesCollected,
GetAcknowledgedByQuorumWALPosition membership aware logic. Currently it
partially duplicates existing one, but we'll drop the latter eventually.
- In python, rename Configuration to MembershipConfiguration for
clarity.
- Add test_quorum_sanity testing new logic.

ref https://github.com/neondatabase/neon/issues/10851
2025-04-10 09:55:37 +00:00
Dmitrii Kovalkov
8a72e6f888 pageserver: add enable_tls_page_service_api (#11508)
## Problem
Page service doesn't use TLS for incoming requests.
- Closes: https://github.com/neondatabase/cloud/issues/27236

## Summary of changes
- Add option `enable_tls_page_service_api` to pageserver config
- Propagate `tls_server_config` to `page_service` if the option is
enabled

No integration tests for now because I didn't find out how to call page
service API from python and AFAIK computes don't support TLS yet
2025-04-10 08:45:17 +00:00
Tristan Partin
a04e33ceb6 Remove --spec-json argument from compute_ctl (#11510)
It isn't used by the production control plane or neon_local. The removal
simplifies compute spec logic just a little bit more since we can remove
any notion of whether we should allow live reconfigurations.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-09 22:39:54 +00:00
Alex Chi Z.
af0be11503 fix(pageserver): ensure gc-compaction gets preempted by L0 (#11512)
## Problem

Part of #9114 

## Summary of changes

Gc-compaction flag was not correctly set, causing it not getting
preempted by L0.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-09 21:41:11 +00:00
Alex Chi Z.
405a17bf0b fix(pageserver): ensure gc-compaction gets preempted by L0 (#11512)
## Problem

Part of #9114 

## Summary of changes

Gc-compaction flag was not correctly set, causing it not getting
preempted by L0.

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-09 20:57:50 +00:00
Erik Grinaker
63ee8e2181 test_runner: ignore .___temp files in evict_random_layers (#11509)
## Problem

`test_location_conf_churn` often fails with `neither image nor delta
layer`, but doesn't say what the file actually is. However, past local
failures have indicated that it might be `.___temp` files.

Touches https://github.com/neondatabase/neon/issues/11348.

## Summary of changes

Ignore `.___temp` files when evicting local layers, and include the file
name in the error message.
2025-04-09 19:03:49 +00:00
Alex Chi Z.
2c21a65b0b feat(pageserver): add gc-compaction time-to-first-item stats (#11475)
## Problem

In some cases gc-compaction doesn't respond to the L0 compaction yield
notifier. I suspect it's stuck on getting the first item, and if so, we
probably need to let L0 yield notifier preempt `next_with_trace`.

## Summary of changes

- Add `time_to_first_kv_pair` to gc-compaction statistics.
- Inverse the ratio so that smaller ratio -> better compaction ratio.

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
2025-04-09 18:07:58 +00:00
Alex Chi Z.
ec66b788e2 fix(pageserver): use different walredo retry setting for gc-compaction (#11497)
## Problem

Not a complete fix for https://github.com/neondatabase/neon/issues/11492
but should work for a short term.

Our current retry strategy for walredo is to retry every request exactly
once. This retry doesn't make sense because it retries all requests
exactly once and each error is expected to cause process restart and
cause future requests to fail. I'll explain it with a scenario of two
threads requesting redos: one with an invalid history (that will cause
walredo to panic) and another that has a correct redo sequence.

First let's look at how we handle retries right now in
do_with_walredo_process. At the beginning of the function it will spawn
a new process if there's no existing one. Then it will continue to redo.
If the process fails, the first process that encounters the error will
remove the walredo process object from the OnceCell, so that the next
time it gets accessed, a new process will be spawned; if it is the last
one that uses the old walredo process, it will kill and wait the process
in `drop(proc)`. I'm skeptical whether this works under races but I
think this is not the root cause of the problem. In this retry handler,
if there are N requests attached to a walredo process and the i-th
request fails (panics the walredo), all other N-i requests will fail and
they need to retry so that they can access a new walredo process.

```
time       ---->
proc        A                 None   B
request 1   ^-----------------^ fail
            uses A for redo   replace with None
request 2      ^-------------------- fail
               uses A for redo
request 3             ^----------------^ fail
                      uses A for redo  last ref, wait for A to be killed
request 4                            ^---------------
                                     None, spawn new process B
```

The problem is with our retry strategy. Normally, for a system that we
want to retry on, the probability of errors for each of the requests are
uncorrelated. However, in walredo, a prior request that panics the
walredo process will cause all future walredo on that process to fail
(that's correlated).

So, back to the situation where we have 2 requests where one will
definitely fail and the other will succeed and we get the following
sequence, where retry attempts = 1,

* new walredo process A starts.
* request 1 (invalid) being processed on A and panics A, waiting for
retry, remove process A from the process object.
* request 2 (valid) being processed on A and receives pipe broken /
poisoned process error, waiting for retry, wait for A to be killed --
this very likely takes a while and cannot finish before request 1 gets
processed again
* new walredo process B starts.
* request 1 (invalid) being processed again on B and panics B, the whole
request fail.
* request 2 (valid) being processed again on B, and get a poisoned error
again.

```
time       ---->
proc        A                 None           B                    None
request 1   ^-----------------^--------------^--------------------^
            spawn A for redo  fail          spawn B for redo     fail
request 2      ^--------------------^-------------------------^------------^
               use A for redo       fail, wait to kill A      B for redo   fail again
```

In such cases, no matter how we set n_attempts, as long as the retry
count applies to all requests, this sequence is bound to fail both
requests because of how they get sequenced; while we could potentially
make request 2 successful.

There are many solutions to this -- like having a separate walredo
manager for compactions, or define which errors are retryable (i.e.,
broken pipe can be retried, while real walredo error won't be retried),
or having a exclusive big lock over the whole redo process (the current
one is very fine-grained). In this patch, we go with a simple approach:
use different retry attempts for different types of requests.

For gc-compaction, the attempt count is set to 0, so that it never
retries and consequently stops the compaction process -- no more redo
will be issued from gc-compaction. Once the walredo process gets
restarted, the normal read requests will proceed normally.

## Summary of changes

Add redo_attempt for each reconstruct value request to set different
retry policies.

---------

Signed-off-by: Alex Chi Z <chi@neon.tech>
Co-authored-by: Erik Grinaker <erik@neon.tech>
2025-04-09 18:01:31 +00:00
Peter Bendel
af12647b9d large tenant oltp benchmark: reindex with downtime (remove concurrently) (#11498)
## Problem

our large oltp benchmark runs very long - we want to remove the duration
of the reindex step.
we don't run concurrent workload anyhow but added "concurrently" only to
have a "prod-like" approach. But if it just doubles the time we report
because it requires two instead of one full table scan we can remove it

## Summary of changes

remove keyword concurrently from the reindex step
2025-04-09 17:11:00 +00:00
Tristan Partin
1c237d0c6d Move compute_ctl claims struct into public API (#11505)
This is preparatory work for teaching neon_local to pass the
Authorization header to compute_ctl.

Signed-off-by: Tristan Partin <tristan@neon.tech>
2025-04-09 16:58:44 +00:00
267 changed files with 9891 additions and 4160 deletions

View File

@@ -19,7 +19,7 @@
!pageserver/
!pgxn/
!proxy/
!object_storage/
!endpoint_storage/
!storage_scrubber/
!safekeeper/
!storage_broker/

View File

@@ -6,6 +6,7 @@ self-hosted-runner:
- small
- small-metal
- small-arm64
- unit-perf
- us-east-2
config-variables:
- AWS_ECR_REGION

View File

@@ -70,6 +70,7 @@ runs:
- name: Install Allure
shell: bash -euxo pipefail {0}
working-directory: /tmp
run: |
if ! which allure; then
ALLURE_ZIP=allure-${ALLURE_VERSION}.zip

View File

@@ -113,8 +113,6 @@ runs:
TEST_OUTPUT: /tmp/test_output
BUILD_TYPE: ${{ inputs.build_type }}
COMPATIBILITY_SNAPSHOT_DIR: /tmp/compatibility_snapshot_pg${{ inputs.pg_version }}
ALLOW_BACKWARD_COMPATIBILITY_BREAKAGE: contains(github.event.pull_request.labels.*.name, 'backward compatibility breakage')
ALLOW_FORWARD_COMPATIBILITY_BREAKAGE: contains(github.event.pull_request.labels.*.name, 'forward compatibility breakage')
RERUN_FAILED: ${{ inputs.rerun_failed }}
PG_VERSION: ${{ inputs.pg_version }}
SANITIZERS: ${{ inputs.sanitizers }}
@@ -135,6 +133,7 @@ runs:
fi
PERF_REPORT_DIR="$(realpath test_runner/perf-report-local)"
echo "PERF_REPORT_DIR=${PERF_REPORT_DIR}" >> ${GITHUB_ENV}
rm -rf $PERF_REPORT_DIR
TEST_SELECTION="test_runner/${{ inputs.test_selection }}"
@@ -211,11 +210,12 @@ runs:
--verbose \
-rA $TEST_SELECTION $EXTRA_PARAMS
if [[ "${{ inputs.save_perf_report }}" == "true" ]]; then
export REPORT_FROM="$PERF_REPORT_DIR"
export REPORT_TO="$PLATFORM"
scripts/generate_and_push_perf_report.sh
fi
- name: Upload performance report
if: ${{ !cancelled() && inputs.save_perf_report == 'true' }}
shell: bash -euxo pipefail {0}
run: |
export REPORT_FROM="${PERF_REPORT_DIR}"
scripts/generate_and_push_perf_report.sh
- name: Upload compatibility snapshot
# Note, that we use `github.base_ref` which is a target branch for a PR

View File

@@ -272,10 +272,13 @@ jobs:
# run pageserver tests with different settings
for get_vectored_concurrent_io in sequential sidecar-task; do
for io_engine in std-fs tokio-epoll-uring ; do
NEON_PAGESERVER_UNIT_TEST_GET_VECTORED_CONCURRENT_IO=$get_vectored_concurrent_io \
NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$io_engine \
${cov_prefix} \
cargo nextest run $CARGO_FLAGS $CARGO_FEATURES -E 'package(pageserver)'
for io_mode in buffered direct direct-rw ; do
NEON_PAGESERVER_UNIT_TEST_GET_VECTORED_CONCURRENT_IO=$get_vectored_concurrent_io \
NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOENGINE=$io_engine \
NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IOMODE=$io_mode \
${cov_prefix} \
cargo nextest run $CARGO_FLAGS $CARGO_FEATURES -E 'package(pageserver)'
done
done
done
@@ -346,7 +349,7 @@ jobs:
contents: read
statuses: write
needs: [ build-neon ]
runs-on: ${{ fromJSON(format('["self-hosted", "{0}"]', inputs.arch == 'arm64' && 'large-arm64' || 'large')) }}
runs-on: ${{ fromJSON(format('["self-hosted", "{0}"]', inputs.arch == 'arm64' && 'large-arm64' || 'large-metal')) }}
container:
image: ${{ inputs.build-tools-image }}
credentials:
@@ -392,6 +395,7 @@ jobs:
BUILD_TAG: ${{ inputs.build-tag }}
PAGESERVER_VIRTUAL_FILE_IO_ENGINE: tokio-epoll-uring
PAGESERVER_GET_VECTORED_CONCURRENT_IO: sidecar-task
PAGESERVER_VIRTUAL_FILE_IO_MODE: direct
USE_LFC: ${{ matrix.lfc_state == 'with-lfc' && 'true' || 'false' }}
# Temporary disable this step until we figure out why it's so flaky

View File

@@ -53,10 +53,13 @@ jobs:
|| inputs.component-name == 'Compute' && 'release-compute'
}}
run: |
today=$(date +'%Y-%m-%d')
echo "title=${COMPONENT_NAME} release ${today}" | tee -a ${GITHUB_OUTPUT}
echo "rc-branch=rc/${RELEASE_BRANCH}/${today}" | tee -a ${GITHUB_OUTPUT}
echo "release-branch=${RELEASE_BRANCH}" | tee -a ${GITHUB_OUTPUT}
now_date=$(date -u +'%Y-%m-%d')
now_time=$(date -u +'%H-%M-%Z')
{
echo "title=${COMPONENT_NAME} release ${now_date}"
echo "rc-branch=rc/${RELEASE_BRANCH}/${now_date}_${now_time}"
echo "release-branch=${RELEASE_BRANCH}"
} | tee -a ${GITHUB_OUTPUT}
- name: Configure git
run: |

View File

@@ -165,5 +165,5 @@ jobs:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
CURRENT_SHA: ${{ github.sha }}
run: |
RELEASE_PR_RUN_ID=$(gh api "/repos/${GITHUB_REPOSITORY}/actions/runs?head_sha=$CURRENT_SHA" | jq '[.workflow_runs[] | select(.name == "Build and Test") | select(.head_branch | test("^rc/release(-(proxy|compute))?/[0-9]{4}-[0-9]{2}-[0-9]{2}$"; "s"))] | first | .id // ("Failed to find Build and Test run from RC PR!" | halt_error(1))')
RELEASE_PR_RUN_ID=$(gh api "/repos/${GITHUB_REPOSITORY}/actions/runs?head_sha=$CURRENT_SHA" | jq '[.workflow_runs[] | select(.name == "Build and Test") | select(.head_branch | test("^rc/release.*$"; "s"))] | first | .id // ("Failed to find Build and Test run from RC PR!" | halt_error(1))')
echo "release-pr-run-id=$RELEASE_PR_RUN_ID" | tee -a $GITHUB_OUTPUT

View File

@@ -284,7 +284,7 @@ jobs:
statuses: write
contents: write
pull-requests: write
runs-on: [ self-hosted, small-metal ]
runs-on: [ self-hosted, unit-perf ]
container:
image: ${{ needs.build-build-tools-image.outputs.image }}-bookworm
credentials:
@@ -323,6 +323,8 @@ jobs:
PERF_TEST_RESULT_CONNSTR: "${{ secrets.PERF_TEST_RESULT_CONNSTR }}"
TEST_RESULT_CONNSTR: "${{ secrets.REGRESS_TEST_RESULT_CONNSTR_NEW }}"
PAGESERVER_VIRTUAL_FILE_IO_ENGINE: tokio-epoll-uring
PAGESERVER_GET_VECTORED_CONCURRENT_IO: sidecar-task
PAGESERVER_VIRTUAL_FILE_IO_MODE: direct
SYNC_BETWEEN_TESTS: true
# XXX: no coverage data handling here, since benchmarks are run on release builds,
# while coverage is currently collected for the debug ones
@@ -1236,7 +1238,7 @@ jobs:
env:
GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
run: |
TIMEOUT=1800 # 30 minutes, usually it takes ~2-3 minutes, but if runners are busy, it might take longer
TIMEOUT=5400 # 90 minutes, usually it takes ~2-3 minutes, but if runners are busy, it might take longer
INTERVAL=15 # try each N seconds
last_status="" # a variable to carry the last status of the "build-and-upload-extensions" context
@@ -1271,7 +1273,7 @@ jobs:
exit 1
deploy:
needs: [ check-permissions, push-neon-image-dev, push-compute-image-dev, push-neon-image-prod, push-compute-image-prod, meta, build-and-test-locally, trigger-custom-extensions-build-and-wait ]
needs: [ check-permissions, push-neon-image-dev, push-compute-image-dev, push-neon-image-prod, push-compute-image-prod, meta, trigger-custom-extensions-build-and-wait ]
# `!failure() && !cancelled()` is required because the workflow depends on the job that can be skipped: `push-neon-image-prod` and `push-compute-image-prod`
if: ${{ contains(fromJSON('["push-main", "storage-release", "proxy-release", "compute-release"]'), needs.meta.outputs.run-kind) && !failure() && !cancelled() }}
permissions:

View File

@@ -27,15 +27,17 @@ jobs:
- name: Fast forwarding
uses: sequoia-pgp/fast-forward@ea7628bedcb0b0b96e94383ada458d812fca4979
# See https://docs.github.com/en/graphql/reference/enums#mergestatestatus
if: ${{ github.event.pull_request.mergeable_state == 'clean' }}
if: ${{ contains(fromJSON('["clean", "unstable"]'), github.event.pull_request.mergeable_state) }}
with:
merge: true
comment: on-error
github_token: ${{ secrets.CI_ACCESS_TOKEN }}
- name: Comment if mergeable_state is not clean
if: ${{ github.event.pull_request.mergeable_state != 'clean' }}
if: ${{ !contains(fromJSON('["clean", "unstable"]'), github.event.pull_request.mergeable_state) }}
env:
GH_TOKEN: ${{ secrets.CI_ACCESS_TOKEN }}
run: |
gh pr comment ${{ github.event.pull_request.number }} \
--repo "${GITHUB_REPOSITORY}" \
--body "Not trying to forward pull-request, because \`mergeable_state\` is \`${{ github.event.pull_request.mergeable_state }}\`, not \`clean\`."
--body "Not trying to forward pull-request, because \`mergeable_state\` is \`${{ github.event.pull_request.mergeable_state }}\`, not \`clean\` or \`unstable\`."

View File

@@ -30,7 +30,7 @@ permissions:
statuses: write # require for posting a status update
env:
DEFAULT_PG_VERSION: 16
DEFAULT_PG_VERSION: 17
PLATFORM: neon-captest-new
AWS_DEFAULT_REGION: eu-central-1
@@ -42,6 +42,8 @@ jobs:
github-event-name: ${{ github.event_name }}
build-build-tools-image:
permissions:
packages: write
needs: [ check-permissions ]
uses: ./.github/workflows/build-build-tools-image.yml
secrets: inherit

93
.github/workflows/random-ops-test.yml vendored Normal file
View File

@@ -0,0 +1,93 @@
name: Random Operations Test
on:
schedule:
# * is a special character in YAML so you have to quote this string
# ┌───────────── minute (0 - 59)
# │ ┌───────────── hour (0 - 23)
# │ │ ┌───────────── day of the month (1 - 31)
# │ │ │ ┌───────────── month (1 - 12 or JAN-DEC)
# │ │ │ │ ┌───────────── day of the week (0 - 6 or SUN-SAT)
- cron: '23 */2 * * *' # runs every 2 hours
workflow_dispatch:
inputs:
random_seed:
type: number
description: 'The random seed'
required: false
default: 0
num_operations:
type: number
description: "The number of operations to test"
default: 250
defaults:
run:
shell: bash -euxo pipefail {0}
permissions: {}
env:
DEFAULT_PG_VERSION: 16
PLATFORM: neon-captest-new
AWS_DEFAULT_REGION: eu-central-1
jobs:
run-random-rests:
env:
POSTGRES_DISTRIB_DIR: /tmp/neon/pg_install
runs-on: small
permissions:
id-token: write
statuses: write
strategy:
fail-fast: false
matrix:
pg-version: [16, 17]
container:
image: ghcr.io/neondatabase/build-tools:pinned-bookworm
credentials:
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
options: --init
steps:
- name: Harden the runner (Audit all outbound calls)
uses: step-security/harden-runner@4d991eb9b905ef189e4c376166672c3f2f230481 # v2.11.0
with:
egress-policy: audit
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
- name: Download Neon artifact
uses: ./.github/actions/download
with:
name: neon-${{ runner.os }}-${{ runner.arch }}-release-artifact
path: /tmp/neon/
prefix: latest
aws-oicd-role-arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
- name: Run tests
uses: ./.github/actions/run-python-test-set
with:
build_type: remote
test_selection: random_ops
run_in_parallel: false
extra_params: -m remote_cluster
pg_version: ${{ matrix.pg-version }}
aws-oicd-role-arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
env:
NEON_API_KEY: ${{ secrets.NEON_STAGING_API_KEY }}
RANDOM_SEED: ${{ inputs.random_seed }}
NUM_OPERATIONS: ${{ inputs.num_operations }}
- name: Create Allure report
if: ${{ !cancelled() }}
id: create-allure-report
uses: ./.github/actions/allure-report-generate
with:
store-test-results-into-db: true
aws-oicd-role-arn: ${{ vars.DEV_AWS_OIDC_ROLE_ARN }}
env:
REGRESS_TEST_RESULT_CONNSTR_NEW: ${{ secrets.REGRESS_TEST_RESULT_CONNSTR_NEW }}

149
Cargo.lock generated
View File

@@ -40,7 +40,7 @@ dependencies = [
"getrandom 0.2.11",
"once_cell",
"version_check",
"zerocopy",
"zerocopy 0.7.31",
]
[[package]]
@@ -1416,6 +1416,7 @@ name = "control_plane"
version = "0.1.0"
dependencies = [
"anyhow",
"base64 0.13.1",
"camino",
"clap",
"comfy-table",
@@ -1425,10 +1426,12 @@ dependencies = [
"humantime",
"humantime-serde",
"hyper 0.14.30",
"jsonwebtoken",
"nix 0.27.1",
"once_cell",
"pageserver_api",
"pageserver_client",
"pem",
"postgres_backend",
"postgres_connection",
"regex",
@@ -1437,6 +1440,8 @@ dependencies = [
"scopeguard",
"serde",
"serde_json",
"sha2",
"spki 0.7.3",
"storage_broker",
"thiserror 1.0.69",
"tokio",
@@ -2032,6 +2037,33 @@ dependencies = [
"zeroize",
]
[[package]]
name = "endpoint_storage"
version = "0.0.1"
dependencies = [
"anyhow",
"axum",
"axum-extra",
"camino",
"camino-tempfile",
"futures",
"http-body-util",
"itertools 0.10.5",
"jsonwebtoken",
"prometheus",
"rand 0.8.5",
"remote_storage",
"serde",
"serde_json",
"test-log",
"tokio",
"tokio-util",
"tower 0.5.2",
"tracing",
"utils",
"workspace_hack",
]
[[package]]
name = "enum-map"
version = "2.5.0"
@@ -2817,6 +2849,7 @@ dependencies = [
"hyper 0.14.30",
"itertools 0.10.5",
"jemalloc_pprof",
"jsonwebtoken",
"metrics",
"once_cell",
"pprof",
@@ -2837,6 +2870,7 @@ dependencies = [
"utils",
"uuid",
"workspace_hack",
"x509-cert",
]
[[package]]
@@ -3991,33 +4025,6 @@ dependencies = [
"memchr",
]
[[package]]
name = "object_storage"
version = "0.0.1"
dependencies = [
"anyhow",
"axum",
"axum-extra",
"camino",
"camino-tempfile",
"futures",
"http-body-util",
"itertools 0.10.5",
"jsonwebtoken",
"prometheus",
"rand 0.8.5",
"remote_storage",
"serde",
"serde_json",
"test-log",
"tokio",
"tokio-util",
"tower 0.5.2",
"tracing",
"utils",
"workspace_hack",
]
[[package]]
name = "once_cell"
version = "1.20.2"
@@ -4268,6 +4275,7 @@ dependencies = [
"hyper 0.14.30",
"indoc",
"itertools 0.10.5",
"jsonwebtoken",
"md5",
"metrics",
"nix 0.27.1",
@@ -4277,6 +4285,7 @@ dependencies = [
"pageserver_api",
"pageserver_client",
"pageserver_compaction",
"pem",
"pin-project-lite",
"postgres-protocol",
"postgres-types",
@@ -4344,6 +4353,7 @@ dependencies = [
"humantime-serde",
"itertools 0.10.5",
"nix 0.27.1",
"once_cell",
"postgres_backend",
"postgres_ffi",
"rand 0.8.5",
@@ -4405,9 +4415,9 @@ dependencies = [
[[package]]
name = "papaya"
version = "0.2.0"
version = "0.2.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "aab21828b6b5952fdadd6c377728ffae53ec3a21b2febc47319ab65741f7e2fd"
checksum = "6827e3fc394523c21d4464d02c0bb1c19966ea4a58a9844ad6d746214179d2bc"
dependencies = [
"equivalent",
"seize",
@@ -5194,7 +5204,7 @@ dependencies = [
"walkdir",
"workspace_hack",
"x509-cert",
"zerocopy",
"zerocopy 0.8.24",
]
[[package]]
@@ -5584,7 +5594,7 @@ dependencies = [
"wasm-bindgen-futures",
"wasm-streams",
"web-sys",
"webpki-roots 0.26.1",
"webpki-roots",
"winreg",
]
@@ -5684,9 +5694,9 @@ dependencies = [
[[package]]
name = "ring"
version = "0.17.13"
version = "0.17.14"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "70ac5d832aa16abd7d1def883a8545280c20a60f523a370aa3a9617c2b8550ee"
checksum = "a4689e6c2294d81e88dc6261c768b63bc4fcdb852be6d1352498b114f61383b7"
dependencies = [
"cc",
"cfg-if",
@@ -5987,10 +5997,12 @@ dependencies = [
"humantime",
"hyper 0.14.30",
"itertools 0.10.5",
"jsonwebtoken",
"metrics",
"once_cell",
"pageserver_api",
"parking_lot 0.12.1",
"pem",
"postgres-protocol",
"postgres_backend",
"postgres_ffi",
@@ -6183,13 +6195,13 @@ checksum = "224e328af6e080cddbab3c770b1cf50f0351ba0577091ef2410c3951d835ff87"
[[package]]
name = "sentry"
version = "0.32.3"
version = "0.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "00421ed8fa0c995f07cde48ba6c89e80f2b312f74ff637326f392fbfd23abe02"
checksum = "255914a8e53822abd946e2ce8baa41d4cded6b8e938913b7f7b9da5b7ab44335"
dependencies = [
"httpdate",
"reqwest",
"rustls 0.21.12",
"rustls 0.23.18",
"sentry-backtrace",
"sentry-contexts",
"sentry-core",
@@ -6197,14 +6209,14 @@ dependencies = [
"sentry-tracing",
"tokio",
"ureq",
"webpki-roots 0.25.2",
"webpki-roots",
]
[[package]]
name = "sentry-backtrace"
version = "0.32.3"
version = "0.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a79194074f34b0cbe5dd33896e5928bbc6ab63a889bd9df2264af5acb186921e"
checksum = "00293cd332a859961f24fd69258f7e92af736feaeb91020cff84dac4188a4302"
dependencies = [
"backtrace",
"once_cell",
@@ -6214,9 +6226,9 @@ dependencies = [
[[package]]
name = "sentry-contexts"
version = "0.32.3"
version = "0.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eba8870c5dba2bfd9db25c75574a11429f6b95957b0a78ac02e2970dd7a5249a"
checksum = "961990f9caa76476c481de130ada05614cd7f5aa70fb57c2142f0e09ad3fb2aa"
dependencies = [
"hostname",
"libc",
@@ -6228,9 +6240,9 @@ dependencies = [
[[package]]
name = "sentry-core"
version = "0.32.3"
version = "0.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "46a75011ea1c0d5c46e9e57df03ce81f5c7f0a9e199086334a1f9c0a541e0826"
checksum = "1a6409d845707d82415c800290a5d63be5e3df3c2e417b0997c60531dfbd35ef"
dependencies = [
"once_cell",
"rand 0.8.5",
@@ -6241,9 +6253,9 @@ dependencies = [
[[package]]
name = "sentry-panic"
version = "0.32.3"
version = "0.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2eaa3ecfa3c8750c78dcfd4637cfa2598b95b52897ed184b4dc77fcf7d95060d"
checksum = "609b1a12340495ce17baeec9e08ff8ed423c337c1a84dffae36a178c783623f3"
dependencies = [
"sentry-backtrace",
"sentry-core",
@@ -6251,9 +6263,9 @@ dependencies = [
[[package]]
name = "sentry-tracing"
version = "0.32.3"
version = "0.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f715932bf369a61b7256687c6f0554141b7ce097287e30e3f7ed6e9de82498fe"
checksum = "49f4e86402d5c50239dc7d8fd3f6d5e048221d5fcb4e026d8d50ab57fe4644cb"
dependencies = [
"sentry-backtrace",
"sentry-core",
@@ -6263,9 +6275,9 @@ dependencies = [
[[package]]
name = "sentry-types"
version = "0.32.3"
version = "0.37.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4519c900ce734f7a0eb7aba0869dfb225a7af8820634a7dd51449e3b093cfb7c"
checksum = "3d3f117b8755dbede8260952de2aeb029e20f432e72634e8969af34324591631"
dependencies = [
"debugid",
"hex",
@@ -6699,8 +6711,6 @@ version = "0.1.0"
dependencies = [
"anyhow",
"async-stream",
"aws-config",
"aws-sdk-s3",
"camino",
"chrono",
"clap",
@@ -7789,7 +7799,7 @@ dependencies = [
"rustls 0.23.18",
"rustls-pki-types",
"url",
"webpki-roots 0.26.1",
"webpki-roots",
]
[[package]]
@@ -7871,6 +7881,7 @@ dependencies = [
"metrics",
"nix 0.27.1",
"once_cell",
"pem",
"pin-project-lite",
"postgres_connection",
"pprof",
@@ -8156,12 +8167,6 @@ dependencies = [
"wasm-bindgen",
]
[[package]]
name = "webpki-roots"
version = "0.25.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "14247bb57be4f377dfb94c72830b8ce8fc6beac03cf4bf7b9732eadd414123fc"
[[package]]
name = "webpki-roots"
version = "0.26.1"
@@ -8469,6 +8474,8 @@ dependencies = [
"regex-syntax 0.8.2",
"reqwest",
"rustls 0.23.18",
"rustls-pki-types",
"rustls-webpki 0.102.8",
"scopeguard",
"sec1 0.7.3",
"serde",
@@ -8497,7 +8504,6 @@ dependencies = [
"tracing-log",
"url",
"uuid",
"zerocopy",
"zeroize",
"zstd",
"zstd-safe",
@@ -8601,8 +8607,16 @@ version = "0.7.31"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1c4061bedbb353041c12f413700357bec76df2c7e2ca8e4df8bac24c6bf68e3d"
dependencies = [
"byteorder",
"zerocopy-derive",
"zerocopy-derive 0.7.31",
]
[[package]]
name = "zerocopy"
version = "0.8.24"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2586fea28e186957ef732a5f8b3be2da217d65c5969d4b1e17f973ebbe876879"
dependencies = [
"zerocopy-derive 0.8.24",
]
[[package]]
@@ -8616,6 +8630,17 @@ dependencies = [
"syn 2.0.100",
]
[[package]]
name = "zerocopy-derive"
version = "0.8.24"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a996a8f63c5c4448cd959ac1bab0aaa3306ccfd060472f85943ee0750f0169be"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.100",
]
[[package]]
name = "zerofrom"
version = "0.1.5"

View File

@@ -40,7 +40,7 @@ members = [
"libs/proxy/postgres-protocol2",
"libs/proxy/postgres-types2",
"libs/proxy/tokio-postgres2",
"object_storage",
"endpoint_storage",
]
[workspace.package]
@@ -141,6 +141,7 @@ parking_lot = "0.12"
parquet = { version = "53", default-features = false, features = ["zstd"] }
parquet_derive = "53"
pbkdf2 = { version = "0.12.1", features = ["simple", "std"] }
pem = "3.0.3"
pin-project-lite = "0.2"
pprof = { version = "0.14", features = ["criterion", "flamegraph", "frame-pointer", "prost-codec"] }
procfs = "0.16"
@@ -163,7 +164,7 @@ scopeguard = "1.1"
sysinfo = "0.29.2"
sd-notify = "0.4.1"
send-future = "0.1.0"
sentry = { version = "0.32", default-features = false, features = ["backtrace", "contexts", "panic", "rustls", "reqwest" ] }
sentry = { version = "0.37", default-features = false, features = ["backtrace", "contexts", "panic", "rustls", "reqwest" ] }
serde = { version = "1.0", features = ["derive"] }
serde_json = "1"
serde_path_to_error = "0.1"
@@ -174,6 +175,7 @@ signal-hook = "0.3"
smallvec = "1.11"
smol_str = { version = "0.2.0", features = ["serde"] }
socket2 = "0.5"
spki = "0.7.3"
strum = "0.26"
strum_macros = "0.26"
"subtle" = "2.5.0"
@@ -218,7 +220,7 @@ uuid = { version = "1.6.1", features = ["v4", "v7", "serde"] }
walkdir = "2.3.2"
rustls-native-certs = "0.8"
whoami = "1.5.1"
zerocopy = { version = "0.7", features = ["derive"] }
zerocopy = { version = "0.8", features = ["derive", "simd"] }
json-structural-diff = { version = "0.2.0" }
x509-cert = { version = "0.2.5" }

View File

@@ -89,7 +89,7 @@ RUN set -e \
--bin storage_broker \
--bin storage_controller \
--bin proxy \
--bin object_storage \
--bin endpoint_storage \
--bin neon_local \
--bin storage_scrubber \
--locked --release
@@ -122,7 +122,7 @@ COPY --from=build --chown=neon:neon /home/nonroot/target/release/safekeeper
COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_broker /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_controller /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/proxy /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/object_storage /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/endpoint_storage /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/neon_local /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_scrubber /usr/local/bin

View File

@@ -270,7 +270,7 @@ By default, this runs both debug and release modes, and all supported postgres v
testing locally, it is convenient to run just one set of permutations, like this:
```sh
DEFAULT_PG_VERSION=16 BUILD_TYPE=release ./scripts/pytest
DEFAULT_PG_VERSION=17 BUILD_TYPE=release ./scripts/pytest
```
## Flamegraphs

View File

@@ -173,7 +173,7 @@ RUN curl -fsSL "https://github.com/protocolbuffers/protobuf/releases/download/v$
&& rm -rf protoc.zip protoc
# s5cmd
ENV S5CMD_VERSION=2.2.2
ENV S5CMD_VERSION=2.3.0
RUN curl -sL "https://github.com/peak/s5cmd/releases/download/v${S5CMD_VERSION}/s5cmd_${S5CMD_VERSION}_Linux-$(uname -m | sed 's/x86_64/64bit/g' | sed 's/aarch64/arm64/g').tar.gz" | tar zxvf - s5cmd \
&& chmod +x s5cmd \
&& mv s5cmd /usr/local/bin/s5cmd
@@ -206,7 +206,7 @@ RUN curl "https://awscli.amazonaws.com/awscli-exe-linux-$(uname -m).zip" -o "aws
&& rm awscliv2.zip
# Mold: A Modern Linker
ENV MOLD_VERSION=v2.34.1
ENV MOLD_VERSION=v2.37.1
RUN set -e \
&& git clone https://github.com/rui314/mold.git \
&& mkdir mold/build \
@@ -268,7 +268,7 @@ WORKDIR /home/nonroot
RUN echo -e "--retry-connrefused\n--connect-timeout 15\n--retry 5\n--max-time 300\n" > /home/nonroot/.curlrc
# Python
ENV PYTHON_VERSION=3.11.10 \
ENV PYTHON_VERSION=3.11.12 \
PYENV_ROOT=/home/nonroot/.pyenv \
PATH=/home/nonroot/.pyenv/shims:/home/nonroot/.pyenv/bin:/home/nonroot/.poetry/bin:$PATH
RUN set -e \
@@ -296,12 +296,12 @@ ENV RUSTC_VERSION=1.86.0
ENV RUSTUP_HOME="/home/nonroot/.rustup"
ENV PATH="/home/nonroot/.cargo/bin:${PATH}"
ARG RUSTFILT_VERSION=0.2.1
ARG CARGO_HAKARI_VERSION=0.9.33
ARG CARGO_DENY_VERSION=0.16.2
ARG CARGO_HACK_VERSION=0.6.33
ARG CARGO_NEXTEST_VERSION=0.9.85
ARG CARGO_HAKARI_VERSION=0.9.36
ARG CARGO_DENY_VERSION=0.18.2
ARG CARGO_HACK_VERSION=0.6.36
ARG CARGO_NEXTEST_VERSION=0.9.94
ARG CARGO_CHEF_VERSION=0.1.71
ARG CARGO_DIESEL_CLI_VERSION=2.2.6
ARG CARGO_DIESEL_CLI_VERSION=2.2.9
RUN curl -sSO https://static.rust-lang.org/rustup/dist/$(uname -m)-unknown-linux-gnu/rustup-init && whoami && \
chmod +x rustup-init && \
./rustup-init -y --default-toolchain ${RUSTC_VERSION} && \

View File

@@ -12,3 +12,5 @@ disallowed-macros = [
# cannot disallow this, because clippy finds used from tokio macros
#"tokio::pin",
]
allow-unwrap-in-tests = true

View File

@@ -1677,7 +1677,7 @@ RUN set -e \
&& apt clean && rm -rf /var/lib/apt/lists/*
# Use `dist_man_MANS=` to skip manpage generation (which requires python3/pandoc)
ENV PGBOUNCER_TAG=pgbouncer_1_22_1
ENV PGBOUNCER_TAG=pgbouncer_1_24_1
RUN set -e \
&& git clone --recurse-submodules --depth 1 --branch ${PGBOUNCER_TAG} https://github.com/pgbouncer/pgbouncer.git pgbouncer \
&& cd pgbouncer \

View File

@@ -1,265 +0,0 @@
commit 00aa659afc9c7336ab81036edec3017168aabf40
Author: Heikki Linnakangas <heikki@neon.tech>
Date: Tue Nov 12 16:59:19 2024 +0200
Temporarily disable test that depends on timezone
diff --git a/tests/expected/generalization.out b/tests/expected/generalization.out
index 23ef5fa..9e60deb 100644
--- a/ext-src/pg_anon-src/tests/expected/generalization.out
+++ b/ext-src/pg_anon-src/tests/expected/generalization.out
@@ -284,12 +284,9 @@ SELECT anon.generalize_tstzrange('19041107','century');
["Tue Jan 01 00:00:00 1901 PST","Mon Jan 01 00:00:00 2001 PST")
(1 row)
-SELECT anon.generalize_tstzrange('19041107','millennium');
- generalize_tstzrange
------------------------------------------------------------------
- ["Thu Jan 01 00:00:00 1001 PST","Mon Jan 01 00:00:00 2001 PST")
-(1 row)
-
+-- temporarily disabled, see:
+-- https://gitlab.com/dalibo/postgresql_anonymizer/-/commit/199f0a392b37c59d92ae441fb8f037e094a11a52#note_2148017485
+--SELECT anon.generalize_tstzrange('19041107','millennium');
-- generalize_daterange
SELECT anon.generalize_daterange('19041107');
generalize_daterange
diff --git a/tests/sql/generalization.sql b/tests/sql/generalization.sql
index b868344..b4fc977 100644
--- a/ext-src/pg_anon-src/tests/sql/generalization.sql
+++ b/ext-src/pg_anon-src/tests/sql/generalization.sql
@@ -61,7 +61,9 @@ SELECT anon.generalize_tstzrange('19041107','month');
SELECT anon.generalize_tstzrange('19041107','year');
SELECT anon.generalize_tstzrange('19041107','decade');
SELECT anon.generalize_tstzrange('19041107','century');
-SELECT anon.generalize_tstzrange('19041107','millennium');
+-- temporarily disabled, see:
+-- https://gitlab.com/dalibo/postgresql_anonymizer/-/commit/199f0a392b37c59d92ae441fb8f037e094a11a52#note_2148017485
+--SELECT anon.generalize_tstzrange('19041107','millennium');
-- generalize_daterange
SELECT anon.generalize_daterange('19041107');
commit 7dd414ee75f2875cffb1d6ba474df1f135a6fc6f
Author: Alexey Masterov <alexeymasterov@neon.tech>
Date: Fri May 31 06:34:26 2024 +0000
These alternative expected files were added to consider the neon features
diff --git a/ext-src/pg_anon-src/tests/expected/permissions_masked_role_1.out b/ext-src/pg_anon-src/tests/expected/permissions_masked_role_1.out
new file mode 100644
index 0000000..2539cfd
--- /dev/null
+++ b/ext-src/pg_anon-src/tests/expected/permissions_masked_role_1.out
@@ -0,0 +1,101 @@
+BEGIN;
+CREATE EXTENSION anon CASCADE;
+NOTICE: installing required extension "pgcrypto"
+SELECT anon.init();
+ init
+------
+ t
+(1 row)
+
+CREATE ROLE mallory_the_masked_user;
+SECURITY LABEL FOR anon ON ROLE mallory_the_masked_user IS 'MASKED';
+CREATE TABLE t1(i INT);
+ALTER TABLE t1 ADD COLUMN t TEXT;
+SECURITY LABEL FOR anon ON COLUMN t1.t
+IS 'MASKED WITH VALUE NULL';
+INSERT INTO t1 VALUES (1,'test');
+--
+-- We're checking the owner's permissions
+--
+-- see
+-- https://postgresql-anonymizer.readthedocs.io/en/latest/SECURITY/#permissions
+--
+SET ROLE mallory_the_masked_user;
+SELECT anon.pseudo_first_name(0) IS NOT NULL;
+ ?column?
+----------
+ t
+(1 row)
+
+-- SHOULD FAIL
+DO $$
+BEGIN
+ PERFORM anon.init();
+ EXCEPTION WHEN insufficient_privilege
+ THEN RAISE NOTICE 'insufficient_privilege';
+END$$;
+NOTICE: insufficient_privilege
+-- SHOULD FAIL
+DO $$
+BEGIN
+ PERFORM anon.anonymize_table('t1');
+ EXCEPTION WHEN insufficient_privilege
+ THEN RAISE NOTICE 'insufficient_privilege';
+END$$;
+NOTICE: insufficient_privilege
+-- SHOULD FAIL
+SAVEPOINT fail_start_engine;
+SELECT anon.start_dynamic_masking();
+ERROR: Only supersusers can start the dynamic masking engine.
+CONTEXT: PL/pgSQL function anon.start_dynamic_masking(boolean) line 18 at RAISE
+ROLLBACK TO fail_start_engine;
+RESET ROLE;
+SELECT anon.start_dynamic_masking();
+ start_dynamic_masking
+-----------------------
+ t
+(1 row)
+
+SET ROLE mallory_the_masked_user;
+SELECT * FROM mask.t1;
+ i | t
+---+---
+ 1 |
+(1 row)
+
+-- SHOULD FAIL
+DO $$
+BEGIN
+ SELECT * FROM public.t1;
+ EXCEPTION WHEN insufficient_privilege
+ THEN RAISE NOTICE 'insufficient_privilege';
+END$$;
+NOTICE: insufficient_privilege
+-- SHOULD FAIL
+SAVEPOINT fail_stop_engine;
+SELECT anon.stop_dynamic_masking();
+ERROR: Only supersusers can stop the dynamic masking engine.
+CONTEXT: PL/pgSQL function anon.stop_dynamic_masking() line 18 at RAISE
+ROLLBACK TO fail_stop_engine;
+RESET ROLE;
+SELECT anon.stop_dynamic_masking();
+NOTICE: The previous priviledges of 'mallory_the_masked_user' are not restored. You need to grant them manually.
+ stop_dynamic_masking
+----------------------
+ t
+(1 row)
+
+SET ROLE mallory_the_masked_user;
+SELECT COUNT(*)=1 FROM anon.pg_masking_rules;
+ ?column?
+----------
+ t
+(1 row)
+
+-- SHOULD FAIL
+SAVEPOINT fail_seclabel_on_role;
+SECURITY LABEL FOR anon ON ROLE mallory_the_masked_user IS NULL;
+ERROR: permission denied
+DETAIL: The current user must have the CREATEROLE attribute.
+ROLLBACK TO fail_seclabel_on_role;
+ROLLBACK;
diff --git a/ext-src/pg_anon-src/tests/expected/permissions_owner_1.out b/ext-src/pg_anon-src/tests/expected/permissions_owner_1.out
new file mode 100644
index 0000000..8b090fe
--- /dev/null
+++ b/ext-src/pg_anon-src/tests/expected/permissions_owner_1.out
@@ -0,0 +1,104 @@
+BEGIN;
+CREATE EXTENSION anon CASCADE;
+NOTICE: installing required extension "pgcrypto"
+SELECT anon.init();
+ init
+------
+ t
+(1 row)
+
+CREATE ROLE oscar_the_owner;
+ALTER DATABASE :DBNAME OWNER TO oscar_the_owner;
+CREATE ROLE mallory_the_masked_user;
+SECURITY LABEL FOR anon ON ROLE mallory_the_masked_user IS 'MASKED';
+--
+-- We're checking the owner's permissions
+--
+-- see
+-- https://postgresql-anonymizer.readthedocs.io/en/latest/SECURITY/#permissions
+--
+SET ROLE oscar_the_owner;
+SELECT anon.pseudo_first_name(0) IS NOT NULL;
+ ?column?
+----------
+ t
+(1 row)
+
+-- SHOULD FAIL
+DO $$
+BEGIN
+ PERFORM anon.init();
+ EXCEPTION WHEN insufficient_privilege
+ THEN RAISE NOTICE 'insufficient_privilege';
+END$$;
+NOTICE: insufficient_privilege
+CREATE TABLE t1(i INT);
+ALTER TABLE t1 ADD COLUMN t TEXT;
+SECURITY LABEL FOR anon ON COLUMN t1.t
+IS 'MASKED WITH VALUE NULL';
+INSERT INTO t1 VALUES (1,'test');
+SELECT anon.anonymize_table('t1');
+ anonymize_table
+-----------------
+ t
+(1 row)
+
+SELECT * FROM t1;
+ i | t
+---+---
+ 1 |
+(1 row)
+
+UPDATE t1 SET t='test' WHERE i=1;
+-- SHOULD FAIL
+SAVEPOINT fail_start_engine;
+SELECT anon.start_dynamic_masking();
+ start_dynamic_masking
+-----------------------
+ t
+(1 row)
+
+ROLLBACK TO fail_start_engine;
+RESET ROLE;
+SELECT anon.start_dynamic_masking();
+ start_dynamic_masking
+-----------------------
+ t
+(1 row)
+
+SET ROLE oscar_the_owner;
+SELECT * FROM t1;
+ i | t
+---+------
+ 1 | test
+(1 row)
+
+--SELECT * FROM mask.t1;
+-- SHOULD FAIL
+SAVEPOINT fail_stop_engine;
+SELECT anon.stop_dynamic_masking();
+ERROR: permission denied for schema mask
+CONTEXT: SQL statement "DROP VIEW mask.t1;"
+PL/pgSQL function anon.mask_drop_view(oid) line 3 at EXECUTE
+SQL statement "SELECT anon.mask_drop_view(oid)
+ FROM pg_catalog.pg_class
+ WHERE relnamespace=quote_ident(pg_catalog.current_setting('anon.sourceschema'))::REGNAMESPACE
+ AND relkind IN ('r','p','f')"
+PL/pgSQL function anon.stop_dynamic_masking() line 22 at PERFORM
+ROLLBACK TO fail_stop_engine;
+RESET ROLE;
+SELECT anon.stop_dynamic_masking();
+NOTICE: The previous priviledges of 'mallory_the_masked_user' are not restored. You need to grant them manually.
+ stop_dynamic_masking
+----------------------
+ t
+(1 row)
+
+SET ROLE oscar_the_owner;
+-- SHOULD FAIL
+SAVEPOINT fail_seclabel_on_role;
+SECURITY LABEL FOR anon ON ROLE mallory_the_masked_user IS NULL;
+ERROR: permission denied
+DETAIL: The current user must have the CREATEROLE attribute.
+ROLLBACK TO fail_seclabel_on_role;
+ROLLBACK;

View File

@@ -11,6 +11,14 @@ index bf6edcb..89b4c7f 100644
USE_PGXS = 1 # use pgxs if not in contrib directory
PGXS := $(shell $(PG_CONFIG) --pgxs)
diff --git a/regress/expected/init-extension.out b/regress/expected/init-extension.out
index 9f2e171..f6e4f8d 100644
--- a/regress/expected/init-extension.out
+++ b/regress/expected/init-extension.out
@@ -1,3 +1,2 @@
SET client_min_messages = warning;
CREATE EXTENSION pg_repack;
-RESET client_min_messages;
diff --git a/regress/expected/nosuper.out b/regress/expected/nosuper.out
index 8d0a94e..63b68bf 100644
--- a/regress/expected/nosuper.out
@@ -42,6 +50,14 @@ index 8d0a94e..63b68bf 100644
INFO: repacking table "public.tbl_cluster"
ERROR: query failed: ERROR: current transaction is aborted, commands ignored until end of transaction block
DETAIL: query was: RESET lock_timeout
diff --git a/regress/sql/init-extension.sql b/regress/sql/init-extension.sql
index 9f2e171..f6e4f8d 100644
--- a/regress/sql/init-extension.sql
+++ b/regress/sql/init-extension.sql
@@ -1,3 +1,2 @@
SET client_min_messages = warning;
CREATE EXTENSION pg_repack;
-RESET client_min_messages;
diff --git a/regress/sql/nosuper.sql b/regress/sql/nosuper.sql
index 072f0fa..dbe60f8 100644
--- a/regress/sql/nosuper.sql

View File

@@ -15,7 +15,7 @@ index 7a4b88c..56678af 100644
HEADERS = src/halfvec.h src/sparsevec.h src/vector.h
diff --git a/src/hnswbuild.c b/src/hnswbuild.c
index b667478..dc95d89 100644
index b667478..1298aa1 100644
--- a/src/hnswbuild.c
+++ b/src/hnswbuild.c
@@ -843,9 +843,17 @@ HnswParallelBuildMain(dsm_segment *seg, shm_toc *toc)
@@ -36,7 +36,7 @@ index b667478..dc95d89 100644
/* Close relations within worker */
index_close(indexRel, indexLockmode);
table_close(heapRel, heapLockmode);
@@ -1100,12 +1108,39 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
@@ -1100,13 +1108,25 @@ BuildIndex(Relation heap, Relation index, IndexInfo *indexInfo,
SeedRandom(42);
#endif
@@ -48,32 +48,17 @@ index b667478..dc95d89 100644
BuildGraph(buildstate, forkNum);
- if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM)
+#ifdef NEON_SMGR
+ smgr_finish_unlogged_build_phase_1(RelationGetSmgr(index));
+#endif
+
+ if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM) {
if (RelationNeedsWAL(index) || forkNum == INIT_FORKNUM)
log_newpage_range(index, forkNum, 0, RelationGetNumberOfBlocksInFork(index, forkNum), true);
+#ifdef NEON_SMGR
+ {
+#if PG_VERSION_NUM >= 160000
+ RelFileLocator rlocator = RelationGetSmgr(index)->smgr_rlocator.locator;
+#else
+ RelFileNode rlocator = RelationGetSmgr(index)->smgr_rnode.node;
+#endif
+ if (set_lwlsn_block_range_hook)
+ set_lwlsn_block_range_hook(XactLastRecEnd, rlocator,
+ MAIN_FORKNUM, 0, RelationGetNumberOfBlocks(index));
+ if (set_lwlsn_relation_hook)
+ set_lwlsn_relation_hook(XactLastRecEnd, rlocator, MAIN_FORKNUM);
+ }
+#endif
+ }
+
+#ifdef NEON_SMGR
+ smgr_end_unlogged_build(RelationGetSmgr(index));
+#endif
+
FreeBuildState(buildstate);
}

View File

@@ -1,5 +1,5 @@
diff --git a/src/ruminsert.c b/src/ruminsert.c
index 255e616..7a2240f 100644
index 255e616..1c6edb7 100644
--- a/src/ruminsert.c
+++ b/src/ruminsert.c
@@ -628,6 +628,10 @@ rumbuild(Relation heap, Relation index, struct IndexInfo *indexInfo)
@@ -24,24 +24,12 @@ index 255e616..7a2240f 100644
/*
* Write index to xlog
*/
@@ -713,6 +721,22 @@ rumbuild(Relation heap, Relation index, struct IndexInfo *indexInfo)
@@ -713,6 +721,10 @@ rumbuild(Relation heap, Relation index, struct IndexInfo *indexInfo)
UnlockReleaseBuffer(buffer);
}
+#ifdef NEON_SMGR
+ {
+#if PG_VERSION_NUM >= 160000
+ RelFileLocator rlocator = RelationGetSmgr(index)->smgr_rlocator.locator;
+#else
+ RelFileNode rlocator = RelationGetSmgr(index)->smgr_rnode.node;
+#endif
+ if (set_lwlsn_block_range_hook)
+ set_lwlsn_block_range_hook(XactLastRecEnd, rlocator, MAIN_FORKNUM, 0, RelationGetNumberOfBlocks(index));
+ if (set_lwlsn_relation_hook)
+ set_lwlsn_relation_hook(XactLastRecEnd, rlocator, MAIN_FORKNUM);
+
+ smgr_end_unlogged_build(index->rd_smgr);
+ }
+ smgr_end_unlogged_build(index->rd_smgr);
+#endif
+
/*

View File

@@ -22,7 +22,7 @@ commands:
- name: local_proxy
user: postgres
sysvInitAction: respawn
shell: '/usr/local/bin/local_proxy --config-path /etc/local_proxy/config.json --pid-path /etc/local_proxy/pid --http 0.0.0.0:10432'
shell: 'RUST_LOG="info,proxy::serverless::sql_over_http=warn" /usr/local/bin/local_proxy --config-path /etc/local_proxy/config.json --pid-path /etc/local_proxy/pid --http 0.0.0.0:10432'
- name: postgres-exporter
user: nobody
sysvInitAction: respawn

View File

@@ -22,7 +22,7 @@ commands:
- name: local_proxy
user: postgres
sysvInitAction: respawn
shell: '/usr/local/bin/local_proxy --config-path /etc/local_proxy/config.json --pid-path /etc/local_proxy/pid --http 0.0.0.0:10432'
shell: 'RUST_LOG="info,proxy::serverless::sql_over_http=warn" /usr/local/bin/local_proxy --config-path /etc/local_proxy/config.json --pid-path /etc/local_proxy/pid --http 0.0.0.0:10432'
- name: postgres-exporter
user: nobody
sysvInitAction: respawn

View File

@@ -29,13 +29,12 @@
//! ```sh
//! compute_ctl -D /var/db/postgres/compute \
//! -C 'postgresql://cloud_admin@localhost/postgres' \
//! -S /var/db/postgres/specs/current.json \
//! -c /var/db/postgres/configs/config.json \
//! -b /usr/local/bin/postgres \
//! -r http://pg-ext-s3-gateway \
//! ```
use std::ffi::OsString;
use std::fs::File;
use std::path::Path;
use std::process::exit;
use std::sync::mpsc;
use std::thread;
@@ -43,8 +42,7 @@ use std::time::Duration;
use anyhow::{Context, Result};
use clap::Parser;
use compute_api::responses::ComputeCtlConfig;
use compute_api::spec::ComputeSpec;
use compute_api::responses::ComputeConfig;
use compute_tools::compute::{
BUILD_TAG, ComputeNode, ComputeNodeParams, forward_termination_signal,
};
@@ -59,24 +57,13 @@ use tracing::{error, info};
use url::Url;
use utils::failpoint_support;
// Compatibility hack: if the control plane specified any remote-ext-config
// use the default value for extension storage proxy gateway.
// Remove this once the control plane is updated to pass the gateway URL
fn parse_remote_ext_config(arg: &str) -> Result<String> {
if arg.starts_with("http") {
Ok(arg.trim_end_matches('/').to_string())
} else {
Ok("http://pg-ext-s3-gateway".to_string())
}
}
#[derive(Parser)]
#[command(rename_all = "kebab-case")]
struct Cli {
#[arg(short = 'b', long, default_value = "postgres", env = "POSTGRES_PATH")]
pub pgbin: String,
#[arg(short = 'r', long, value_parser = parse_remote_ext_config)]
#[arg(short = 'r', long)]
pub remote_ext_config: Option<String>,
/// The port to bind the external listening HTTP server to. Clients running
@@ -118,16 +105,19 @@ struct Cli {
#[arg(long)]
pub set_disk_quota_for_fs: Option<String>,
#[arg(short = 's', long = "spec", group = "spec")]
pub spec_json: Option<String>,
#[arg(short = 'S', long, group = "spec-path")]
pub spec_path: Option<OsString>,
#[arg(short = 'c', long)]
pub config: Option<OsString>,
#[arg(short = 'i', long, group = "compute-id")]
pub compute_id: String,
#[arg(short = 'p', long, conflicts_with_all = ["spec", "spec-path"], value_name = "CONTROL_PLANE_API_BASE_URL")]
#[arg(
short = 'p',
long,
conflicts_with = "config",
value_name = "CONTROL_PLANE_API_BASE_URL",
requires = "compute-id"
)]
pub control_plane_uri: Option<String>,
}
@@ -136,7 +126,7 @@ fn main() -> Result<()> {
let scenario = failpoint_support::init();
// For historical reasons, the main thread that processes the spec and launches postgres
// For historical reasons, the main thread that processes the config and launches postgres
// is synchronous, but we always have this tokio runtime available and we "enter" it so
// that you can use tokio::spawn() and tokio::runtime::Handle::current().block_on(...)
// from all parts of compute_ctl.
@@ -152,7 +142,7 @@ fn main() -> Result<()> {
let connstr = Url::parse(&cli.connstr).context("cannot parse connstr as a URL")?;
let cli_spec = try_spec_from_cli(&cli)?;
let config = get_config(&cli)?;
let compute_node = ComputeNode::new(
ComputeNodeParams {
@@ -172,10 +162,8 @@ fn main() -> Result<()> {
cgroup: cli.cgroup,
#[cfg(target_os = "linux")]
vm_monitor_addr: cli.vm_monitor_addr,
live_config_allowed: cli_spec.live_config_allowed,
},
cli_spec.spec,
cli_spec.compute_ctl_config,
config,
)?;
let exit_code = compute_node.run()?;
@@ -200,37 +188,17 @@ async fn init() -> Result<()> {
Ok(())
}
fn try_spec_from_cli(cli: &Cli) -> Result<CliSpecParams> {
// First, try to get cluster spec from the cli argument
if let Some(ref spec_json) = cli.spec_json {
info!("got spec from cli argument {}", spec_json);
return Ok(CliSpecParams {
spec: Some(serde_json::from_str(spec_json)?),
compute_ctl_config: ComputeCtlConfig::default(),
live_config_allowed: false,
});
fn get_config(cli: &Cli) -> Result<ComputeConfig> {
// First, read the config from the path if provided
if let Some(ref config) = cli.config {
let file = File::open(config)?;
return Ok(serde_json::from_reader(&file)?);
}
// Second, try to read it from the file if path is provided
if let Some(ref spec_path) = cli.spec_path {
let file = File::open(Path::new(spec_path))?;
return Ok(CliSpecParams {
spec: Some(serde_json::from_reader(file)?),
compute_ctl_config: ComputeCtlConfig::default(),
live_config_allowed: true,
});
}
if cli.control_plane_uri.is_none() {
panic!("must specify --control-plane-uri");
};
match get_spec_from_control_plane(cli.control_plane_uri.as_ref().unwrap(), &cli.compute_id) {
Ok(resp) => Ok(CliSpecParams {
spec: resp.0,
compute_ctl_config: resp.1,
live_config_allowed: true,
}),
// If the config wasn't provided in the CLI arguments, then retrieve it from
// the control plane
match get_config_from_control_plane(cli.control_plane_uri.as_ref().unwrap(), &cli.compute_id) {
Ok(config) => Ok(config),
Err(e) => {
error!(
"cannot get response from control plane: {}\n\
@@ -242,14 +210,6 @@ fn try_spec_from_cli(cli: &Cli) -> Result<CliSpecParams> {
}
}
struct CliSpecParams {
/// If a spec was provided via CLI or file, the [`ComputeSpec`]
spec: Option<ComputeSpec>,
#[allow(dead_code)]
compute_ctl_config: ComputeCtlConfig,
live_config_allowed: bool,
}
fn deinit_and_exit(exit_code: Option<i32>) -> ! {
// Shutdown trace pipeline gracefully, so that it has a chance to send any
// pending traces before we exit. Shutting down OTEL tracing provider may

View File

@@ -11,7 +11,7 @@ use std::{env, fs};
use anyhow::{Context, Result};
use chrono::{DateTime, Utc};
use compute_api::privilege::Privilege;
use compute_api::responses::{ComputeCtlConfig, ComputeMetrics, ComputeStatus};
use compute_api::responses::{ComputeConfig, ComputeCtlConfig, ComputeMetrics, ComputeStatus};
use compute_api::spec::{
ComputeAudit, ComputeFeature, ComputeMode, ComputeSpec, ExtVersion, PgIdent,
};
@@ -93,20 +93,6 @@ pub struct ComputeNodeParams {
/// the address of extension storage proxy gateway
pub ext_remote_storage: Option<String>,
/// We should only allow live re- / configuration of the compute node if
/// it uses 'pull model', i.e. it can go to control-plane and fetch
/// the latest configuration. Otherwise, there could be a case:
/// - we start compute with some spec provided as argument
/// - we push new spec and it does reconfiguration
/// - but then something happens and compute pod / VM is destroyed,
/// so k8s controller starts it again with the **old** spec
///
/// and the same for empty computes:
/// - we started compute without any spec
/// - we push spec and it does configuration
/// - but then it is restarted without any spec again
pub live_config_allowed: bool,
}
/// Compute node info shared across several `compute_ctl` threads.
@@ -317,11 +303,7 @@ struct StartVmMonitorResult {
}
impl ComputeNode {
pub fn new(
params: ComputeNodeParams,
cli_spec: Option<ComputeSpec>,
compute_ctl_config: ComputeCtlConfig,
) -> Result<Self> {
pub fn new(params: ComputeNodeParams, config: ComputeConfig) -> Result<Self> {
let connstr = params.connstr.as_str();
let conn_conf = postgres::config::Config::from_str(connstr)
.context("cannot build postgres config from connstr")?;
@@ -329,8 +311,8 @@ impl ComputeNode {
.context("cannot build tokio postgres config from connstr")?;
let mut new_state = ComputeState::new();
if let Some(cli_spec) = cli_spec {
let pspec = ParsedSpec::try_from(cli_spec).map_err(|msg| anyhow::anyhow!(msg))?;
if let Some(spec) = config.spec {
let pspec = ParsedSpec::try_from(spec).map_err(|msg| anyhow::anyhow!(msg))?;
new_state.pspec = Some(pspec);
}
@@ -341,7 +323,7 @@ impl ComputeNode {
state: Mutex::new(new_state),
state_changed: Condvar::new(),
ext_download_progress: RwLock::new(HashMap::new()),
compute_ctl_config,
compute_ctl_config: config.compute_ctl_config,
})
}
@@ -537,11 +519,14 @@ impl ComputeNode {
let pspec = compute_state.pspec.as_ref().expect("spec must be set");
info!(
"starting compute for project {}, operation {}, tenant {}, timeline {}, features {:?}, spec.remote_extensions {:?}",
"starting compute for project {}, operation {}, tenant {}, timeline {}, project {}, branch {}, endpoint {}, features {:?}, spec.remote_extensions {:?}",
pspec.spec.cluster.cluster_id.as_deref().unwrap_or("None"),
pspec.spec.operation_uuid.as_deref().unwrap_or("None"),
pspec.tenant_id,
pspec.timeline_id,
pspec.spec.project_id.as_deref().unwrap_or("None"),
pspec.spec.branch_id.as_deref().unwrap_or("None"),
pspec.spec.endpoint_id.as_deref().unwrap_or("None"),
pspec.spec.features,
pspec.spec.remote_extensions,
);
@@ -645,19 +630,42 @@ impl ComputeNode {
});
}
// Configure and start rsyslog for HIPAA if necessary
if let ComputeAudit::Hipaa = pspec.spec.audit_log_level {
let remote_endpoint = std::env::var("AUDIT_LOGGING_ENDPOINT").unwrap_or("".to_string());
if remote_endpoint.is_empty() {
anyhow::bail!("AUDIT_LOGGING_ENDPOINT is empty");
// Configure and start rsyslog for compliance audit logging
match pspec.spec.audit_log_level {
ComputeAudit::Hipaa | ComputeAudit::Extended | ComputeAudit::Full => {
let remote_endpoint =
std::env::var("AUDIT_LOGGING_ENDPOINT").unwrap_or("".to_string());
if remote_endpoint.is_empty() {
anyhow::bail!("AUDIT_LOGGING_ENDPOINT is empty");
}
let log_directory_path = Path::new(&self.params.pgdata).join("log");
let log_directory_path = log_directory_path.to_string_lossy().to_string();
// Add project_id,endpoint_id tag to identify the logs.
//
// These ids are passed from cplane,
// for backwards compatibility (old computes that don't have them),
// we set them to None.
// TODO: Clean up this code when all computes have them.
let tag: Option<String> = match (
pspec.spec.project_id.as_deref(),
pspec.spec.endpoint_id.as_deref(),
) {
(Some(project_id), Some(endpoint_id)) => {
Some(format!("{project_id}/{endpoint_id}"))
}
(Some(project_id), None) => Some(format!("{project_id}/None")),
(None, Some(endpoint_id)) => Some(format!("None,{endpoint_id}")),
(None, None) => None,
};
configure_audit_rsyslog(log_directory_path.clone(), tag, &remote_endpoint)?;
// Launch a background task to clean up the audit logs
launch_pgaudit_gc(log_directory_path);
}
let log_directory_path = Path::new(&self.params.pgdata).join("log");
let log_directory_path = log_directory_path.to_string_lossy().to_string();
configure_audit_rsyslog(log_directory_path.clone(), "hipaa", &remote_endpoint)?;
// Launch a background task to clean up the audit logs
launch_pgaudit_gc(log_directory_path);
_ => {}
}
// Configure and start rsyslog for Postgres logs export

View File

@@ -89,6 +89,15 @@ pub fn write_postgres_conf(
escape_conf_value(&s.to_string())
)?;
}
if let Some(s) = &spec.project_id {
writeln!(file, "neon.project_id={}", escape_conf_value(s))?;
}
if let Some(s) = &spec.branch_id {
writeln!(file, "neon.branch_id={}", escape_conf_value(s))?;
}
if let Some(s) = &spec.endpoint_id {
writeln!(file, "neon.endpoint_id={}", escape_conf_value(s))?;
}
// tls
if let Some(tls_config) = tls_config {
@@ -169,7 +178,7 @@ pub fn write_postgres_conf(
// and don't allow the user or the control plane admin to change them.
match spec.audit_log_level {
ComputeAudit::Disabled => {}
ComputeAudit::Log => {
ComputeAudit::Log | ComputeAudit::Base => {
writeln!(file, "# Managed by compute_ctl base audit settings: start")?;
writeln!(file, "pgaudit.log='ddl,role'")?;
// Disable logging of catalog queries to reduce the noise
@@ -193,16 +202,20 @@ pub fn write_postgres_conf(
}
writeln!(file, "# Managed by compute_ctl base audit settings: end")?;
}
ComputeAudit::Hipaa => {
ComputeAudit::Hipaa | ComputeAudit::Extended | ComputeAudit::Full => {
writeln!(
file,
"# Managed by compute_ctl compliance audit settings: begin"
)?;
// This log level is very verbose
// but this is necessary for HIPAA compliance.
// Exclude 'misc' category, because it doesn't contain anythig relevant.
writeln!(file, "pgaudit.log='all, -misc'")?;
writeln!(file, "pgaudit.log_parameter=on")?;
// Enable logging of parameters.
// This is very verbose and may contain sensitive data.
if spec.audit_log_level == ComputeAudit::Full {
writeln!(file, "pgaudit.log_parameter=on")?;
writeln!(file, "pgaudit.log='all'")?;
} else {
writeln!(file, "pgaudit.log_parameter=off")?;
writeln!(file, "pgaudit.log='all, -misc'")?;
}
// Disable logging of catalog queries
// The catalog doesn't contain sensitive data, so we don't need to audit it.
writeln!(file, "pgaudit.log_catalog=off")?;

View File

@@ -6,4 +6,5 @@ pub(crate) mod request_id;
pub(crate) use json::Json;
pub(crate) use path::Path;
pub(crate) use query::Query;
#[allow(unused)]
pub(crate) use request_id::RequestId;

View File

@@ -1,24 +1,19 @@
use std::{collections::HashSet, net::SocketAddr};
use std::collections::HashSet;
use anyhow::{Result, anyhow};
use axum::{RequestExt, body::Body, extract::ConnectInfo};
use axum::{RequestExt, body::Body};
use axum_extra::{
TypedHeader,
headers::{Authorization, authorization::Bearer},
};
use compute_api::requests::ComputeClaims;
use futures::future::BoxFuture;
use http::{Request, Response, StatusCode};
use jsonwebtoken::{Algorithm, DecodingKey, TokenData, Validation, jwk::JwkSet};
use serde::Deserialize;
use tower_http::auth::AsyncAuthorizeRequest;
use tracing::warn;
use tracing::{debug, warn};
use crate::http::{JsonResponse, extract::RequestId};
#[derive(Clone, Debug, Deserialize)]
pub(in crate::http) struct Claims {
compute_id: String,
}
use crate::http::JsonResponse;
#[derive(Clone, Debug)]
pub(in crate::http) struct Authorize {
@@ -57,31 +52,6 @@ impl AsyncAuthorizeRequest<Body> for Authorize {
let validation = self.validation.clone();
Box::pin(async move {
let request_id = request.extract_parts::<RequestId>().await.unwrap();
// TODO: Remove this stanza after teaching neon_local and the
// regression tests to use a JWT + JWKS.
//
// https://github.com/neondatabase/neon/issues/11316
if cfg!(feature = "testing") {
warn!(%request_id, "Skipping compute_ctl authorization check");
return Ok(request);
}
let connect_info = request
.extract_parts::<ConnectInfo<SocketAddr>>()
.await
.unwrap();
// In the event the request is coming from the loopback interface,
// allow all requests
if connect_info.ip().is_loopback() {
warn!(%request_id, "Bypassed authorization because request is coming from the loopback interface");
return Ok(request);
}
let TypedHeader(Authorization(bearer)) = request
.extract_parts::<TypedHeader<Authorization<Bearer>>>()
.await
@@ -97,7 +67,7 @@ impl AsyncAuthorizeRequest<Body> for Authorize {
if data.claims.compute_id != compute_id {
return Err(JsonResponse::error(
StatusCode::UNAUTHORIZED,
"invalid claims in authorization token",
"invalid compute ID in authorization token claims",
));
}
@@ -112,13 +82,21 @@ impl AsyncAuthorizeRequest<Body> for Authorize {
impl Authorize {
/// Verify the token using the JSON Web Key set and return the token data.
fn verify(jwks: &JwkSet, token: &str, validation: &Validation) -> Result<TokenData<Claims>> {
fn verify(
jwks: &JwkSet,
token: &str,
validation: &Validation,
) -> Result<TokenData<ComputeClaims>> {
debug_assert!(!jwks.keys.is_empty());
debug!("verifying token {}", token);
for jwk in jwks.keys.iter() {
let decoding_key = match DecodingKey::from_jwk(jwk) {
Ok(key) => key,
Err(e) => {
warn!(
"Failed to construct decoding key from {}: {}",
"failed to construct decoding key from {}: {}",
jwk.common.key_id.as_ref().unwrap(),
e
);
@@ -127,11 +105,11 @@ impl Authorize {
}
};
match jsonwebtoken::decode::<Claims>(token, &decoding_key, validation) {
match jsonwebtoken::decode::<ComputeClaims>(token, &decoding_key, validation) {
Ok(data) => return Ok(data),
Err(e) => {
warn!(
"Failed to decode authorization token using {}: {}",
"failed to decode authorization token using {}: {}",
jwk.common.key_id.as_ref().unwrap(),
e
);
@@ -141,6 +119,6 @@ impl Authorize {
}
}
Err(anyhow!("Failed to verify authorization token"))
Err(anyhow!("failed to verify authorization token"))
}
}

View File

@@ -22,13 +22,6 @@ pub(in crate::http) async fn configure(
State(compute): State<Arc<ComputeNode>>,
request: Json<ConfigurationRequest>,
) -> Response {
if !compute.params.live_config_allowed {
return JsonResponse::error(
StatusCode::PRECONDITION_FAILED,
"live configuration is not allowed for this compute node".to_string(),
);
}
let pspec = match ParsedSpec::try_from(request.spec.clone()) {
Ok(p) => p,
Err(e) => return JsonResponse::error(StatusCode::BAD_REQUEST, e),

View File

@@ -1,8 +1,8 @@
use metrics::core::{AtomicF64, Collector, GenericGauge};
use metrics::core::{AtomicF64, AtomicU64, Collector, GenericCounter, GenericGauge};
use metrics::proto::MetricFamily;
use metrics::{
IntCounterVec, IntGaugeVec, UIntGaugeVec, register_gauge, register_int_counter_vec,
register_int_gauge_vec, register_uint_gauge_vec,
IntCounterVec, IntGaugeVec, UIntGaugeVec, register_gauge, register_int_counter,
register_int_counter_vec, register_int_gauge_vec, register_uint_gauge_vec,
};
use once_cell::sync::Lazy;
@@ -19,13 +19,13 @@ pub(crate) static INSTALLED_EXTENSIONS: Lazy<UIntGaugeVec> = Lazy::new(|| {
// but for all our APIs we defined a 'slug'/method/operationId in the OpenAPI spec.
// And it's fair to call it a 'RPC' (Remote Procedure Call).
pub enum CPlaneRequestRPC {
GetSpec,
GetConfig,
}
impl CPlaneRequestRPC {
pub fn as_str(&self) -> &str {
match self {
CPlaneRequestRPC::GetSpec => "GetSpec",
CPlaneRequestRPC::GetConfig => "GetConfig",
}
}
}
@@ -81,6 +81,22 @@ pub(crate) static COMPUTE_CTL_UP: Lazy<IntGaugeVec> = Lazy::new(|| {
.expect("failed to define a metric")
});
pub(crate) static PG_CURR_DOWNTIME_MS: Lazy<GenericGauge<AtomicF64>> = Lazy::new(|| {
register_gauge!(
"compute_pg_current_downtime_ms",
"Non-cumulative duration of Postgres downtime in ms; resets after successful check",
)
.expect("failed to define a metric")
});
pub(crate) static PG_TOTAL_DOWNTIME_MS: Lazy<GenericCounter<AtomicU64>> = Lazy::new(|| {
register_int_counter!(
"compute_pg_downtime_ms_total",
"Cumulative duration of Postgres downtime in ms",
)
.expect("failed to define a metric")
});
pub fn collect() -> Vec<MetricFamily> {
let mut metrics = COMPUTE_CTL_UP.collect();
metrics.extend(INSTALLED_EXTENSIONS.collect());
@@ -88,5 +104,7 @@ pub fn collect() -> Vec<MetricFamily> {
metrics.extend(REMOTE_EXT_REQUESTS_TOTAL.collect());
metrics.extend(DB_MIGRATION_FAILED.collect());
metrics.extend(AUDIT_LOG_DIR_SIZE.collect());
metrics.extend(PG_CURR_DOWNTIME_MS.collect());
metrics.extend(PG_TOTAL_DOWNTIME_MS.collect());
metrics
}

View File

@@ -6,197 +6,294 @@ use chrono::{DateTime, Utc};
use compute_api::responses::ComputeStatus;
use compute_api::spec::ComputeFeature;
use postgres::{Client, NoTls};
use tracing::{debug, error, info, warn};
use tracing::{Level, error, info, instrument, span};
use crate::compute::ComputeNode;
use crate::metrics::{PG_CURR_DOWNTIME_MS, PG_TOTAL_DOWNTIME_MS};
const MONITOR_CHECK_INTERVAL: Duration = Duration::from_millis(500);
// Spin in a loop and figure out the last activity time in the Postgres.
// Then update it in the shared state. This function never errors out.
// NB: the only expected panic is at `Mutex` unwrap(), all other errors
// should be handled gracefully.
fn watch_compute_activity(compute: &ComputeNode) {
// Suppose that `connstr` doesn't change
let connstr = compute.params.connstr.clone();
let conf = compute.get_conn_conf(Some("compute_ctl:activity_monitor"));
struct ComputeMonitor {
compute: Arc<ComputeNode>,
// During startup and configuration we connect to every Postgres database,
// but we don't want to count this as some user activity. So wait until
// the compute fully started before monitoring activity.
wait_for_postgres_start(compute);
/// The moment when Postgres had some activity,
/// that should prevent compute from being suspended.
last_active: Option<DateTime<Utc>>,
// Define `client` outside of the loop to reuse existing connection if it's active.
let mut client = conf.connect(NoTls);
/// The moment when we last tried to check Postgres.
last_checked: DateTime<Utc>,
/// The last moment we did a successful Postgres check.
last_up: DateTime<Utc>,
let mut sleep = false;
let mut prev_active_time: Option<f64> = None;
let mut prev_sessions: Option<i64> = None;
/// Only used for internal statistics change tracking
/// between monitor runs and can be outdated.
active_time: Option<f64>,
/// Only used for internal statistics change tracking
/// between monitor runs and can be outdated.
sessions: Option<i64>,
if compute.has_feature(ComputeFeature::ActivityMonitorExperimental) {
info!("starting experimental activity monitor for {}", connstr);
} else {
info!("starting activity monitor for {}", connstr);
/// Use experimental statistics-based activity monitor. It's no longer
/// 'experimental' per se, as it's enabled for everyone, but we still
/// keep the flag as an option to turn it off in some cases if it will
/// misbehave.
experimental: bool,
}
impl ComputeMonitor {
fn report_down(&self) {
let now = Utc::now();
// Calculate and report current downtime
// (since the last time Postgres was up)
let downtime = now.signed_duration_since(self.last_up);
PG_CURR_DOWNTIME_MS.set(downtime.num_milliseconds() as f64);
// Calculate and update total downtime
// (cumulative duration of Postgres downtime in ms)
let inc = now
.signed_duration_since(self.last_checked)
.num_milliseconds();
PG_TOTAL_DOWNTIME_MS.inc_by(inc as u64);
}
loop {
// We use `continue` a lot, so it's more convenient to sleep at the top of the loop.
// But skip the first sleep, so we can connect to Postgres immediately.
if sleep {
// Should be outside of the mutex lock to allow others to read while we sleep.
thread::sleep(MONITOR_CHECK_INTERVAL);
} else {
sleep = true;
}
fn report_up(&mut self) {
self.last_up = Utc::now();
PG_CURR_DOWNTIME_MS.set(0.0);
}
match &mut client {
Ok(cli) => {
if cli.is_closed() {
info!("connection to Postgres is closed, trying to reconnect");
fn downtime_info(&self) -> String {
format!(
"total_ms: {}, current_ms: {}, last_up: {}",
PG_TOTAL_DOWNTIME_MS.get(),
PG_CURR_DOWNTIME_MS.get(),
self.last_up
)
}
// Connection is closed, reconnect and try again.
client = conf.connect(NoTls);
continue;
}
/// Spin in a loop and figure out the last activity time in the Postgres.
/// Then update it in the shared state. This function never errors out.
/// NB: the only expected panic is at `Mutex` unwrap(), all other errors
/// should be handled gracefully.
#[instrument(skip_all)]
pub fn run(&mut self) {
// Suppose that `connstr` doesn't change
let connstr = self.compute.params.connstr.clone();
let conf = self
.compute
.get_conn_conf(Some("compute_ctl:compute_monitor"));
// This is a new logic, only enable if the feature flag is set.
// TODO: remove this once we are sure that it works OR drop it altogether.
if compute.has_feature(ComputeFeature::ActivityMonitorExperimental) {
// First, check if the total active time or sessions across all databases has changed.
// If it did, it means that user executed some queries. In theory, it can even go down if
// some databases were dropped, but it's still a user activity.
match get_database_stats(cli) {
Ok((active_time, sessions)) => {
let mut detected_activity = false;
// During startup and configuration we connect to every Postgres database,
// but we don't want to count this as some user activity. So wait until
// the compute fully started before monitoring activity.
wait_for_postgres_start(&self.compute);
prev_active_time = match prev_active_time {
Some(prev_active_time) => {
if active_time != prev_active_time {
detected_activity = true;
}
Some(active_time)
}
None => Some(active_time),
};
prev_sessions = match prev_sessions {
Some(prev_sessions) => {
if sessions != prev_sessions {
detected_activity = true;
}
Some(sessions)
}
None => Some(sessions),
};
// Define `client` outside of the loop to reuse existing connection if it's active.
let mut client = conf.connect(NoTls);
if detected_activity {
// Update the last active time and continue, we don't need to
// check backends state change.
compute.update_last_active(Some(Utc::now()));
continue;
}
}
Err(e) => {
error!("could not get database statistics: {}", e);
continue;
}
}
}
info!("starting compute monitor for {}", connstr);
// Second, if database statistics is the same, check all backends state change,
// maybe there is some with more recent activity. `get_backends_state_change()`
// can return None or stale timestamp, so it's `compute.update_last_active()`
// responsibility to check if the new timestamp is more recent than the current one.
// This helps us to discover new sessions, that did nothing yet.
match get_backends_state_change(cli) {
Ok(last_active) => {
compute.update_last_active(last_active);
}
Err(e) => {
error!("could not get backends state change: {}", e);
}
}
// Finally, if there are existing (logical) walsenders, do not suspend.
//
// walproposer doesn't currently show up in pg_stat_replication,
// but protect if it will be
let ws_count_query = "select count(*) from pg_stat_replication where application_name != 'walproposer';";
match cli.query_one(ws_count_query, &[]) {
Ok(r) => match r.try_get::<&str, i64>("count") {
Ok(num_ws) => {
if num_ws > 0 {
compute.update_last_active(Some(Utc::now()));
continue;
}
}
Err(e) => {
warn!("failed to parse walsenders count: {:?}", e);
continue;
}
},
Err(e) => {
warn!("failed to get list of walsenders: {:?}", e);
continue;
}
}
//
// Don't suspend compute if there is an active logical replication subscription
//
// `where pid is not null` to filter out read only computes and subscription on branches
//
let logical_subscriptions_query =
"select count(*) from pg_stat_subscription where pid is not null;";
match cli.query_one(logical_subscriptions_query, &[]) {
Ok(row) => match row.try_get::<&str, i64>("count") {
Ok(num_subscribers) => {
if num_subscribers > 0 {
compute.update_last_active(Some(Utc::now()));
continue;
}
}
Err(e) => {
warn!("failed to parse `pg_stat_subscription` count: {:?}", e);
continue;
}
},
Err(e) => {
warn!(
"failed to get list of active logical replication subscriptions: {:?}",
e
loop {
match &mut client {
Ok(cli) => {
if cli.is_closed() {
info!(
downtime_info = self.downtime_info(),
"connection to Postgres is closed, trying to reconnect"
);
continue;
}
}
//
// Do not suspend compute if autovacuum is running
//
let autovacuum_count_query = "select count(*) from pg_stat_activity where backend_type = 'autovacuum worker'";
match cli.query_one(autovacuum_count_query, &[]) {
Ok(r) => match r.try_get::<&str, i64>("count") {
Ok(num_workers) => {
if num_workers > 0 {
compute.update_last_active(Some(Utc::now()));
continue;
self.report_down();
// Connection is closed, reconnect and try again.
client = conf.connect(NoTls);
} else {
match self.check(cli) {
Ok(_) => {
self.report_up();
self.compute.update_last_active(self.last_active);
}
Err(e) => {
// Although we have many places where we can return errors in `check()`,
// normally it shouldn't happen. I.e., we will likely return error if
// connection got broken, query timed out, Postgres returned invalid data, etc.
// In all such cases it's suspicious, so let's report this as downtime.
self.report_down();
error!(
downtime_info = self.downtime_info(),
"could not check Postgres: {}", e
);
// Reconnect to Postgres just in case. During tests, I noticed
// that queries in `check()` can fail with `connection closed`,
// but `cli.is_closed()` above doesn't detect it. Even if old
// connection is still alive, it will be dropped when we reassign
// `client` to a new connection.
client = conf.connect(NoTls);
}
}
Err(e) => {
warn!("failed to parse autovacuum workers count: {:?}", e);
continue;
}
},
Err(e) => {
warn!("failed to get list of autovacuum workers: {:?}", e);
continue;
}
}
}
Err(e) => {
debug!("could not connect to Postgres: {}, retrying", e);
Err(e) => {
info!(
downtime_info = self.downtime_info(),
"could not connect to Postgres: {}, retrying", e
);
self.report_down();
// Establish a new connection and try again.
client = conf.connect(NoTls);
// Establish a new connection and try again.
client = conf.connect(NoTls);
}
}
// Reset the `last_checked` timestamp and sleep before the next iteration.
self.last_checked = Utc::now();
thread::sleep(MONITOR_CHECK_INTERVAL);
}
}
#[instrument(skip_all)]
fn check(&mut self, cli: &mut Client) -> anyhow::Result<()> {
// This is new logic, only enable if the feature flag is set.
// TODO: remove this once we are sure that it works OR drop it altogether.
if self.experimental {
// Check if the total active time or sessions across all databases has changed.
// If it did, it means that user executed some queries. In theory, it can even go down if
// some databases were dropped, but it's still user activity.
match get_database_stats(cli) {
Ok((active_time, sessions)) => {
let mut detected_activity = false;
if let Some(prev_active_time) = self.active_time {
if active_time != prev_active_time {
detected_activity = true;
}
}
self.active_time = Some(active_time);
if let Some(prev_sessions) = self.sessions {
if sessions != prev_sessions {
detected_activity = true;
}
}
self.sessions = Some(sessions);
if detected_activity {
// Update the last active time and continue, we don't need to
// check backends state change.
self.last_active = Some(Utc::now());
return Ok(());
}
}
Err(e) => {
return Err(anyhow::anyhow!("could not get database statistics: {}", e));
}
}
}
// If database statistics are the same, check all backends for state changes.
// Maybe there are some with more recent activity. `get_backends_state_change()`
// can return None or stale timestamp, so it's `compute.update_last_active()`
// responsibility to check if the new timestamp is more recent than the current one.
// This helps us to discover new sessions that have not done anything yet.
match get_backends_state_change(cli) {
Ok(last_active) => match (last_active, self.last_active) {
(Some(last_active), Some(prev_last_active)) => {
if last_active > prev_last_active {
self.last_active = Some(last_active);
return Ok(());
}
}
(Some(last_active), None) => {
self.last_active = Some(last_active);
return Ok(());
}
_ => {}
},
Err(e) => {
return Err(anyhow::anyhow!(
"could not get backends state change: {}",
e
));
}
}
// If there are existing (logical) walsenders, do not suspend.
//
// N.B. walproposer doesn't currently show up in pg_stat_replication,
// but protect if it will.
const WS_COUNT_QUERY: &str =
"select count(*) from pg_stat_replication where application_name != 'walproposer';";
match cli.query_one(WS_COUNT_QUERY, &[]) {
Ok(r) => match r.try_get::<&str, i64>("count") {
Ok(num_ws) => {
if num_ws > 0 {
self.last_active = Some(Utc::now());
return Ok(());
}
}
Err(e) => {
let err: anyhow::Error = e.into();
return Err(err.context("failed to parse walsenders count"));
}
},
Err(e) => {
return Err(anyhow::anyhow!("failed to get list of walsenders: {}", e));
}
}
// Don't suspend compute if there is an active logical replication subscription
//
// `where pid is not null` to filter out read only computes and subscription on branches
const LOGICAL_SUBSCRIPTIONS_QUERY: &str =
"select count(*) from pg_stat_subscription where pid is not null;";
match cli.query_one(LOGICAL_SUBSCRIPTIONS_QUERY, &[]) {
Ok(row) => match row.try_get::<&str, i64>("count") {
Ok(num_subscribers) => {
if num_subscribers > 0 {
self.last_active = Some(Utc::now());
return Ok(());
}
}
Err(e) => {
return Err(anyhow::anyhow!(
"failed to parse 'pg_stat_subscription' count: {}",
e
));
}
},
Err(e) => {
return Err(anyhow::anyhow!(
"failed to get list of active logical replication subscriptions: {}",
e
));
}
}
// Do not suspend compute if autovacuum is running
const AUTOVACUUM_COUNT_QUERY: &str =
"select count(*) from pg_stat_activity where backend_type = 'autovacuum worker'";
match cli.query_one(AUTOVACUUM_COUNT_QUERY, &[]) {
Ok(r) => match r.try_get::<&str, i64>("count") {
Ok(num_workers) => {
if num_workers > 0 {
self.last_active = Some(Utc::now());
return Ok(());
};
}
Err(e) => {
return Err(anyhow::anyhow!(
"failed to parse autovacuum workers count: {}",
e
));
}
},
Err(e) => {
return Err(anyhow::anyhow!(
"failed to get list of autovacuum workers: {}",
e
));
}
}
Ok(())
}
}
@@ -315,9 +412,24 @@ fn get_backends_state_change(cli: &mut Client) -> anyhow::Result<Option<DateTime
/// Launch a separate compute monitor thread and return its `JoinHandle`.
pub fn launch_monitor(compute: &Arc<ComputeNode>) -> thread::JoinHandle<()> {
let compute = Arc::clone(compute);
let experimental = compute.has_feature(ComputeFeature::ActivityMonitorExperimental);
let now = Utc::now();
let mut monitor = ComputeMonitor {
compute,
last_active: None,
last_checked: now,
last_up: now,
active_time: None,
sessions: None,
experimental,
};
let span = span!(Level::INFO, "compute_monitor");
thread::Builder::new()
.name("compute-monitor".into())
.spawn(move || watch_compute_activity(&compute))
.spawn(move || {
let _enter = span.enter();
monitor.run();
})
.expect("cannot launch compute monitor thread")
}

View File

@@ -50,13 +50,13 @@ fn restart_rsyslog() -> Result<()> {
pub fn configure_audit_rsyslog(
log_directory: String,
tag: &str,
tag: Option<String>,
remote_endpoint: &str,
) -> Result<()> {
let config_content: String = format!(
include_str!("config_template/compute_audit_rsyslog_template.conf"),
log_directory = log_directory,
tag = tag,
tag = tag.unwrap_or("".to_string()),
remote_endpoint = remote_endpoint
);

View File

@@ -3,9 +3,8 @@ use std::path::Path;
use anyhow::{Result, anyhow, bail};
use compute_api::responses::{
ComputeCtlConfig, ControlPlaneComputeStatus, ControlPlaneSpecResponse,
ComputeConfig, ControlPlaneComputeStatus, ControlPlaneConfigResponse,
};
use compute_api::spec::ComputeSpec;
use reqwest::StatusCode;
use tokio_postgres::Client;
use tracing::{error, info, instrument};
@@ -21,7 +20,7 @@ use crate::params::PG_HBA_ALL_MD5;
fn do_control_plane_request(
uri: &str,
jwt: &str,
) -> Result<ControlPlaneSpecResponse, (bool, String, String)> {
) -> Result<ControlPlaneConfigResponse, (bool, String, String)> {
let resp = reqwest::blocking::Client::new()
.get(uri)
.header("Authorization", format!("Bearer {}", jwt))
@@ -29,14 +28,14 @@ fn do_control_plane_request(
.map_err(|e| {
(
true,
format!("could not perform spec request to control plane: {:?}", e),
format!("could not perform request to control plane: {:?}", e),
UNKNOWN_HTTP_STATUS.to_string(),
)
})?;
let status = resp.status();
match status {
StatusCode::OK => match resp.json::<ControlPlaneSpecResponse>() {
StatusCode::OK => match resp.json::<ControlPlaneConfigResponse>() {
Ok(spec_resp) => Ok(spec_resp),
Err(e) => Err((
true,
@@ -69,40 +68,35 @@ fn do_control_plane_request(
}
}
/// Request spec from the control-plane by compute_id. If `NEON_CONTROL_PLANE_TOKEN`
/// env variable is set, it will be used for authorization.
pub fn get_spec_from_control_plane(
base_uri: &str,
compute_id: &str,
) -> Result<(Option<ComputeSpec>, ComputeCtlConfig)> {
/// Request config from the control-plane by compute_id. If
/// `NEON_CONTROL_PLANE_TOKEN` env variable is set, it will be used for
/// authorization.
pub fn get_config_from_control_plane(base_uri: &str, compute_id: &str) -> Result<ComputeConfig> {
let cp_uri = format!("{base_uri}/compute/api/v2/computes/{compute_id}/spec");
let jwt: String = match std::env::var("NEON_CONTROL_PLANE_TOKEN") {
Ok(v) => v,
Err(_) => "".to_string(),
};
let jwt: String = std::env::var("NEON_CONTROL_PLANE_TOKEN").unwrap_or_default();
let mut attempt = 1;
info!("getting spec from control plane: {}", cp_uri);
info!("getting config from control plane: {}", cp_uri);
// Do 3 attempts to get spec from the control plane using the following logic:
// - network error -> then retry
// - compute id is unknown or any other error -> bail out
// - no spec for compute yet (Empty state) -> return Ok(None)
// - got spec -> return Ok(Some(spec))
// - got config -> return Ok(Some(config))
while attempt < 4 {
let result = match do_control_plane_request(&cp_uri, &jwt) {
Ok(spec_resp) => {
Ok(config_resp) => {
CPLANE_REQUESTS_TOTAL
.with_label_values(&[
CPlaneRequestRPC::GetSpec.as_str(),
CPlaneRequestRPC::GetConfig.as_str(),
&StatusCode::OK.to_string(),
])
.inc();
match spec_resp.status {
ControlPlaneComputeStatus::Empty => Ok((None, spec_resp.compute_ctl_config)),
match config_resp.status {
ControlPlaneComputeStatus::Empty => Ok(config_resp.into()),
ControlPlaneComputeStatus::Attached => {
if let Some(spec) = spec_resp.spec {
Ok((Some(spec), spec_resp.compute_ctl_config))
if config_resp.spec.is_some() {
Ok(config_resp.into())
} else {
bail!("compute is attached, but spec is empty")
}
@@ -111,7 +105,7 @@ pub fn get_spec_from_control_plane(
}
Err((retry, msg, status)) => {
CPLANE_REQUESTS_TOTAL
.with_label_values(&[CPlaneRequestRPC::GetSpec.as_str(), &status])
.with_label_values(&[CPlaneRequestRPC::GetConfig.as_str(), &status])
.inc();
if retry {
Err(anyhow!(msg))
@@ -122,7 +116,7 @@ pub fn get_spec_from_control_plane(
};
if let Err(e) = &result {
error!("attempt {} to get spec failed with: {}", attempt, e);
error!("attempt {} to get config failed with: {}", attempt, e);
} else {
return result;
}
@@ -133,13 +127,13 @@ pub fn get_spec_from_control_plane(
// All attempts failed, return error.
Err(anyhow::anyhow!(
"Exhausted all attempts to retrieve the spec from the control plane"
"Exhausted all attempts to retrieve the config from the control plane"
))
}
/// Check `pg_hba.conf` and update if needed to allow external connections.
pub fn update_pg_hba(pgdata_path: &Path) -> Result<()> {
// XXX: consider making it a part of spec.json
// XXX: consider making it a part of config.json
let pghba_path = pgdata_path.join("pg_hba.conf");
if config::line_in_file(&pghba_path, PG_HBA_ALL_MD5)? {
@@ -153,7 +147,7 @@ pub fn update_pg_hba(pgdata_path: &Path) -> Result<()> {
/// Create a standby.signal file
pub fn add_standby_signal(pgdata_path: &Path) -> Result<()> {
// XXX: consider making it a part of spec.json
// XXX: consider making it a part of config.json
let signalfile = pgdata_path.join("standby.signal");
if !signalfile.exists() {

View File

@@ -278,12 +278,12 @@ impl ComputeNode {
// so that all config operations are audit logged.
match spec.audit_log_level
{
ComputeAudit::Hipaa => {
ComputeAudit::Hipaa | ComputeAudit::Extended | ComputeAudit::Full => {
phases.push(CreatePgauditExtension);
phases.push(CreatePgauditlogtofileExtension);
phases.push(DisablePostgresDBPgAudit);
}
ComputeAudit::Log => {
ComputeAudit::Log | ComputeAudit::Base => {
phases.push(CreatePgauditExtension);
phases.push(DisablePostgresDBPgAudit);
}

View File

@@ -6,13 +6,16 @@ license.workspace = true
[dependencies]
anyhow.workspace = true
base64.workspace = true
camino.workspace = true
clap.workspace = true
comfy-table.workspace = true
futures.workspace = true
humantime.workspace = true
jsonwebtoken.workspace = true
nix.workspace = true
once_cell.workspace = true
pem.workspace = true
humantime-serde.workspace = true
hyper0.workspace = true
regex.workspace = true
@@ -20,6 +23,8 @@ reqwest = { workspace = true, features = ["blocking", "json"] }
scopeguard.workspace = true
serde.workspace = true
serde_json.workspace = true
sha2.workspace = true
spki.workspace = true
thiserror.workspace = true
toml.workspace = true
toml_edit.workspace = true

View File

@@ -18,12 +18,11 @@ use anyhow::{Context, Result, anyhow, bail};
use clap::Parser;
use compute_api::spec::ComputeMode;
use control_plane::endpoint::ComputeControlPlane;
use control_plane::endpoint_storage::{ENDPOINT_STORAGE_DEFAULT_PORT, EndpointStorage};
use control_plane::local_env::{
InitForceMode, LocalEnv, NeonBroker, NeonLocalInitConf, NeonLocalInitPageserverConf,
ObjectStorageConf, SafekeeperConf,
EndpointStorageConf, InitForceMode, LocalEnv, NeonBroker, NeonLocalInitConf,
NeonLocalInitPageserverConf, SafekeeperConf,
};
use control_plane::object_storage::OBJECT_STORAGE_DEFAULT_PORT;
use control_plane::object_storage::ObjectStorage;
use control_plane::pageserver::PageServerNode;
use control_plane::safekeeper::SafekeeperNode;
use control_plane::storage_controller::{
@@ -63,7 +62,7 @@ const DEFAULT_PAGESERVER_ID: NodeId = NodeId(1);
const DEFAULT_BRANCH_NAME: &str = "main";
project_git_version!(GIT_VERSION);
const DEFAULT_PG_VERSION: u32 = 16;
const DEFAULT_PG_VERSION: u32 = 17;
const DEFAULT_PAGESERVER_CONTROL_PLANE_API: &str = "http://127.0.0.1:1234/upcall/v1/";
@@ -93,7 +92,7 @@ enum NeonLocalCmd {
#[command(subcommand)]
Safekeeper(SafekeeperCmd),
#[command(subcommand)]
ObjectStorage(ObjectStorageCmd),
EndpointStorage(EndpointStorageCmd),
#[command(subcommand)]
Endpoint(EndpointCmd),
#[command(subcommand)]
@@ -460,14 +459,14 @@ enum SafekeeperCmd {
#[derive(clap::Subcommand)]
#[clap(about = "Manage object storage")]
enum ObjectStorageCmd {
Start(ObjectStorageStartCmd),
Stop(ObjectStorageStopCmd),
enum EndpointStorageCmd {
Start(EndpointStorageStartCmd),
Stop(EndpointStorageStopCmd),
}
#[derive(clap::Args)]
#[clap(about = "Start object storage")]
struct ObjectStorageStartCmd {
struct EndpointStorageStartCmd {
#[clap(short = 't', long, help = "timeout until we fail the command")]
#[arg(default_value = "10s")]
start_timeout: humantime::Duration,
@@ -475,7 +474,7 @@ struct ObjectStorageStartCmd {
#[derive(clap::Args)]
#[clap(about = "Stop object storage")]
struct ObjectStorageStopCmd {
struct EndpointStorageStopCmd {
#[arg(value_enum, default_value = "fast")]
#[clap(
short = 'm',
@@ -552,6 +551,7 @@ enum EndpointCmd {
Start(EndpointStartCmdArgs),
Reconfigure(EndpointReconfigureCmdArgs),
Stop(EndpointStopCmdArgs),
GenerateJwt(EndpointGenerateJwtCmdArgs),
}
#[derive(clap::Args)]
@@ -699,6 +699,13 @@ struct EndpointStopCmdArgs {
mode: String,
}
#[derive(clap::Args)]
#[clap(about = "Generate a JWT for an endpoint")]
struct EndpointGenerateJwtCmdArgs {
#[clap(help = "Postgres endpoint id")]
endpoint_id: String,
}
#[derive(clap::Subcommand)]
#[clap(about = "Manage neon_local branch name mappings")]
enum MappingsCmd {
@@ -789,7 +796,9 @@ fn main() -> Result<()> {
}
NeonLocalCmd::StorageBroker(subcmd) => rt.block_on(handle_storage_broker(&subcmd, env)),
NeonLocalCmd::Safekeeper(subcmd) => rt.block_on(handle_safekeeper(&subcmd, env)),
NeonLocalCmd::ObjectStorage(subcmd) => rt.block_on(handle_object_storage(&subcmd, env)),
NeonLocalCmd::EndpointStorage(subcmd) => {
rt.block_on(handle_endpoint_storage(&subcmd, env))
}
NeonLocalCmd::Endpoint(subcmd) => rt.block_on(handle_endpoint(&subcmd, env)),
NeonLocalCmd::Mappings(subcmd) => handle_mappings(&subcmd, env),
};
@@ -1006,8 +1015,8 @@ fn handle_init(args: &InitCmdArgs) -> anyhow::Result<LocalEnv> {
}
})
.collect(),
object_storage: ObjectStorageConf {
port: OBJECT_STORAGE_DEFAULT_PORT,
endpoint_storage: EndpointStorageConf {
port: ENDPOINT_STORAGE_DEFAULT_PORT,
},
pg_distrib_dir: None,
neon_distrib_dir: None,
@@ -1528,6 +1537,16 @@ async fn handle_endpoint(subcmd: &EndpointCmd, env: &local_env::LocalEnv) -> Res
.with_context(|| format!("postgres endpoint {endpoint_id} is not found"))?;
endpoint.stop(&args.mode, args.destroy)?;
}
EndpointCmd::GenerateJwt(args) => {
let endpoint_id = &args.endpoint_id;
let endpoint = cplane
.endpoints
.get(endpoint_id)
.with_context(|| format!("postgres endpoint {endpoint_id} is not found"))?;
let jwt = endpoint.generate_jwt()?;
print!("{jwt}");
}
}
Ok(())
@@ -1717,12 +1736,15 @@ async fn handle_safekeeper(subcmd: &SafekeeperCmd, env: &local_env::LocalEnv) ->
Ok(())
}
async fn handle_object_storage(subcmd: &ObjectStorageCmd, env: &local_env::LocalEnv) -> Result<()> {
use ObjectStorageCmd::*;
let storage = ObjectStorage::from_env(env);
async fn handle_endpoint_storage(
subcmd: &EndpointStorageCmd,
env: &local_env::LocalEnv,
) -> Result<()> {
use EndpointStorageCmd::*;
let storage = EndpointStorage::from_env(env);
// In tests like test_forward_compatibility or test_graceful_cluster_restart
// old neon binaries (without object_storage) are present
// old neon binaries (without endpoint_storage) are present
if !storage.bin.exists() {
eprintln!(
"{} binary not found. Ignore if this is a compatibility test",
@@ -1732,13 +1754,13 @@ async fn handle_object_storage(subcmd: &ObjectStorageCmd, env: &local_env::Local
}
match subcmd {
Start(ObjectStorageStartCmd { start_timeout }) => {
Start(EndpointStorageStartCmd { start_timeout }) => {
if let Err(e) = storage.start(start_timeout).await {
eprintln!("object_storage start failed: {e}");
eprintln!("endpoint_storage start failed: {e}");
exit(1);
}
}
Stop(ObjectStorageStopCmd { stop_mode }) => {
Stop(EndpointStorageStopCmd { stop_mode }) => {
let immediate = match stop_mode {
StopMode::Fast => false,
StopMode::Immediate => true,
@@ -1848,10 +1870,10 @@ async fn handle_start_all_impl(
}
js.spawn(async move {
ObjectStorage::from_env(env)
EndpointStorage::from_env(env)
.start(&retry_timeout)
.await
.map_err(|e| e.context("start object_storage"))
.map_err(|e| e.context("start endpoint_storage"))
});
})();
@@ -1950,9 +1972,9 @@ async fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
}
}
let storage = ObjectStorage::from_env(env);
let storage = EndpointStorage::from_env(env);
if let Err(e) = storage.stop(immediate) {
eprintln!("object_storage stop failed: {:#}", e);
eprintln!("endpoint_storage stop failed: {:#}", e);
}
for ps_conf in &env.pageservers {

View File

@@ -29,7 +29,7 @@
//! compute.log - log output of `compute_ctl` and `postgres`
//! endpoint.json - serialized `EndpointConf` struct
//! postgresql.conf - postgresql settings
//! spec.json - passed to `compute_ctl`
//! config.json - passed to `compute_ctl`
//! pgdata/
//! postgresql.conf - copy of postgresql.conf created by `compute_ctl`
//! zenith.signal
@@ -42,20 +42,30 @@ use std::path::PathBuf;
use std::process::Command;
use std::str::FromStr;
use std::sync::Arc;
use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
use std::time::{Duration, Instant};
use anyhow::{Context, Result, anyhow, bail};
use compute_api::requests::ConfigurationRequest;
use compute_api::responses::{ComputeCtlConfig, ComputeStatus, ComputeStatusResponse};
use compute_api::requests::{ComputeClaims, ConfigurationRequest};
use compute_api::responses::{
ComputeConfig, ComputeCtlConfig, ComputeStatus, ComputeStatusResponse, TlsConfig,
};
use compute_api::spec::{
Cluster, ComputeAudit, ComputeFeature, ComputeMode, ComputeSpec, Database, PgIdent,
RemoteExtSpec, Role,
};
use jsonwebtoken::jwk::{
AlgorithmParameters, CommonParameters, EllipticCurve, Jwk, JwkSet, KeyAlgorithm, KeyOperations,
OctetKeyPairParameters, OctetKeyPairType, PublicKeyUse,
};
use nix::sys::signal::{Signal, kill};
use pageserver_api::shard::ShardStripeSize;
use pem::Pem;
use reqwest::header::CONTENT_TYPE;
use safekeeper_api::membership::SafekeeperGeneration;
use serde::{Deserialize, Serialize};
use sha2::{Digest, Sha256};
use spki::der::Decode;
use spki::{SubjectPublicKeyInfo, SubjectPublicKeyInfoRef};
use tracing::debug;
use url::Host;
use utils::id::{NodeId, TenantId, TimelineId};
@@ -80,6 +90,7 @@ pub struct EndpointConf {
drop_subscriptions_before_start: bool,
features: Vec<ComputeFeature>,
cluster: Option<Cluster>,
compute_ctl_config: ComputeCtlConfig,
}
//
@@ -135,6 +146,37 @@ impl ComputeControlPlane {
.unwrap_or(self.base_port)
}
/// Create a JSON Web Key Set. This ideally matches the way we create a JWKS
/// from the production control plane.
fn create_jwks_from_pem(pem: &Pem) -> Result<JwkSet> {
let spki: SubjectPublicKeyInfoRef = SubjectPublicKeyInfo::from_der(pem.contents())?;
let public_key = spki.subject_public_key.raw_bytes();
let mut hasher = Sha256::new();
hasher.update(public_key);
let key_hash = hasher.finalize();
Ok(JwkSet {
keys: vec![Jwk {
common: CommonParameters {
public_key_use: Some(PublicKeyUse::Signature),
key_operations: Some(vec![KeyOperations::Verify]),
key_algorithm: Some(KeyAlgorithm::EdDSA),
key_id: Some(base64::encode_config(key_hash, base64::URL_SAFE_NO_PAD)),
x509_url: None::<String>,
x509_chain: None::<Vec<String>>,
x509_sha1_fingerprint: None::<String>,
x509_sha256_fingerprint: None::<String>,
},
algorithm: AlgorithmParameters::OctetKeyPair(OctetKeyPairParameters {
key_type: OctetKeyPairType::OctetKeyPair,
curve: EllipticCurve::Ed25519,
x: base64::encode_config(public_key, base64::URL_SAFE_NO_PAD),
}),
}],
})
}
#[allow(clippy::too_many_arguments)]
pub fn new_endpoint(
&mut self,
@@ -152,6 +194,10 @@ impl ComputeControlPlane {
let pg_port = pg_port.unwrap_or_else(|| self.get_port());
let external_http_port = external_http_port.unwrap_or_else(|| self.get_port() + 1);
let internal_http_port = internal_http_port.unwrap_or_else(|| external_http_port + 1);
let compute_ctl_config = ComputeCtlConfig {
jwks: Self::create_jwks_from_pem(&self.env.read_public_key()?)?,
tls: None::<TlsConfig>,
};
let ep = Arc::new(Endpoint {
endpoint_id: endpoint_id.to_owned(),
pg_address: SocketAddr::new(IpAddr::from(Ipv4Addr::LOCALHOST), pg_port),
@@ -179,6 +225,7 @@ impl ComputeControlPlane {
reconfigure_concurrency: 1,
features: vec![],
cluster: None,
compute_ctl_config: compute_ctl_config.clone(),
});
ep.create_endpoint_dir()?;
@@ -198,6 +245,7 @@ impl ComputeControlPlane {
reconfigure_concurrency: 1,
features: vec![],
cluster: None,
compute_ctl_config,
})?,
)?;
std::fs::write(
@@ -240,7 +288,6 @@ impl ComputeControlPlane {
///////////////////////////////////////////////////////////////////////////////
#[derive(Debug)]
pub struct Endpoint {
/// used as the directory name
endpoint_id: String,
@@ -269,6 +316,9 @@ pub struct Endpoint {
features: Vec<ComputeFeature>,
// Cluster settings
cluster: Option<Cluster>,
/// The compute_ctl config for the endpoint's compute.
compute_ctl_config: ComputeCtlConfig,
}
#[derive(PartialEq, Eq)]
@@ -331,6 +381,7 @@ impl Endpoint {
drop_subscriptions_before_start: conf.drop_subscriptions_before_start,
features: conf.features,
cluster: conf.cluster,
compute_ctl_config: conf.compute_ctl_config,
})
}
@@ -578,6 +629,13 @@ impl Endpoint {
Ok(safekeeper_connstrings)
}
/// Generate a JWT with the correct claims.
pub fn generate_jwt(&self) -> Result<String> {
self.env.generate_auth_token(&ComputeClaims {
compute_id: self.endpoint_id.clone(),
})
}
#[allow(clippy::too_many_arguments)]
pub async fn start(
&self,
@@ -619,87 +677,97 @@ impl Endpoint {
remote_extensions = None;
};
// Create spec file
let mut spec = ComputeSpec {
skip_pg_catalog_updates: self.skip_pg_catalog_updates,
format_version: 1.0,
operation_uuid: None,
features: self.features.clone(),
swap_size_bytes: None,
disk_quota_bytes: None,
disable_lfc_resizing: None,
cluster: Cluster {
cluster_id: None, // project ID: not used
name: None, // project name: not used
state: None,
roles: if create_test_user {
vec![Role {
// Create config file
let config = {
let mut spec = ComputeSpec {
skip_pg_catalog_updates: self.skip_pg_catalog_updates,
format_version: 1.0,
operation_uuid: None,
features: self.features.clone(),
swap_size_bytes: None,
disk_quota_bytes: None,
disable_lfc_resizing: None,
cluster: Cluster {
cluster_id: None, // project ID: not used
name: None, // project name: not used
state: None,
roles: if create_test_user {
vec![Role {
name: PgIdent::from_str("test").unwrap(),
encrypted_password: None,
options: None,
}]
} else {
Vec::new()
},
databases: if create_test_user {
vec![Database {
name: PgIdent::from_str("neondb").unwrap(),
owner: PgIdent::from_str("test").unwrap(),
options: None,
restrict_conn: false,
invalid: false,
}]
} else {
Vec::new()
},
settings: None,
postgresql_conf: Some(postgresql_conf.clone()),
},
delta_operations: None,
tenant_id: Some(self.tenant_id),
timeline_id: Some(self.timeline_id),
project_id: None,
branch_id: None,
endpoint_id: Some(self.endpoint_id.clone()),
mode: self.mode,
pageserver_connstring: Some(pageserver_connstring),
safekeepers_generation: safekeepers_generation.map(|g| g.into_inner()),
safekeeper_connstrings,
storage_auth_token: auth_token.clone(),
remote_extensions,
pgbouncer_settings: None,
shard_stripe_size: Some(shard_stripe_size),
local_proxy_config: None,
reconfigure_concurrency: self.reconfigure_concurrency,
drop_subscriptions_before_start: self.drop_subscriptions_before_start,
audit_log_level: ComputeAudit::Disabled,
logs_export_host: None::<String>,
};
// this strange code is needed to support respec() in tests
if self.cluster.is_some() {
debug!("Cluster is already set in the endpoint spec, using it");
spec.cluster = self.cluster.clone().unwrap();
debug!("spec.cluster {:?}", spec.cluster);
// fill missing fields again
if create_test_user {
spec.cluster.roles.push(Role {
name: PgIdent::from_str("test").unwrap(),
encrypted_password: None,
options: None,
}]
} else {
Vec::new()
},
databases: if create_test_user {
vec![Database {
});
spec.cluster.databases.push(Database {
name: PgIdent::from_str("neondb").unwrap(),
owner: PgIdent::from_str("test").unwrap(),
options: None,
restrict_conn: false,
invalid: false,
}]
} else {
Vec::new()
},
settings: None,
postgresql_conf: Some(postgresql_conf.clone()),
},
delta_operations: None,
tenant_id: Some(self.tenant_id),
timeline_id: Some(self.timeline_id),
mode: self.mode,
pageserver_connstring: Some(pageserver_connstring),
safekeepers_generation: safekeepers_generation.map(|g| g.into_inner()),
safekeeper_connstrings,
storage_auth_token: auth_token.clone(),
remote_extensions,
pgbouncer_settings: None,
shard_stripe_size: Some(shard_stripe_size),
local_proxy_config: None,
reconfigure_concurrency: self.reconfigure_concurrency,
drop_subscriptions_before_start: self.drop_subscriptions_before_start,
audit_log_level: ComputeAudit::Disabled,
logs_export_host: None::<String>,
});
}
spec.cluster.postgresql_conf = Some(postgresql_conf);
}
ComputeConfig {
spec: Some(spec),
compute_ctl_config: self.compute_ctl_config.clone(),
}
};
// this strange code is needed to support respec() in tests
if self.cluster.is_some() {
debug!("Cluster is already set in the endpoint spec, using it");
spec.cluster = self.cluster.clone().unwrap();
debug!("spec.cluster {:?}", spec.cluster);
// fill missing fields again
if create_test_user {
spec.cluster.roles.push(Role {
name: PgIdent::from_str("test").unwrap(),
encrypted_password: None,
options: None,
});
spec.cluster.databases.push(Database {
name: PgIdent::from_str("neondb").unwrap(),
owner: PgIdent::from_str("test").unwrap(),
options: None,
restrict_conn: false,
invalid: false,
});
}
spec.cluster.postgresql_conf = Some(postgresql_conf);
}
let spec_path = self.endpoint_path().join("spec.json");
std::fs::write(spec_path, serde_json::to_string_pretty(&spec)?)?;
let config_path = self.endpoint_path().join("config.json");
std::fs::write(config_path, serde_json::to_string_pretty(&config)?)?;
// Open log file. We'll redirect the stdout and stderr of `compute_ctl` to it.
let logfile = std::fs::OpenOptions::new()
@@ -725,10 +793,8 @@ impl Endpoint {
])
.args(["--pgdata", self.pgdata().to_str().unwrap()])
.args(["--connstr", &conn_str])
.args([
"--spec-path",
self.endpoint_path().join("spec.json").to_str().unwrap(),
])
.arg("--config")
.arg(self.endpoint_path().join("config.json").as_os_str())
.args([
"--pgbin",
self.env
@@ -739,16 +805,7 @@ impl Endpoint {
])
// TODO: It would be nice if we generated compute IDs with the same
// algorithm as the real control plane.
.args([
"--compute-id",
&format!(
"compute-{}",
SystemTime::now()
.duration_since(UNIX_EPOCH)
.unwrap()
.as_secs()
),
])
.args(["--compute-id", &self.endpoint_id])
.stdin(std::process::Stdio::null())
.stderr(logfile.try_clone()?)
.stdout(logfile);
@@ -846,6 +903,7 @@ impl Endpoint {
self.external_http_address.port()
),
)
.bearer_auth(self.generate_jwt()?)
.send()
.await?;
@@ -870,10 +928,12 @@ impl Endpoint {
stripe_size: Option<ShardStripeSize>,
safekeepers: Option<Vec<NodeId>>,
) -> Result<()> {
let mut spec: ComputeSpec = {
let spec_path = self.endpoint_path().join("spec.json");
let file = std::fs::File::open(spec_path)?;
serde_json::from_reader(file)?
let (mut spec, compute_ctl_config) = {
let config_path = self.endpoint_path().join("config.json");
let file = std::fs::File::open(config_path)?;
let config: ComputeConfig = serde_json::from_reader(file)?;
(config.spec.unwrap(), config.compute_ctl_config)
};
let postgresql_conf = self.read_postgresql_conf()?;
@@ -920,10 +980,11 @@ impl Endpoint {
self.external_http_address.port()
))
.header(CONTENT_TYPE.as_str(), "application/json")
.bearer_auth(self.generate_jwt()?)
.body(
serde_json::to_string(&ConfigurationRequest {
spec,
compute_ctl_config: ComputeCtlConfig::default(),
compute_ctl_config,
})
.unwrap(),
)

View File

@@ -1,34 +1,33 @@
use crate::background_process::{self, start_process, stop_process};
use crate::local_env::LocalEnv;
use anyhow::anyhow;
use anyhow::{Context, Result};
use camino::Utf8PathBuf;
use std::io::Write;
use std::time::Duration;
/// Directory within .neon which will be used by default for LocalFs remote storage.
pub const OBJECT_STORAGE_REMOTE_STORAGE_DIR: &str = "local_fs_remote_storage/object_storage";
pub const OBJECT_STORAGE_DEFAULT_PORT: u16 = 9993;
pub const ENDPOINT_STORAGE_REMOTE_STORAGE_DIR: &str = "local_fs_remote_storage/endpoint_storage";
pub const ENDPOINT_STORAGE_DEFAULT_PORT: u16 = 9993;
pub struct ObjectStorage {
pub struct EndpointStorage {
pub bin: Utf8PathBuf,
pub data_dir: Utf8PathBuf,
pub pemfile: Utf8PathBuf,
pub port: u16,
}
impl ObjectStorage {
pub fn from_env(env: &LocalEnv) -> ObjectStorage {
ObjectStorage {
bin: Utf8PathBuf::from_path_buf(env.object_storage_bin()).unwrap(),
data_dir: Utf8PathBuf::from_path_buf(env.object_storage_data_dir()).unwrap(),
impl EndpointStorage {
pub fn from_env(env: &LocalEnv) -> EndpointStorage {
EndpointStorage {
bin: Utf8PathBuf::from_path_buf(env.endpoint_storage_bin()).unwrap(),
data_dir: Utf8PathBuf::from_path_buf(env.endpoint_storage_data_dir()).unwrap(),
pemfile: Utf8PathBuf::from_path_buf(env.public_key_path.clone()).unwrap(),
port: env.object_storage.port,
port: env.endpoint_storage.port,
}
}
fn config_path(&self) -> Utf8PathBuf {
self.data_dir.join("object_storage.json")
self.data_dir.join("endpoint_storage.json")
}
fn listen_addr(&self) -> Utf8PathBuf {
@@ -49,7 +48,7 @@ impl ObjectStorage {
let cfg = Cfg {
listen: self.listen_addr(),
pemfile: parent.join(self.pemfile.clone()),
local_path: parent.join(OBJECT_STORAGE_REMOTE_STORAGE_DIR),
local_path: parent.join(ENDPOINT_STORAGE_REMOTE_STORAGE_DIR),
r#type: "LocalFs".to_string(),
};
std::fs::create_dir_all(self.config_path().parent().unwrap())?;
@@ -59,24 +58,19 @@ impl ObjectStorage {
}
pub async fn start(&self, retry_timeout: &Duration) -> Result<()> {
println!("Starting s3 proxy at {}", self.listen_addr());
println!("Starting endpoint_storage at {}", self.listen_addr());
std::io::stdout().flush().context("flush stdout")?;
let process_status_check = || async {
tokio::time::sleep(Duration::from_millis(500)).await;
let res = reqwest::Client::new()
.get(format!("http://{}/metrics", self.listen_addr()))
.send()
.await;
match res {
Ok(response) if response.status().is_success() => Ok(true),
Ok(_) => Err(anyhow!("Failed to query /metrics")),
Err(e) => Err(anyhow!("Failed to check node status: {e}")),
let res = reqwest::Client::new().get(format!("http://{}/metrics", self.listen_addr()));
match res.send().await {
Ok(res) => Ok(res.status().is_success()),
Err(_) => Ok(false),
}
};
let res = start_process(
"object_storage",
"endpoint_storage",
&self.data_dir.clone().into_std_path_buf(),
&self.bin.clone().into_std_path_buf(),
vec![self.config_path().to_string()],
@@ -94,14 +88,14 @@ impl ObjectStorage {
}
pub fn stop(&self, immediate: bool) -> anyhow::Result<()> {
stop_process(immediate, "object_storage", &self.pid_file())
stop_process(immediate, "endpoint_storage", &self.pid_file())
}
fn log_file(&self) -> Utf8PathBuf {
self.data_dir.join("object_storage.log")
self.data_dir.join("endpoint_storage.log")
}
fn pid_file(&self) -> Utf8PathBuf {
self.data_dir.join("object_storage.pid")
self.data_dir.join("endpoint_storage.pid")
}
}

View File

@@ -9,8 +9,8 @@
mod background_process;
pub mod broker;
pub mod endpoint;
pub mod endpoint_storage;
pub mod local_env;
pub mod object_storage;
pub mod pageserver;
pub mod postgresql_conf;
pub mod safekeeper;

View File

@@ -12,17 +12,18 @@ use std::{env, fs};
use anyhow::{Context, bail};
use clap::ValueEnum;
use pem::Pem;
use postgres_backend::AuthType;
use reqwest::Url;
use serde::{Deserialize, Serialize};
use utils::auth::encode_from_key_file;
use utils::id::{NodeId, TenantId, TenantTimelineId, TimelineId};
use crate::object_storage::{OBJECT_STORAGE_REMOTE_STORAGE_DIR, ObjectStorage};
use crate::endpoint_storage::{ENDPOINT_STORAGE_REMOTE_STORAGE_DIR, EndpointStorage};
use crate::pageserver::{PAGESERVER_REMOTE_STORAGE_DIR, PageServerNode};
use crate::safekeeper::SafekeeperNode;
pub const DEFAULT_PG_VERSION: u32 = 16;
pub const DEFAULT_PG_VERSION: u32 = 17;
//
// This data structures represents neon_local CLI config
@@ -56,6 +57,7 @@ pub struct LocalEnv {
// used to issue tokens during e.g pg start
pub private_key_path: PathBuf,
/// Path to environment's public key
pub public_key_path: PathBuf,
pub broker: NeonBroker,
@@ -70,7 +72,7 @@ pub struct LocalEnv {
pub safekeepers: Vec<SafekeeperConf>,
pub object_storage: ObjectStorageConf,
pub endpoint_storage: EndpointStorageConf,
// Control plane upcall API for pageserver: if None, we will not run storage_controller If set, this will
// be propagated into each pageserver's configuration.
@@ -108,7 +110,7 @@ pub struct OnDiskConfig {
)]
pub pageservers: Vec<PageServerConf>,
pub safekeepers: Vec<SafekeeperConf>,
pub object_storage: ObjectStorageConf,
pub endpoint_storage: EndpointStorageConf,
pub control_plane_api: Option<Url>,
pub control_plane_hooks_api: Option<Url>,
pub control_plane_compute_hook_api: Option<Url>,
@@ -142,7 +144,7 @@ pub struct NeonLocalInitConf {
pub storage_controller: Option<NeonStorageControllerConf>,
pub pageservers: Vec<NeonLocalInitPageserverConf>,
pub safekeepers: Vec<SafekeeperConf>,
pub object_storage: ObjectStorageConf,
pub endpoint_storage: EndpointStorageConf,
pub control_plane_api: Option<Url>,
pub control_plane_hooks_api: Option<Url>,
pub generate_local_ssl_certs: bool,
@@ -150,7 +152,7 @@ pub struct NeonLocalInitConf {
#[derive(Serialize, Default, Deserialize, PartialEq, Eq, Clone, Debug)]
#[serde(default)]
pub struct ObjectStorageConf {
pub struct EndpointStorageConf {
pub port: u16,
}
@@ -411,8 +413,8 @@ impl LocalEnv {
self.pg_dir(pg_version, "lib")
}
pub fn object_storage_bin(&self) -> PathBuf {
self.neon_distrib_dir.join("object_storage")
pub fn endpoint_storage_bin(&self) -> PathBuf {
self.neon_distrib_dir.join("endpoint_storage")
}
pub fn pageserver_bin(&self) -> PathBuf {
@@ -448,8 +450,8 @@ impl LocalEnv {
self.base_data_dir.join("safekeepers").join(data_dir_name)
}
pub fn object_storage_data_dir(&self) -> PathBuf {
self.base_data_dir.join("object_storage")
pub fn endpoint_storage_data_dir(&self) -> PathBuf {
self.base_data_dir.join("endpoint_storage")
}
pub fn get_pageserver_conf(&self, id: NodeId) -> anyhow::Result<&PageServerConf> {
@@ -613,7 +615,7 @@ impl LocalEnv {
control_plane_compute_hook_api: _,
branch_name_mappings,
generate_local_ssl_certs,
object_storage,
endpoint_storage,
} = on_disk_config;
LocalEnv {
base_data_dir: repopath.to_owned(),
@@ -630,7 +632,7 @@ impl LocalEnv {
control_plane_hooks_api,
branch_name_mappings,
generate_local_ssl_certs,
object_storage,
endpoint_storage,
}
};
@@ -740,7 +742,7 @@ impl LocalEnv {
control_plane_compute_hook_api: None,
branch_name_mappings: self.branch_name_mappings.clone(),
generate_local_ssl_certs: self.generate_local_ssl_certs,
object_storage: self.object_storage.clone(),
endpoint_storage: self.endpoint_storage.clone(),
},
)
}
@@ -758,11 +760,11 @@ impl LocalEnv {
// this function is used only for testing purposes in CLI e g generate tokens during init
pub fn generate_auth_token<S: Serialize>(&self, claims: &S) -> anyhow::Result<String> {
let private_key_path = self.get_private_key_path();
let key_data = fs::read(private_key_path)?;
encode_from_key_file(claims, &key_data)
let key = self.read_private_key()?;
encode_from_key_file(claims, &key)
}
/// Get the path to the private key.
pub fn get_private_key_path(&self) -> PathBuf {
if self.private_key_path.is_absolute() {
self.private_key_path.to_path_buf()
@@ -771,6 +773,29 @@ impl LocalEnv {
}
}
/// Get the path to the public key.
pub fn get_public_key_path(&self) -> PathBuf {
if self.public_key_path.is_absolute() {
self.public_key_path.to_path_buf()
} else {
self.base_data_dir.join(&self.public_key_path)
}
}
/// Read the contents of the private key file.
pub fn read_private_key(&self) -> anyhow::Result<Pem> {
let private_key_path = self.get_private_key_path();
let pem = pem::parse(fs::read(private_key_path)?)?;
Ok(pem)
}
/// Read the contents of the public key file.
pub fn read_public_key(&self) -> anyhow::Result<Pem> {
let public_key_path = self.get_public_key_path();
let pem = pem::parse(fs::read(public_key_path)?)?;
Ok(pem)
}
/// Materialize the [`NeonLocalInitConf`] to disk. Called during [`neon_local init`].
pub fn init(conf: NeonLocalInitConf, force: &InitForceMode) -> anyhow::Result<()> {
let base_path = base_path();
@@ -824,7 +849,7 @@ impl LocalEnv {
control_plane_api,
generate_local_ssl_certs,
control_plane_hooks_api,
object_storage,
endpoint_storage,
} = conf;
// Find postgres binaries.
@@ -876,7 +901,7 @@ impl LocalEnv {
control_plane_hooks_api,
branch_name_mappings: Default::default(),
generate_local_ssl_certs,
object_storage,
endpoint_storage,
};
if generate_local_ssl_certs {
@@ -904,13 +929,13 @@ impl LocalEnv {
.context("pageserver init failed")?;
}
ObjectStorage::from_env(&env)
EndpointStorage::from_env(&env)
.init()
.context("object storage init failed")?;
// setup remote remote location for default LocalFs remote storage
std::fs::create_dir_all(env.base_data_dir.join(PAGESERVER_REMOTE_STORAGE_DIR))?;
std::fs::create_dir_all(env.base_data_dir.join(OBJECT_STORAGE_REMOTE_STORAGE_DIR))?;
std::fs::create_dir_all(env.base_data_dir.join(ENDPOINT_STORAGE_REMOTE_STORAGE_DIR))?;
env.persist_config()
}
@@ -956,6 +981,7 @@ fn generate_auth_keys(private_key_path: &Path, public_key_path: &Path) -> anyhow
String::from_utf8_lossy(&keygen_output.stderr)
);
}
// Extract the public key from the private key file
//
// openssl pkey -in auth_private_key.pem -pubout -out auth_public_key.pem
@@ -972,6 +998,7 @@ fn generate_auth_keys(private_key_path: &Path, public_key_path: &Path) -> anyhow
String::from_utf8_lossy(&keygen_output.stderr)
);
}
Ok(())
}
@@ -980,7 +1007,7 @@ fn generate_ssl_ca_cert(cert_path: &Path, key_path: &Path) -> anyhow::Result<()>
// -out rootCA.crt -keyout rootCA.key
let keygen_output = Command::new("openssl")
.args([
"req", "-x509", "-newkey", "rsa:2048", "-nodes", "-days", "36500",
"req", "-x509", "-newkey", "ed25519", "-nodes", "-days", "36500",
])
.args(["-subj", "/CN=Neon Local CA"])
.args(["-out", cert_path.to_str().unwrap()])
@@ -1010,7 +1037,7 @@ fn generate_ssl_cert(
// -subj "/CN=localhost" -addext "subjectAltName=DNS:localhost,IP:127.0.0.1"
let keygen_output = Command::new("openssl")
.args(["req", "-new", "-nodes"])
.args(["-newkey", "rsa:2048"])
.args(["-newkey", "ed25519"])
.args(["-subj", "/CN=localhost"])
.args(["-addext", "subjectAltName=DNS:localhost,IP:127.0.0.1"])
.args(["-keyout", key_path.to_str().unwrap()])

View File

@@ -413,6 +413,11 @@ impl PageServerNode {
.map(serde_json::from_str)
.transpose()
.context("Failed to parse 'compaction_algorithm' json")?,
compaction_shard_ancestor: settings
.remove("compaction_shard_ancestor")
.map(|x| x.parse::<bool>())
.transpose()
.context("Failed to parse 'compaction_shard_ancestor' as a bool")?,
compaction_l0_first: settings
.remove("compaction_l0_first")
.map(|x| x.parse::<bool>())
@@ -535,6 +540,11 @@ impl PageServerNode {
.map(|x| x.parse::<bool>())
.transpose()
.context("Failed to parse 'gc_compaction_enabled' as bool")?,
gc_compaction_verification: settings
.remove("gc_compaction_verification")
.map(|x| x.parse::<bool>())
.transpose()
.context("Failed to parse 'gc_compaction_verification' as bool")?,
gc_compaction_initial_threshold_kb: settings
.remove("gc_compaction_initial_threshold_kb")
.map(|x| x.parse::<u64>())

View File

@@ -13,9 +13,12 @@ use pageserver_api::controller_api::{
NodeConfigureRequest, NodeDescribeResponse, NodeRegisterRequest, TenantCreateRequest,
TenantCreateResponse, TenantLocateResponse,
};
use pageserver_api::models::{TenantConfigRequest, TimelineCreateRequest, TimelineInfo};
use pageserver_api::models::{
TenantConfig, TenantConfigRequest, TimelineCreateRequest, TimelineInfo,
};
use pageserver_api::shard::TenantShardId;
use pageserver_client::mgmt_api::ResponseErrorMessageExt;
use pem::Pem;
use postgres_backend::AuthType;
use reqwest::{Certificate, Method};
use serde::de::DeserializeOwned;
@@ -32,8 +35,8 @@ use crate::local_env::{LocalEnv, NeonStorageControllerConf};
pub struct StorageController {
env: LocalEnv,
private_key: Option<Vec<u8>>,
public_key: Option<String>,
private_key: Option<Pem>,
public_key: Option<Pem>,
client: reqwest::Client,
config: NeonStorageControllerConf,
@@ -82,7 +85,8 @@ impl NeonStorageControllerStopArgs {
pub struct AttachHookRequest {
pub tenant_shard_id: TenantShardId,
pub node_id: Option<NodeId>,
pub generation_override: Option<i32>,
pub generation_override: Option<i32>, // only new tenants
pub config: Option<TenantConfig>, // only new tenants
}
#[derive(Serialize, Deserialize)]
@@ -113,7 +117,9 @@ impl StorageController {
AuthType::Trust => (None, None),
AuthType::NeonJWT => {
let private_key_path = env.get_private_key_path();
let private_key = fs::read(private_key_path).expect("failed to read private key");
let private_key =
pem::parse(fs::read(private_key_path).expect("failed to read private key"))
.expect("failed to parse PEM file");
// If pageserver auth is enabled, this implicitly enables auth for this service,
// using the same credentials.
@@ -135,9 +141,13 @@ impl StorageController {
.expect("Empty key dir")
.expect("Error reading key dir");
std::fs::read_to_string(dent.path()).expect("Can't read public key")
pem::parse(std::fs::read_to_string(dent.path()).expect("Can't read public key"))
.expect("Failed to parse PEM file")
} else {
std::fs::read_to_string(&public_key_path).expect("Can't read public key")
pem::parse(
std::fs::read_to_string(&public_key_path).expect("Can't read public key"),
)
.expect("Failed to parse PEM file")
};
(Some(private_key), Some(public_key))
}
@@ -805,6 +815,7 @@ impl StorageController {
tenant_shard_id,
node_id: Some(pageserver_id),
generation_override: None,
config: None,
};
let response = self

View File

@@ -45,9 +45,7 @@ allow = [
"ISC",
"MIT",
"MPL-2.0",
"OpenSSL",
"Unicode-3.0",
"Zlib",
]
confidence-threshold = 0.8
exceptions = [
@@ -56,14 +54,6 @@ exceptions = [
{ allow = ["Zlib"], name = "const_format", version = "*" },
]
[[licenses.clarify]]
name = "ring"
version = "*"
expression = "MIT AND ISC AND OpenSSL"
license-files = [
{ path = "LICENSE", hash = 0xbd0eed23 }
]
[licenses.private]
ignore = true
registries = []
@@ -116,7 +106,11 @@ name = "openssl"
unknown-registry = "warn"
unknown-git = "warn"
allow-registry = ["https://github.com/rust-lang/crates.io-index"]
allow-git = []
allow-git = [
# Crate pinned to commit in origin repo due to opentelemetry version.
# TODO: Remove this once crate is fetched from crates.io again.
"https://github.com/mattiapenati/tower-otel",
]
[sources.allow-org]
github = [

View File

@@ -1,4 +1,3 @@
# Example docker compose configuration
The configuration in this directory is used for testing Neon docker images: it is
@@ -8,3 +7,13 @@ you can experiment with a miniature Neon system, use `cargo neon` rather than co
This configuration does not start the storage controller, because the controller
needs a way to reconfigure running computes, and no such thing exists in this setup.
## Generating the JWKS for a compute
```shell
openssl genpkey -algorithm Ed25519 -out private-key.pem
openssl pkey -in private-key.pem -pubout -out public-key.pem
openssl pkey -pubin -inform pem -in public-key.pem -pubout -outform der -out public-key.der
key="$(xxd -plain -cols 32 -s -32 public-key.der)"
key_id="$(printf '%s' "$key" | sha256sum | awk '{ print $1 }' | basenc --base64url --wrap=0)"
x="$(printf '%s' "$key" | basenc --base64url --wrap=0)"
```

View File

@@ -0,0 +1,3 @@
-----BEGIN PRIVATE KEY-----
MC4CAQAwBQYDK2VwBCIEIOmnRbzt2AJ0d+S3aU1hiYOl/tXpvz1FmWBfwHYBgOma
-----END PRIVATE KEY-----

Binary file not shown.

View File

@@ -0,0 +1,3 @@
-----BEGIN PUBLIC KEY-----
MCowBQYDK2VwAyEADY0al/U0bgB3+9fUGk+3PKWnsck9OyxN5DjHIN6Xep0=
-----END PUBLIC KEY-----

View File

@@ -11,8 +11,8 @@ generate_id() {
PG_VERSION=${PG_VERSION:-14}
SPEC_FILE_ORG=/var/db/postgres/specs/spec.json
SPEC_FILE=/tmp/spec.json
CONFIG_FILE_ORG=/var/db/postgres/configs/config.json
CONFIG_FILE=/tmp/config.json
echo "Waiting pageserver become ready."
while ! nc -z pageserver 6400; do
@@ -20,7 +20,7 @@ while ! nc -z pageserver 6400; do
done
echo "Page server is ready."
cp ${SPEC_FILE_ORG} ${SPEC_FILE}
cp ${CONFIG_FILE_ORG} ${CONFIG_FILE}
if [ -n "${TENANT_ID:-}" ] && [ -n "${TIMELINE_ID:-}" ]; then
tenant_id=${TENANT_ID}
@@ -73,17 +73,17 @@ else
ulid_extension=ulid
fi
echo "Adding pgx_ulid"
shared_libraries=$(jq -r '.cluster.settings[] | select(.name=="shared_preload_libraries").value' ${SPEC_FILE})
sed -i "s/${shared_libraries}/${shared_libraries},${ulid_extension}/" ${SPEC_FILE}
shared_libraries=$(jq -r '.spec.cluster.settings[] | select(.name=="shared_preload_libraries").value' ${CONFIG_FILE})
sed -i "s/${shared_libraries}/${shared_libraries},${ulid_extension}/" ${CONFIG_FILE}
echo "Overwrite tenant id and timeline id in spec file"
sed -i "s/TENANT_ID/${tenant_id}/" ${SPEC_FILE}
sed -i "s/TIMELINE_ID/${timeline_id}/" ${SPEC_FILE}
sed -i "s/TENANT_ID/${tenant_id}/" ${CONFIG_FILE}
sed -i "s/TIMELINE_ID/${timeline_id}/" ${CONFIG_FILE}
cat ${SPEC_FILE}
cat ${CONFIG_FILE}
echo "Start compute node"
/usr/local/bin/compute_ctl --pgdata /var/db/postgres/compute \
-C "postgresql://cloud_admin@localhost:55433/postgres" \
-b /usr/local/bin/postgres \
--compute-id "compute-$RANDOM" \
-S ${SPEC_FILE}
--config "$CONFIG_FILE"

View File

@@ -0,0 +1,160 @@
{
"spec": {
"format_version": 1.0,
"timestamp": "2022-10-12T18:00:00.000Z",
"operation_uuid": "0f657b36-4b0f-4a2d-9c2e-1dcd615e7d8c",
"cluster": {
"cluster_id": "docker_compose",
"name": "docker_compose_test",
"state": "restarted",
"roles": [
{
"name": "cloud_admin",
"encrypted_password": "b093c0d3b281ba6da1eacc608620abd8",
"options": null
}
],
"databases": [
],
"settings": [
{
"name": "fsync",
"value": "off",
"vartype": "bool"
},
{
"name": "wal_level",
"value": "logical",
"vartype": "enum"
},
{
"name": "wal_log_hints",
"value": "on",
"vartype": "bool"
},
{
"name": "log_connections",
"value": "on",
"vartype": "bool"
},
{
"name": "port",
"value": "55433",
"vartype": "integer"
},
{
"name": "shared_buffers",
"value": "1MB",
"vartype": "string"
},
{
"name": "max_connections",
"value": "100",
"vartype": "integer"
},
{
"name": "listen_addresses",
"value": "0.0.0.0",
"vartype": "string"
},
{
"name": "max_wal_senders",
"value": "10",
"vartype": "integer"
},
{
"name": "max_replication_slots",
"value": "10",
"vartype": "integer"
},
{
"name": "wal_sender_timeout",
"value": "5s",
"vartype": "string"
},
{
"name": "wal_keep_size",
"value": "0",
"vartype": "integer"
},
{
"name": "password_encryption",
"value": "md5",
"vartype": "enum"
},
{
"name": "restart_after_crash",
"value": "off",
"vartype": "bool"
},
{
"name": "synchronous_standby_names",
"value": "walproposer",
"vartype": "string"
},
{
"name": "shared_preload_libraries",
"value": "neon,pg_cron,timescaledb,pg_stat_statements",
"vartype": "string"
},
{
"name": "neon.safekeepers",
"value": "safekeeper1:5454,safekeeper2:5454,safekeeper3:5454",
"vartype": "string"
},
{
"name": "neon.timeline_id",
"value": "TIMELINE_ID",
"vartype": "string"
},
{
"name": "neon.tenant_id",
"value": "TENANT_ID",
"vartype": "string"
},
{
"name": "neon.pageserver_connstring",
"value": "host=pageserver port=6400",
"vartype": "string"
},
{
"name": "max_replication_write_lag",
"value": "500MB",
"vartype": "string"
},
{
"name": "max_replication_flush_lag",
"value": "10GB",
"vartype": "string"
},
{
"name": "cron.database",
"value": "postgres",
"vartype": "string"
}
]
},
"delta_operations": [
]
},
"compute_ctl_config": {
"jwks": {
"keys": [
{
"use": "sig",
"key_ops": [
"verify"
],
"alg": "EdDSA",
"kid": "ZGIxMzAzOGY0YWQwODk2ODU1MTk1NzMxMDFkYmUyOWU2NzZkOWNjNjMyMGRkZGJjOWY0MjdjYWVmNzE1MjUyOAo=",
"kty": "OKP",
"crv": "Ed25519",
"x": "MGQ4ZDFhOTdmNTM0NmUwMDc3ZmJkN2Q0MWE0ZmI3M2NhNWE3YjFjOTNkM2IyYzRkZTQzOGM3MjBkZTk3N2E5ZAo="
}
]
}
}
}

View File

@@ -1,141 +0,0 @@
{
"format_version": 1.0,
"timestamp": "2022-10-12T18:00:00.000Z",
"operation_uuid": "0f657b36-4b0f-4a2d-9c2e-1dcd615e7d8c",
"cluster": {
"cluster_id": "docker_compose",
"name": "docker_compose_test",
"state": "restarted",
"roles": [
{
"name": "cloud_admin",
"encrypted_password": "b093c0d3b281ba6da1eacc608620abd8",
"options": null
}
],
"databases": [
],
"settings": [
{
"name": "fsync",
"value": "off",
"vartype": "bool"
},
{
"name": "wal_level",
"value": "logical",
"vartype": "enum"
},
{
"name": "wal_log_hints",
"value": "on",
"vartype": "bool"
},
{
"name": "log_connections",
"value": "on",
"vartype": "bool"
},
{
"name": "port",
"value": "55433",
"vartype": "integer"
},
{
"name": "shared_buffers",
"value": "1MB",
"vartype": "string"
},
{
"name": "max_connections",
"value": "100",
"vartype": "integer"
},
{
"name": "listen_addresses",
"value": "0.0.0.0",
"vartype": "string"
},
{
"name": "max_wal_senders",
"value": "10",
"vartype": "integer"
},
{
"name": "max_replication_slots",
"value": "10",
"vartype": "integer"
},
{
"name": "wal_sender_timeout",
"value": "5s",
"vartype": "string"
},
{
"name": "wal_keep_size",
"value": "0",
"vartype": "integer"
},
{
"name": "password_encryption",
"value": "md5",
"vartype": "enum"
},
{
"name": "restart_after_crash",
"value": "off",
"vartype": "bool"
},
{
"name": "synchronous_standby_names",
"value": "walproposer",
"vartype": "string"
},
{
"name": "shared_preload_libraries",
"value": "neon,pg_cron,timescaledb,pg_stat_statements",
"vartype": "string"
},
{
"name": "neon.safekeepers",
"value": "safekeeper1:5454,safekeeper2:5454,safekeeper3:5454",
"vartype": "string"
},
{
"name": "neon.timeline_id",
"value": "TIMELINE_ID",
"vartype": "string"
},
{
"name": "neon.tenant_id",
"value": "TENANT_ID",
"vartype": "string"
},
{
"name": "neon.pageserver_connstring",
"value": "host=pageserver port=6400",
"vartype": "string"
},
{
"name": "max_replication_write_lag",
"value": "500MB",
"vartype": "string"
},
{
"name": "max_replication_flush_lag",
"value": "10GB",
"vartype": "string"
},
{
"name": "cron.database",
"value": "postgres",
"vartype": "string"
}
]
},
"delta_operations": [
]
}

View File

@@ -159,7 +159,7 @@ services:
#- RUST_BACKTRACE=1
# Mount the test files directly, for faster editing cycle.
volumes:
- ./compute_wrapper/var/db/postgres/specs/:/var/db/postgres/specs/
- ./compute_wrapper/var/db/postgres/configs/:/var/db/postgres/configs/
- ./compute_wrapper/shell/:/shell/
ports:
- 55433:55433 # pg protocol handler

View File

@@ -0,0 +1,8 @@
EXTENSION = pg_jsonschema
DATA = pg_jsonschema--1.0.sql
REGRESS = jsonschema_valid_api jsonschema_edge_cases
REGRESS_OPTS = --load-extension=pg_jsonschema
PG_CONFIG ?= pg_config
PGXS := $(shell $(PG_CONFIG) --pgxs)
include $(PGXS)

View File

@@ -0,0 +1,87 @@
-- Schema with enums, nulls, extra properties disallowed
SELECT jsonschema_is_valid('{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json);
jsonschema_is_valid
---------------------
t
(1 row)
-- Valid enum and null email
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "active", "email": null}'::json
);
jsonschema_validation_errors
------------------------------
{}
(1 row)
-- Invalid enum value
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "disabled", "email": null}'::json
);
jsonschema_validation_errors
----------------------------------------------------------------------
{"\"disabled\" is not one of [\"active\",\"inactive\",\"pending\"]"}
(1 row)
-- Invalid email format (assuming format is validated)
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "active", "email": "not-an-email"}'::json
);
jsonschema_validation_errors
-----------------------------------------
{"\"not-an-email\" is not a \"email\""}
(1 row)
-- Extra property not allowed
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "active", "extra": "should not be here"}'::json
);
jsonschema_validation_errors
--------------------------------------------------------------------
{"Additional properties are not allowed ('extra' was unexpected)"}
(1 row)

View File

@@ -0,0 +1,65 @@
-- Define schema
SELECT jsonschema_is_valid('{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json);
jsonschema_is_valid
---------------------
t
(1 row)
-- Valid instance
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json,
'{"username": "alice", "age": 25}'::json
);
jsonschema_validation_errors
------------------------------
{}
(1 row)
-- Invalid instance: missing required "username"
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json,
'{"age": 25}'::json
);
jsonschema_validation_errors
-----------------------------------------
{"\"username\" is a required property"}
(1 row)
-- Invalid instance: wrong type for "age"
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json,
'{"username": "bob", "age": "twenty"}'::json
);
jsonschema_validation_errors
-------------------------------------------
{"\"twenty\" is not of type \"integer\""}
(1 row)

View File

@@ -0,0 +1,66 @@
-- Schema with enums, nulls, extra properties disallowed
SELECT jsonschema_is_valid('{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json);
-- Valid enum and null email
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "active", "email": null}'::json
);
-- Invalid enum value
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "disabled", "email": null}'::json
);
-- Invalid email format (assuming format is validated)
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "active", "email": "not-an-email"}'::json
);
-- Extra property not allowed
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"status": { "type": "string", "enum": ["active", "inactive", "pending"] },
"email": { "type": ["string", "null"], "format": "email" }
},
"required": ["status"],
"additionalProperties": false
}'::json,
'{"status": "active", "extra": "should not be here"}'::json
);

View File

@@ -0,0 +1,48 @@
-- Define schema
SELECT jsonschema_is_valid('{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json);
-- Valid instance
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json,
'{"username": "alice", "age": 25}'::json
);
-- Invalid instance: missing required "username"
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json,
'{"age": 25}'::json
);
-- Invalid instance: wrong type for "age"
SELECT jsonschema_validation_errors(
'{
"type": "object",
"properties": {
"username": { "type": "string" },
"age": { "type": "integer" }
},
"required": ["username"]
}'::json,
'{"username": "bob", "age": "twenty"}'::json
);

View File

@@ -0,0 +1,9 @@
EXTENSION = pg_session_jwt
REGRESS = basic_functions
REGRESS_OPTS = --load-extension=$(EXTENSION)
export PGOPTIONS = -c pg_session_jwt.jwk={"crv":"Ed25519","kty":"OKP","x":"R_Abz-63zJ00l-IraL5fQhwkhGVZCSooQFV5ntC3C7M"}
PG_CONFIG ?= pg_config
PGXS := $(shell $(PG_CONFIG) --pgxs)
include $(PGXS)

View File

@@ -0,0 +1,35 @@
-- Basic functionality tests for pg_session_jwt
-- Test auth.init() function
SELECT auth.init();
init
------
(1 row)
-- Test an invalid JWT
SELECT auth.jwt_session_init('INVALID-JWT');
ERROR: invalid JWT encoding
-- Test creating a session with an expired JWT
SELECT auth.jwt_session_init('eyJhbGciOiJFZERTQSJ9.eyJleHAiOjE3NDI1NjQ0MzIsImlhdCI6MTc0MjU2NDI1MiwianRpIjo0MjQyNDIsInN1YiI6InVzZXIxMjMifQ.A6FwKuaSduHB9O7Gz37g0uoD_U9qVS0JNtT7YABGVgB7HUD1AMFc9DeyhNntWBqncg8k5brv-hrNTuUh5JYMAw');
ERROR: Token used after it has expired
-- Test creating a session with a valid JWT
SELECT auth.jwt_session_init('eyJhbGciOiJFZERTQSJ9.eyJleHAiOjQ4OTYxNjQyNTIsImlhdCI6MTc0MjU2NDI1MiwianRpIjo0MzQzNDMsInN1YiI6InVzZXIxMjMifQ.2TXVgjb6JSUq6_adlvp-m_SdOxZSyGS30RS9TLB0xu2N83dMSs2NybwE1NMU8Fb0tcAZR_ET7M2rSxbTrphfCg');
jwt_session_init
------------------
(1 row)
-- Test auth.session() function
SELECT auth.session();
session
-------------------------------------------------------------------------
{"exp": 4896164252, "iat": 1742564252, "jti": 434343, "sub": "user123"}
(1 row)
-- Test auth.user_id() function
SELECT auth.user_id() AS user_id;
user_id
---------
user123
(1 row)

View File

@@ -0,0 +1,19 @@
-- Basic functionality tests for pg_session_jwt
-- Test auth.init() function
SELECT auth.init();
-- Test an invalid JWT
SELECT auth.jwt_session_init('INVALID-JWT');
-- Test creating a session with an expired JWT
SELECT auth.jwt_session_init('eyJhbGciOiJFZERTQSJ9.eyJleHAiOjE3NDI1NjQ0MzIsImlhdCI6MTc0MjU2NDI1MiwianRpIjo0MjQyNDIsInN1YiI6InVzZXIxMjMifQ.A6FwKuaSduHB9O7Gz37g0uoD_U9qVS0JNtT7YABGVgB7HUD1AMFc9DeyhNntWBqncg8k5brv-hrNTuUh5JYMAw');
-- Test creating a session with a valid JWT
SELECT auth.jwt_session_init('eyJhbGciOiJFZERTQSJ9.eyJleHAiOjQ4OTYxNjQyNTIsImlhdCI6MTc0MjU2NDI1MiwianRpIjo0MzQzNDMsInN1YiI6InVzZXIxMjMifQ.2TXVgjb6JSUq6_adlvp-m_SdOxZSyGS30RS9TLB0xu2N83dMSs2NybwE1NMU8Fb0tcAZR_ET7M2rSxbTrphfCg');
-- Test auth.session() function
SELECT auth.session();
-- Test auth.user_id() function
SELECT auth.user_id() AS user_id;

View File

@@ -1,5 +1,5 @@
[package]
name = "object_storage"
name = "endpoint_storage"
version = "0.0.1"
edition.workspace = true
license.workspace = true

View File

@@ -2,7 +2,7 @@ use anyhow::anyhow;
use axum::body::{Body, Bytes};
use axum::response::{IntoResponse, Response};
use axum::{Router, http::StatusCode};
use object_storage::{PrefixS3Path, S3Path, Storage, bad_request, internal_error, not_found, ok};
use endpoint_storage::{PrefixS3Path, S3Path, Storage, bad_request, internal_error, not_found, ok};
use remote_storage::TimeoutOrCancel;
use remote_storage::{DownloadError, DownloadOpts, GenericRemoteStorage, RemotePath};
use std::{sync::Arc, time::SystemTime, time::UNIX_EPOCH};
@@ -46,12 +46,12 @@ async fn metrics() -> Result {
async fn get(S3Path { path }: S3Path, state: State) -> Result {
info!(%path, "downloading");
let download_err = |e| {
if let DownloadError::NotFound = e {
info!(%path, %e, "downloading"); // 404 is not an issue of _this_ service
let download_err = |err| {
if let DownloadError::NotFound = err {
info!(%path, %err, "downloading"); // 404 is not an issue of _this_ service
return not_found(&path);
}
internal_error(e, &path, "downloading")
internal_error(err, &path, "downloading")
};
let cancel = state.cancel.clone();
let opts = &DownloadOpts::default();
@@ -249,7 +249,7 @@ mod tests {
};
let proxy = Storage {
auth: object_storage::JwtAuth::new(TEST_PUB_KEY_ED25519).unwrap(),
auth: endpoint_storage::JwtAuth::new(TEST_PUB_KEY_ED25519).unwrap(),
storage,
cancel: cancel.clone(),
max_upload_file_limit: usize::MAX,
@@ -343,14 +343,14 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
TimelineId::from_array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 5, 7]);
const ENDPOINT_ID: &str = "ep-winter-frost-a662z3vg";
fn token() -> String {
let claims = object_storage::Claims {
let claims = endpoint_storage::Claims {
tenant_id: TENANT_ID,
timeline_id: TIMELINE_ID,
endpoint_id: ENDPOINT_ID.into(),
exp: u64::MAX,
};
let key = jsonwebtoken::EncodingKey::from_ed_pem(TEST_PRIV_KEY_ED25519).unwrap();
let header = jsonwebtoken::Header::new(object_storage::VALIDATION_ALGO);
let header = jsonwebtoken::Header::new(endpoint_storage::VALIDATION_ALGO);
jsonwebtoken::encode(&header, &claims, &key).unwrap()
}
@@ -364,7 +364,10 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
vec![TIMELINE_ID.to_string(), TimelineId::generate().to_string()],
vec![ENDPOINT_ID, "ep-ololo"]
)
.skip(1);
// first one is fully valid path, second path is valid for GET as
// read paths may have different endpoint if tenant and timeline matches
// (needed for prewarming RO->RW replica)
.skip(2);
for ((uri, method), (tenant, timeline, endpoint)) in iproduct!(routes(), args) {
info!(%uri, %method, %tenant, %timeline, %endpoint);
@@ -475,6 +478,16 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
requests_chain(chain.into_iter(), |_| token()).await;
}
#[testlog(tokio::test)]
async fn read_other_endpoint_data() {
let uri = format!("/{TENANT_ID}/{TIMELINE_ID}/other_endpoint/key");
let chain = vec![
(uri.clone(), "GET", "", StatusCode::NOT_FOUND, false),
(uri.clone(), "PUT", "", StatusCode::UNAUTHORIZED, false),
];
requests_chain(chain.into_iter(), |_| token()).await;
}
fn delete_prefix_token(uri: &str) -> String {
use serde::Serialize;
let parts = uri.split("/").collect::<Vec<&str>>();
@@ -482,7 +495,7 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
struct PrefixClaims {
tenant_id: TenantId,
timeline_id: Option<TimelineId>,
endpoint_id: Option<object_storage::EndpointId>,
endpoint_id: Option<endpoint_storage::EndpointId>,
exp: u64,
}
let claims = PrefixClaims {
@@ -492,7 +505,7 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
exp: u64::MAX,
};
let key = jsonwebtoken::EncodingKey::from_ed_pem(TEST_PRIV_KEY_ED25519).unwrap();
let header = jsonwebtoken::Header::new(object_storage::VALIDATION_ALGO);
let header = jsonwebtoken::Header::new(endpoint_storage::VALIDATION_ALGO);
jsonwebtoken::encode(&header, &claims, &key).unwrap()
}

View File

@@ -169,10 +169,19 @@ impl FromRequestParts<Arc<Storage>> for S3Path {
.auth
.decode(bearer.token())
.map_err(|e| bad_request(e, "decoding token"))?;
// Read paths may have different endpoint ids. For readonly -> readwrite replica
// prewarming, endpoint must read other endpoint's data.
let endpoint_id = if parts.method == axum::http::Method::GET {
claims.endpoint_id.clone()
} else {
path.endpoint_id.clone()
};
let route = Claims {
tenant_id: path.tenant_id,
timeline_id: path.timeline_id,
endpoint_id: path.endpoint_id.clone(),
endpoint_id,
exp: claims.exp,
};
if route != claims {

View File

@@ -1,4 +1,4 @@
//! `object_storage` is a service which provides API for uploading and downloading
//! `endpoint_storage` is a service which provides API for uploading and downloading
//! files. It is used by compute and control plane for accessing LFC prewarm data.
//! This service is deployed either as a separate component or as part of compute image
//! for large computes.
@@ -33,7 +33,7 @@ async fn main() -> anyhow::Result<()> {
let config: String = std::env::args().skip(1).take(1).collect();
if config.is_empty() {
anyhow::bail!("Usage: object_storage config.json")
anyhow::bail!("Usage: endpoint_storage config.json")
}
info!("Reading config from {config}");
let config = std::fs::read_to_string(config.clone())?;
@@ -41,7 +41,7 @@ async fn main() -> anyhow::Result<()> {
info!("Reading pemfile from {}", config.pemfile.clone());
let pemfile = std::fs::read(config.pemfile.clone())?;
info!("Loading public key from {}", config.pemfile.clone());
let auth = object_storage::JwtAuth::new(&pemfile)?;
let auth = endpoint_storage::JwtAuth::new(&pemfile)?;
let listener = tokio::net::TcpListener::bind(config.listen).await.unwrap();
info!("listening on {}", listener.local_addr().unwrap());
@@ -50,7 +50,7 @@ async fn main() -> anyhow::Result<()> {
let cancel = tokio_util::sync::CancellationToken::new();
app::check_storage_permissions(&storage, cancel.clone()).await?;
let proxy = std::sync::Arc::new(object_storage::Storage {
let proxy = std::sync::Arc::new(endpoint_storage::Storage {
auth,
storage,
cancel: cancel.clone(),

View File

@@ -5,6 +5,14 @@ use crate::privilege::Privilege;
use crate::responses::ComputeCtlConfig;
use crate::spec::{ComputeSpec, ExtVersion, PgIdent};
/// When making requests to the `compute_ctl` external HTTP server, the client
/// must specify a set of claims in `Authorization` header JWTs such that
/// `compute_ctl` can authorize the request.
#[derive(Clone, Debug, Deserialize, Serialize)]
pub struct ComputeClaims {
pub compute_id: String,
}
/// Request of the /configure API
///
/// We now pass only `spec` in the configuration request, but later we can

View File

@@ -14,6 +14,32 @@ pub struct GenericAPIError {
pub error: String,
}
/// All configuration parameters necessary for a compute. When
/// [`ComputeConfig::spec`] is provided, it means that the compute is attached
/// to a tenant. [`ComputeConfig::compute_ctl_config`] will always be provided
/// and contains parameters necessary for operating `compute_ctl` independently
/// of whether a tenant is attached to the compute or not.
///
/// This also happens to be the body of `compute_ctl`'s /configure request.
#[derive(Debug, Deserialize, Serialize)]
pub struct ComputeConfig {
/// The compute spec
pub spec: Option<ComputeSpec>,
/// The compute_ctl configuration
#[allow(dead_code)]
pub compute_ctl_config: ComputeCtlConfig,
}
impl From<ControlPlaneConfigResponse> for ComputeConfig {
fn from(value: ControlPlaneConfigResponse) -> Self {
Self {
spec: value.spec,
compute_ctl_config: value.compute_ctl_config,
}
}
}
#[derive(Debug, Clone, Serialize)]
pub struct ExtensionInstallResponse {
pub extension: PgIdent,
@@ -134,7 +160,7 @@ pub struct CatalogObjects {
pub databases: Vec<Database>,
}
#[derive(Clone, Debug, Deserialize, Serialize)]
#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
pub struct ComputeCtlConfig {
/// Set of JSON web keys that the compute can use to authenticate
/// communication from the control plane.
@@ -153,7 +179,7 @@ impl Default for ComputeCtlConfig {
}
}
#[derive(Clone, Debug, Deserialize, Serialize)]
#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
pub struct TlsConfig {
pub key_path: String,
pub cert_path: String,
@@ -161,7 +187,7 @@ pub struct TlsConfig {
/// Response of the `/computes/{compute_id}/spec` control-plane API.
#[derive(Deserialize, Debug)]
pub struct ControlPlaneSpecResponse {
pub struct ControlPlaneConfigResponse {
pub spec: Option<ComputeSpec>,
pub status: ControlPlaneComputeStatus,
pub compute_ctl_config: ComputeCtlConfig,

View File

@@ -1,8 +1,8 @@
//! `ComputeSpec` represents the contents of the spec.json file.
//!
//! The spec.json file is used to pass information to 'compute_ctl'. It contains
//! all the information needed to start up the right version of PostgreSQL,
//! and connect it to the storage nodes.
//! The ComputeSpec contains all the information needed to start up
//! the right version of PostgreSQL, and connect it to the storage nodes.
//! It can be passed as part of the `config.json`, or the control plane can
//! provide it by calling the compute_ctl's `/compute_ctl` endpoint, or
//! compute_ctl can fetch it by calling the control plane's API.
use std::collections::HashMap;
use indexmap::IndexMap;
@@ -104,6 +104,12 @@ pub struct ComputeSpec {
pub timeline_id: Option<TimelineId>,
pub pageserver_connstring: Option<String>,
// More neon ids that we expose to the compute_ctl
// and to postgres as neon extension GUCs.
pub project_id: Option<String>,
pub branch_id: Option<String>,
pub endpoint_id: Option<String>,
/// Safekeeper membership config generation. It is put in
/// neon.safekeepers GUC and serves two purposes:
/// 1) Non zero value forces walproposer to use membership configurations.
@@ -159,13 +165,7 @@ pub struct ComputeSpec {
#[serde(default)] // Default false
pub drop_subscriptions_before_start: bool,
/// Log level for audit logging:
///
/// Disabled - no audit logging. This is the default.
/// log - log masked statements to the postgres log using pgaudit extension
/// hipaa - log unmasked statements to the file using pgaudit and pgauditlogtofile extension
///
/// Extensions should be present in shared_preload_libraries
/// Log level for compute audit logging
#[serde(default)]
pub audit_log_level: ComputeAudit,
@@ -242,13 +242,22 @@ impl RemoteExtSpec {
match self.extension_data.get(real_ext_name) {
Some(_ext_data) => {
// We have decided to use the Go naming convention due to Kubernetes.
let arch = match std::env::consts::ARCH {
"x86_64" => "amd64",
"aarch64" => "arm64",
arch => arch,
};
// Construct the path to the extension archive
// BUILD_TAG/PG_MAJOR_VERSION/extensions/EXTENSION_NAME.tar.zst
//
// Keep it in sync with path generation in
// https://github.com/neondatabase/build-custom-extensions/tree/main
let archive_path_str =
format!("{build_tag}/{pg_major_version}/extensions/{real_ext_name}.tar.zst");
let archive_path_str = format!(
"{build_tag}/{arch}/{pg_major_version}/extensions/{real_ext_name}.tar.zst"
);
Ok((
real_ext_name.to_string(),
RemotePath::from_string(&archive_path_str)?,
@@ -289,14 +298,25 @@ impl ComputeMode {
}
/// Log level for audit logging
/// Disabled, log, hipaa
/// Default is Disabled
#[derive(Clone, Debug, Default, Eq, PartialEq, Deserialize, Serialize)]
pub enum ComputeAudit {
#[default]
Disabled,
// Deprecated, use Base instead
Log,
// (pgaudit.log = 'ddl', pgaudit.log_parameter='off')
// logged to the standard postgresql log stream
Base,
// Deprecated, use Full or Extended instead
Hipaa,
// (pgaudit.log = 'all, -misc', pgaudit.log_parameter='off')
// logged to separate files collected by rsyslog
// into dedicated log storage with strict access
Extended,
// (pgaudit.log='all', pgaudit.log_parameter='on'),
// logged to separate files collected by rsyslog
// into dedicated log storage with strict access.
Full,
}
#[derive(Clone, Debug, Default, Deserialize, Serialize, PartialEq, Eq)]

View File

@@ -14,6 +14,7 @@ futures.workspace = true
hyper0.workspace = true
itertools.workspace = true
jemalloc_pprof.workspace = true
jsonwebtoken.workspace = true
once_cell.workspace = true
pprof.workspace = true
regex.workspace = true
@@ -30,6 +31,7 @@ tokio.workspace = true
tracing.workspace = true
url.workspace = true
uuid.workspace = true
x509-cert.workspace = true
# to use tokio channels as streams, this is faster to compile than async_stream
# why is it only here? no other crate should use it, streams are rarely needed.

View File

@@ -8,6 +8,7 @@ use bytes::{Bytes, BytesMut};
use hyper::header::{AUTHORIZATION, CONTENT_DISPOSITION, CONTENT_TYPE, HeaderName};
use hyper::http::HeaderValue;
use hyper::{Body, Method, Request, Response};
use jsonwebtoken::TokenData;
use metrics::{Encoder, IntCounter, TextEncoder, register_int_counter};
use once_cell::sync::Lazy;
use pprof::ProfilerGuardBuilder;
@@ -618,7 +619,7 @@ pub fn auth_middleware<B: hyper::body::HttpBody + Send + Sync + 'static>(
})?;
let token = parse_token(header_value)?;
let data = auth.decode(token).map_err(|err| {
let data: TokenData<Claims> = auth.decode(token).map_err(|err| {
warn!("Authentication error: {err}");
// Rely on From<AuthError> for ApiError impl
err

View File

@@ -4,6 +4,8 @@ use futures::StreamExt;
use futures::stream::FuturesUnordered;
use hyper0::Body;
use hyper0::server::conn::Http;
use metrics::{IntCounterVec, register_int_counter_vec};
use once_cell::sync::Lazy;
use routerify::{RequestService, RequestServiceBuilder};
use tokio::io::{AsyncRead, AsyncWrite};
use tokio_rustls::TlsAcceptor;
@@ -26,6 +28,24 @@ pub struct Server {
tls_acceptor: Option<TlsAcceptor>,
}
static CONNECTION_STARTED_COUNT: Lazy<IntCounterVec> = Lazy::new(|| {
register_int_counter_vec!(
"http_server_connection_started_total",
"Number of established http/https connections",
&["scheme"]
)
.expect("failed to define a metric")
});
static CONNECTION_ERROR_COUNT: Lazy<IntCounterVec> = Lazy::new(|| {
register_int_counter_vec!(
"http_server_connection_errors_total",
"Number of occured connection errors by type",
&["type"]
)
.expect("failed to define a metric")
});
impl Server {
pub fn new(
request_service: Arc<RequestServiceBuilder<Body, ApiError>>,
@@ -60,6 +80,15 @@ impl Server {
false
}
let tcp_error_cnt = CONNECTION_ERROR_COUNT.with_label_values(&["tcp"]);
let tls_error_cnt = CONNECTION_ERROR_COUNT.with_label_values(&["tls"]);
let http_error_cnt = CONNECTION_ERROR_COUNT.with_label_values(&["http"]);
let https_error_cnt = CONNECTION_ERROR_COUNT.with_label_values(&["https"]);
let panic_error_cnt = CONNECTION_ERROR_COUNT.with_label_values(&["panic"]);
let http_connection_cnt = CONNECTION_STARTED_COUNT.with_label_values(&["http"]);
let https_connection_cnt = CONNECTION_STARTED_COUNT.with_label_values(&["https"]);
let mut connections = FuturesUnordered::new();
loop {
tokio::select! {
@@ -67,6 +96,7 @@ impl Server {
let (tcp_stream, remote_addr) = match stream {
Ok(stream) => stream,
Err(err) => {
tcp_error_cnt.inc();
if !suppress_io_error(&err) {
info!("Failed to accept TCP connection: {err:#}");
}
@@ -78,11 +108,18 @@ impl Server {
let tls_acceptor = self.tls_acceptor.clone();
let cancel = cancel.clone();
let tls_error_cnt = tls_error_cnt.clone();
let http_error_cnt = http_error_cnt.clone();
let https_error_cnt = https_error_cnt.clone();
let http_connection_cnt = http_connection_cnt.clone();
let https_connection_cnt = https_connection_cnt.clone();
connections.push(tokio::spawn(
async move {
match tls_acceptor {
Some(tls_acceptor) => {
// Handle HTTPS connection.
https_connection_cnt.inc();
let tls_stream = tokio::select! {
tls_stream = tls_acceptor.accept(tcp_stream) => tls_stream,
_ = cancel.cancelled() => return,
@@ -90,6 +127,7 @@ impl Server {
let tls_stream = match tls_stream {
Ok(tls_stream) => tls_stream,
Err(err) => {
tls_error_cnt.inc();
if !suppress_io_error(&err) {
info!(%remote_addr, "Failed to accept TLS connection: {err:#}");
}
@@ -97,6 +135,7 @@ impl Server {
}
};
if let Err(err) = Self::serve_connection(tls_stream, service, cancel).await {
https_error_cnt.inc();
if !suppress_hyper_error(&err) {
info!(%remote_addr, "Failed to serve HTTPS connection: {err:#}");
}
@@ -104,7 +143,9 @@ impl Server {
}
None => {
// Handle HTTP connection.
http_connection_cnt.inc();
if let Err(err) = Self::serve_connection(tcp_stream, service, cancel).await {
http_error_cnt.inc();
if !suppress_hyper_error(&err) {
info!(%remote_addr, "Failed to serve HTTP connection: {err:#}");
}
@@ -115,6 +156,7 @@ impl Server {
}
Some(conn) = connections.next() => {
if let Err(err) = conn {
panic_error_cnt.inc();
error!("Connection panicked: {err:#}");
}
}
@@ -122,6 +164,7 @@ impl Server {
// Wait for graceful shutdown of all connections.
while let Some(conn) = connections.next().await {
if let Err(err) = conn {
panic_error_cnt.inc();
error!("Connection panicked: {err:#}");
}
}

View File

@@ -3,11 +3,14 @@ use std::{sync::Arc, time::Duration};
use anyhow::Context;
use arc_swap::ArcSwap;
use camino::Utf8Path;
use metrics::{IntCounterVec, UIntGaugeVec, register_int_counter_vec, register_uint_gauge_vec};
use once_cell::sync::Lazy;
use rustls::{
pki_types::{CertificateDer, PrivateKeyDer},
pki_types::{CertificateDer, PrivateKeyDer, UnixTime},
server::{ClientHello, ResolvesServerCert},
sign::CertifiedKey,
};
use x509_cert::der::Reader;
pub async fn load_cert_chain(filename: &Utf8Path) -> anyhow::Result<Vec<CertificateDer<'static>>> {
let cert_data = tokio::fs::read(filename)
@@ -53,6 +56,76 @@ pub async fn load_certified_key(
Ok(certified_key)
}
/// rustls's CertifiedKey with extra parsed fields used for metrics.
struct ParsedCertifiedKey {
certified_key: CertifiedKey,
expiration_time: UnixTime,
}
/// Parse expiration time from an X509 certificate.
fn parse_expiration_time(cert: &CertificateDer<'_>) -> anyhow::Result<UnixTime> {
let parsed_cert = x509_cert::der::SliceReader::new(cert)
.context("Failed to parse cerficiate")?
.decode::<x509_cert::Certificate>()
.context("Failed to parse cerficiate")?;
Ok(UnixTime::since_unix_epoch(
parsed_cert
.tbs_certificate
.validity
.not_after
.to_unix_duration(),
))
}
async fn load_and_parse_certified_key(
key_filename: &Utf8Path,
cert_filename: &Utf8Path,
) -> anyhow::Result<ParsedCertifiedKey> {
let certified_key = load_certified_key(key_filename, cert_filename).await?;
let expiration_time = parse_expiration_time(certified_key.end_entity_cert()?)?;
Ok(ParsedCertifiedKey {
certified_key,
expiration_time,
})
}
static CERT_EXPIRATION_TIME: Lazy<UIntGaugeVec> = Lazy::new(|| {
register_uint_gauge_vec!(
"tls_certs_expiration_time_seconds",
"Expiration time of the loaded certificate since unix epoch in seconds",
&["resolver_name"]
)
.expect("failed to define a metric")
});
static CERT_RELOAD_STARTED_COUNTER: Lazy<IntCounterVec> = Lazy::new(|| {
register_int_counter_vec!(
"tls_certs_reload_started_total",
"Number of certificate reload loop iterations started",
&["resolver_name"]
)
.expect("failed to define a metric")
});
static CERT_RELOAD_UPDATED_COUNTER: Lazy<IntCounterVec> = Lazy::new(|| {
register_int_counter_vec!(
"tls_certs_reload_updated_total",
"Number of times the certificate was updated to the new one",
&["resolver_name"]
)
.expect("failed to define a metric")
});
static CERT_RELOAD_FAILED_COUNTER: Lazy<IntCounterVec> = Lazy::new(|| {
register_int_counter_vec!(
"tls_certs_reload_failed_total",
"Number of times the certificate reload failed",
&["resolver_name"]
)
.expect("failed to define a metric")
});
/// Implementation of [`rustls::server::ResolvesServerCert`] which reloads certificates from
/// the disk periodically.
#[derive(Debug)]
@@ -63,16 +136,28 @@ pub struct ReloadingCertificateResolver {
impl ReloadingCertificateResolver {
/// Creates a new Resolver by loading certificate and private key from FS and
/// creating tokio::task to reload them with provided reload_period.
/// resolver_name is used as metric's label.
pub async fn new(
resolver_name: &str,
key_filename: &Utf8Path,
cert_filename: &Utf8Path,
reload_period: Duration,
) -> anyhow::Result<Arc<Self>> {
// Create metrics for current resolver.
let cert_expiration_time = CERT_EXPIRATION_TIME.with_label_values(&[resolver_name]);
let cert_reload_started_counter =
CERT_RELOAD_STARTED_COUNTER.with_label_values(&[resolver_name]);
let cert_reload_updated_counter =
CERT_RELOAD_UPDATED_COUNTER.with_label_values(&[resolver_name]);
let cert_reload_failed_counter =
CERT_RELOAD_FAILED_COUNTER.with_label_values(&[resolver_name]);
let parsed_key = load_and_parse_certified_key(key_filename, cert_filename).await?;
let this = Arc::new(Self {
certified_key: ArcSwap::from_pointee(
load_certified_key(key_filename, cert_filename).await?,
),
certified_key: ArcSwap::from_pointee(parsed_key.certified_key),
});
cert_expiration_time.set(parsed_key.expiration_time.as_secs());
tokio::spawn({
let weak_this = Arc::downgrade(&this);
@@ -88,17 +173,22 @@ impl ReloadingCertificateResolver {
Some(this) => this,
None => break, // Resolver has been destroyed, exit.
};
match load_certified_key(&key_filename, &cert_filename).await {
Ok(new_certified_key) => {
if new_certified_key.cert == this.certified_key.load().cert {
cert_reload_started_counter.inc();
match load_and_parse_certified_key(&key_filename, &cert_filename).await {
Ok(parsed_key) => {
if parsed_key.certified_key.cert == this.certified_key.load().cert {
tracing::debug!("Certificate has not changed since last reloading");
} else {
tracing::info!("Certificate has been reloaded");
this.certified_key.store(Arc::new(new_certified_key));
this.certified_key.store(Arc::new(parsed_key.certified_key));
cert_expiration_time.set(parsed_key.expiration_time.as_secs());
cert_reload_updated_counter.inc();
}
last_reload_failed = false;
}
Err(err) => {
cert_reload_failed_counter.inc();
// Note: Reloading certs may fail if it conflicts with the script updating
// the files at the same time. Warn only if the error is persistent.
if last_reload_failed {

View File

@@ -35,6 +35,7 @@ nix = {workspace = true, optional = true}
reqwest.workspace = true
rand.workspace = true
tracing-utils.workspace = true
once_cell.workspace = true
[dev-dependencies]
bincode.workspace = true

View File

@@ -180,6 +180,8 @@ pub struct ConfigToml {
#[serde(skip_serializing_if = "Option::is_none")]
pub generate_unarchival_heatmap: Option<bool>,
pub tracing: Option<Tracing>,
pub enable_tls_page_service_api: bool,
pub dev_mode: bool,
}
#[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
@@ -206,6 +208,10 @@ pub struct PageServicePipeliningConfigPipelined {
/// Causes runtime errors if larger than max get_vectored batch size.
pub max_batch_size: NonZeroUsize,
pub execution: PageServiceProtocolPipelinedExecutionStrategy,
// The default below is such that new versions of the software can start
// with the old configuration.
#[serde(default)]
pub batching: PageServiceProtocolPipelinedBatchingStrategy,
}
#[derive(Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
@@ -215,6 +221,19 @@ pub enum PageServiceProtocolPipelinedExecutionStrategy {
Tasks,
}
#[derive(Default, Debug, Clone, Copy, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
#[serde(rename_all = "kebab-case")]
pub enum PageServiceProtocolPipelinedBatchingStrategy {
/// All get page requests in a batch will be at the same LSN
#[default]
UniformLsn,
/// Get page requests in a batch may be at different LSN
///
/// One key cannot be present more than once at different LSNs in
/// the same batch.
ScatteredLsn,
}
#[derive(Debug, Clone, PartialEq, Eq, serde::Serialize, serde::Deserialize)]
#[serde(tag = "mode", rename_all = "kebab-case")]
pub enum GetVectoredConcurrentIo {
@@ -361,6 +380,8 @@ pub struct TenantConfigToml {
/// size exceeds `compaction_upper_limit * checkpoint_distance`.
pub compaction_upper_limit: usize,
pub compaction_algorithm: crate::models::CompactionAlgorithmSettings,
/// If true, enable shard ancestor compaction (enabled by default).
pub compaction_shard_ancestor: bool,
/// If true, compact down L0 across all tenant timelines before doing regular compaction. L0
/// compaction must be responsive to avoid read amp during heavy ingestion. Defaults to true.
pub compaction_l0_first: bool,
@@ -451,6 +472,8 @@ pub struct TenantConfigToml {
// gc-compaction related configs
/// Enable automatic gc-compaction trigger on this tenant.
pub gc_compaction_enabled: bool,
/// Enable verification of gc-compaction results.
pub gc_compaction_verification: bool,
/// The initial threshold for gc-compaction in KB. Once the total size of layers below the gc-horizon is above this threshold,
/// gc-compaction will be triggered.
pub gc_compaction_initial_threshold_kb: u64,
@@ -612,9 +635,12 @@ impl Default for ConfigToml {
page_service_pipelining: if !cfg!(test) {
PageServicePipeliningConfig::Serial
} else {
// Do not turn this into the default until scattered reads have been
// validated and rolled-out fully.
PageServicePipeliningConfig::Pipelined(PageServicePipeliningConfigPipelined {
max_batch_size: NonZeroUsize::new(32).unwrap(),
execution: PageServiceProtocolPipelinedExecutionStrategy::ConcurrentFutures,
batching: PageServiceProtocolPipelinedBatchingStrategy::ScatteredLsn,
})
},
get_vectored_concurrent_io: if !cfg!(test) {
@@ -631,6 +657,8 @@ impl Default for ConfigToml {
load_previous_heatmap: None,
generate_unarchival_heatmap: None,
tracing: None,
enable_tls_page_service_api: false,
dev_mode: false,
}
}
}
@@ -653,12 +681,13 @@ pub mod tenant_conf_defaults {
pub const DEFAULT_COMPACTION_PERIOD: &str = "20 s";
pub const DEFAULT_COMPACTION_THRESHOLD: usize = 10;
pub const DEFAULT_COMPACTION_SHARD_ANCESTOR: bool = true;
// This value needs to be tuned to avoid OOM. We have 3/4*CPUs threads for L0 compaction, that's
// 3/4*16=9 on most of our pageservers. Compacting 20 layers requires about 1 GB memory (could
// be reduced later by optimizing L0 hole calculation to avoid loading all keys into memory). So
// with this config, we can get a maximum peak compaction usage of 9 GB.
pub const DEFAULT_COMPACTION_UPPER_LIMIT: usize = 20;
// 3/4*8=6 on most of our pageservers. Compacting 10 layers requires a maximum of
// DEFAULT_CHECKPOINT_DISTANCE*10 memory, that's 2560MB. So with this config, we can get a maximum peak
// compaction usage of 15360MB.
pub const DEFAULT_COMPACTION_UPPER_LIMIT: usize = 10;
// Enable L0 compaction pass and semaphore by default. L0 compaction must be responsive to avoid
// read amp.
pub const DEFAULT_COMPACTION_L0_FIRST: bool = true;
@@ -675,8 +704,11 @@ pub mod tenant_conf_defaults {
// Relevant: https://github.com/neondatabase/neon/issues/3394
pub const DEFAULT_GC_PERIOD: &str = "1 hr";
pub const DEFAULT_IMAGE_CREATION_THRESHOLD: usize = 3;
// If there are more than threshold * compaction_threshold (that is 3 * 10 in the default config) L0 layers, image
// layer creation will end immediately. Set to 0 to disable.
// Currently, any value other than 0 will trigger image layer creation preemption immediately with L0 backpressure
// without looking at the exact number of L0 layers.
// It was expected to have the following behavior:
// > If there are more than threshold * compaction_threshold (that is 3 * 10 in the default config) L0 layers, image
// > layer creation will end immediately. Set to 0 to disable.
pub const DEFAULT_IMAGE_CREATION_PREEMPT_THRESHOLD: usize = 3;
pub const DEFAULT_PITR_INTERVAL: &str = "7 days";
pub const DEFAULT_WALRECEIVER_CONNECT_TIMEOUT: &str = "10 seconds";
@@ -690,6 +722,7 @@ pub mod tenant_conf_defaults {
// image layers should be created.
pub const DEFAULT_IMAGE_LAYER_CREATION_CHECK_THRESHOLD: u8 = 2;
pub const DEFAULT_GC_COMPACTION_ENABLED: bool = false;
pub const DEFAULT_GC_COMPACTION_VERIFICATION: bool = true;
pub const DEFAULT_GC_COMPACTION_INITIAL_THRESHOLD_KB: u64 = 5 * 1024 * 1024; // 5GB
pub const DEFAULT_GC_COMPACTION_RATIO_PERCENT: u64 = 100;
}
@@ -709,6 +742,7 @@ impl Default for TenantConfigToml {
compaction_algorithm: crate::models::CompactionAlgorithmSettings {
kind: DEFAULT_COMPACTION_ALGORITHM,
},
compaction_shard_ancestor: DEFAULT_COMPACTION_SHARD_ANCESTOR,
compaction_l0_first: DEFAULT_COMPACTION_L0_FIRST,
compaction_l0_semaphore: DEFAULT_COMPACTION_L0_SEMAPHORE,
l0_flush_delay_threshold: None,
@@ -744,6 +778,7 @@ impl Default for TenantConfigToml {
wal_receiver_protocol_override: None,
rel_size_v2_enabled: false,
gc_compaction_enabled: DEFAULT_GC_COMPACTION_ENABLED,
gc_compaction_verification: DEFAULT_GC_COMPACTION_VERIFICATION,
gc_compaction_initial_threshold_kb: DEFAULT_GC_COMPACTION_INITIAL_THRESHOLD_KB,
gc_compaction_ratio_percent: DEFAULT_GC_COMPACTION_RATIO_PERCENT,
sampling_ratio: None,

View File

@@ -7,7 +7,8 @@ use std::time::{Duration, Instant};
/// API (`/control/v1` prefix). Implemented by the server
/// in [`storage_controller::http`]
use serde::{Deserialize, Serialize};
use utils::id::{NodeId, TenantId};
use utils::id::{NodeId, TenantId, TimelineId};
use utils::lsn::Lsn;
use crate::models::{PageserverUtilization, ShardParameters, TenantConfig};
use crate::shard::{ShardStripeSize, TenantShardId};
@@ -499,6 +500,15 @@ pub struct SafekeeperSchedulingPolicyRequest {
pub scheduling_policy: SkSchedulingPolicy,
}
/// Import request for safekeeper timelines.
#[derive(Serialize, Deserialize, Clone)]
pub struct TimelineImportRequest {
pub tenant_id: TenantId,
pub timeline_id: TimelineId,
pub start_lsn: Lsn,
pub sk_set: Vec<NodeId>,
}
#[cfg(test)]
mod test {
use serde_json;

View File

@@ -927,7 +927,7 @@ impl Key {
/// Guaranteed to return `Ok()` if [`Self::is_rel_block_key`] returns `true` for `key`.
#[inline(always)]
pub fn to_rel_block(self) -> anyhow::Result<(RelTag, BlockNumber)> {
pub fn to_rel_block(self) -> Result<(RelTag, BlockNumber), ToRelBlockError> {
Ok(match self.field1 {
0x00 => (
RelTag {
@@ -938,7 +938,7 @@ impl Key {
},
self.field6,
),
_ => anyhow::bail!("unexpected value kind 0x{:02x}", self.field1),
_ => return Err(ToRelBlockError(self.field1)),
})
}
}
@@ -951,6 +951,17 @@ impl std::str::FromStr for Key {
}
}
#[derive(Debug)]
pub struct ToRelBlockError(u8);
impl fmt::Display for ToRelBlockError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "unexpected value kind 0x{:02x}", self.0)
}
}
impl std::error::Error for ToRelBlockError {}
#[cfg(test)]
mod tests {
use std::str::FromStr;

View File

@@ -320,6 +320,35 @@ pub struct TimelineCreateRequest {
pub mode: TimelineCreateRequestMode,
}
impl TimelineCreateRequest {
pub fn mode_tag(&self) -> &'static str {
match &self.mode {
TimelineCreateRequestMode::Branch { .. } => "branch",
TimelineCreateRequestMode::ImportPgdata { .. } => "import",
TimelineCreateRequestMode::Bootstrap { .. } => "bootstrap",
}
}
pub fn is_import(&self) -> bool {
matches!(self.mode, TimelineCreateRequestMode::ImportPgdata { .. })
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq)]
pub enum ShardImportStatus {
InProgress,
Done,
Error(String),
}
impl ShardImportStatus {
pub fn is_terminal(&self) -> bool {
match self {
ShardImportStatus::InProgress => false,
ShardImportStatus::Done | ShardImportStatus::Error(_) => true,
}
}
}
/// Storage controller specific extensions to [`TimelineInfo`].
#[derive(Serialize, Deserialize, Clone)]
pub struct TimelineCreateResponseStorcon {
@@ -526,6 +555,8 @@ pub struct TenantConfigPatch {
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub compaction_algorithm: FieldPatch<CompactionAlgorithmSettings>,
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub compaction_shard_ancestor: FieldPatch<bool>,
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub compaction_l0_first: FieldPatch<bool>,
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub compaction_l0_semaphore: FieldPatch<bool>,
@@ -576,6 +607,8 @@ pub struct TenantConfigPatch {
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub gc_compaction_enabled: FieldPatch<bool>,
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub gc_compaction_verification: FieldPatch<bool>,
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub gc_compaction_initial_threshold_kb: FieldPatch<u64>,
#[serde(skip_serializing_if = "FieldPatch::is_noop")]
pub gc_compaction_ratio_percent: FieldPatch<u64>,
@@ -613,6 +646,9 @@ pub struct TenantConfig {
#[serde(skip_serializing_if = "Option::is_none")]
pub compaction_algorithm: Option<CompactionAlgorithmSettings>,
#[serde(skip_serializing_if = "Option::is_none")]
pub compaction_shard_ancestor: Option<bool>,
#[serde(skip_serializing_if = "Option::is_none")]
pub compaction_l0_first: Option<bool>,
@@ -696,6 +732,9 @@ pub struct TenantConfig {
#[serde(skip_serializing_if = "Option::is_none")]
pub gc_compaction_enabled: Option<bool>,
#[serde(skip_serializing_if = "Option::is_none")]
pub gc_compaction_verification: Option<bool>,
#[serde(skip_serializing_if = "Option::is_none")]
pub gc_compaction_initial_threshold_kb: Option<u64>,
@@ -719,6 +758,7 @@ impl TenantConfig {
mut compaction_threshold,
mut compaction_upper_limit,
mut compaction_algorithm,
mut compaction_shard_ancestor,
mut compaction_l0_first,
mut compaction_l0_semaphore,
mut l0_flush_delay_threshold,
@@ -744,6 +784,7 @@ impl TenantConfig {
mut wal_receiver_protocol_override,
mut rel_size_v2_enabled,
mut gc_compaction_enabled,
mut gc_compaction_verification,
mut gc_compaction_initial_threshold_kb,
mut gc_compaction_ratio_percent,
mut sampling_ratio,
@@ -766,6 +807,9 @@ impl TenantConfig {
.compaction_upper_limit
.apply(&mut compaction_upper_limit);
patch.compaction_algorithm.apply(&mut compaction_algorithm);
patch
.compaction_shard_ancestor
.apply(&mut compaction_shard_ancestor);
patch.compaction_l0_first.apply(&mut compaction_l0_first);
patch
.compaction_l0_semaphore
@@ -835,6 +879,9 @@ impl TenantConfig {
patch
.gc_compaction_enabled
.apply(&mut gc_compaction_enabled);
patch
.gc_compaction_verification
.apply(&mut gc_compaction_verification);
patch
.gc_compaction_initial_threshold_kb
.apply(&mut gc_compaction_initial_threshold_kb);
@@ -851,6 +898,7 @@ impl TenantConfig {
compaction_threshold,
compaction_upper_limit,
compaction_algorithm,
compaction_shard_ancestor,
compaction_l0_first,
compaction_l0_semaphore,
l0_flush_delay_threshold,
@@ -876,6 +924,7 @@ impl TenantConfig {
wal_receiver_protocol_override,
rel_size_v2_enabled,
gc_compaction_enabled,
gc_compaction_verification,
gc_compaction_initial_threshold_kb,
gc_compaction_ratio_percent,
sampling_ratio,
@@ -910,6 +959,9 @@ impl TenantConfig {
.as_ref()
.unwrap_or(&global_conf.compaction_algorithm)
.clone(),
compaction_shard_ancestor: self
.compaction_shard_ancestor
.unwrap_or(global_conf.compaction_shard_ancestor),
compaction_l0_first: self
.compaction_l0_first
.unwrap_or(global_conf.compaction_l0_first),
@@ -974,6 +1026,9 @@ impl TenantConfig {
gc_compaction_enabled: self
.gc_compaction_enabled
.unwrap_or(global_conf.gc_compaction_enabled),
gc_compaction_verification: self
.gc_compaction_verification
.unwrap_or(global_conf.gc_compaction_verification),
gc_compaction_initial_threshold_kb: self
.gc_compaction_initial_threshold_kb
.unwrap_or(global_conf.gc_compaction_initial_threshold_kb),
@@ -1791,8 +1846,34 @@ pub mod virtual_file {
}
impl IoMode {
pub const fn preferred() -> Self {
Self::Buffered
pub fn preferred() -> Self {
// The default behavior when running Rust unit tests without any further
// flags is to use the newest behavior if available on the platform (Direct).
// The CI uses the following environment variable to unit tests for all
// different modes.
// NB: the Python regression & perf tests have their own defaults management
// that writes pageserver.toml; they do not use this variable.
if cfg!(test) {
use once_cell::sync::Lazy;
static CACHED: Lazy<IoMode> = Lazy::new(|| {
utils::env::var_serde_json_string(
"NEON_PAGESERVER_UNIT_TEST_VIRTUAL_FILE_IO_MODE",
)
.unwrap_or({
#[cfg(target_os = "linux")]
{
IoMode::Direct
}
#[cfg(not(target_os = "linux"))]
{
IoMode::Buffered
}
})
});
*CACHED
} else {
IoMode::Buffered
}
}
}

View File

@@ -4,10 +4,10 @@
//! See docs/rfcs/025-generation-numbers.md
use serde::{Deserialize, Serialize};
use utils::id::NodeId;
use utils::id::{NodeId, TimelineId};
use crate::controller_api::NodeRegisterRequest;
use crate::models::LocationConfigMode;
use crate::models::{LocationConfigMode, ShardImportStatus};
use crate::shard::TenantShardId;
/// Upcall message sent by the pageserver to the configured `control_plane_api` on
@@ -62,3 +62,10 @@ pub struct ValidateResponseTenant {
pub id: TenantShardId,
pub valid: bool,
}
#[derive(Serialize, Deserialize)]
pub struct PutTimelineImportStatusRequest {
pub tenant_shard_id: TenantShardId,
pub timeline_id: TimelineId,
pub status: ShardImportStatus,
}

View File

@@ -14,8 +14,9 @@ use anyhow::{Context, Result};
use azure_core::request_options::{IfMatchCondition, MaxResults, Metadata, Range};
use azure_core::{Continuable, HttpClient, RetryOptions, TransportOptions};
use azure_storage::StorageCredentials;
use azure_storage_blobs::blob::CopyStatus;
use azure_storage_blobs::blob::operations::GetBlobBuilder;
use azure_storage_blobs::blob::{Blob, CopyStatus};
use azure_storage_blobs::container::operations::ListBlobsBuilder;
use azure_storage_blobs::prelude::{ClientBuilder, ContainerClient};
use bytes::Bytes;
use futures::FutureExt;
@@ -253,53 +254,15 @@ impl AzureBlobStorage {
download
}
async fn permit(
&self,
kind: RequestKind,
cancel: &CancellationToken,
) -> Result<tokio::sync::SemaphorePermit<'_>, Cancelled> {
let acquire = self.concurrency_limiter.acquire(kind);
tokio::select! {
permit = acquire => Ok(permit.expect("never closed")),
_ = cancel.cancelled() => Err(Cancelled),
}
}
pub fn container_name(&self) -> &str {
&self.container_name
}
}
fn to_azure_metadata(metadata: StorageMetadata) -> Metadata {
let mut res = Metadata::new();
for (k, v) in metadata.0.into_iter() {
res.insert(k, v);
}
res
}
fn to_download_error(error: azure_core::Error) -> DownloadError {
if let Some(http_err) = error.as_http_error() {
match http_err.status() {
StatusCode::NotFound => DownloadError::NotFound,
StatusCode::NotModified => DownloadError::Unmodified,
StatusCode::BadRequest => DownloadError::BadInput(anyhow::Error::new(error)),
_ => DownloadError::Other(anyhow::Error::new(error)),
}
} else {
DownloadError::Other(error.into())
}
}
impl RemoteStorage for AzureBlobStorage {
fn list_streaming(
fn list_streaming_for_fn<T: Default + ListingCollector>(
&self,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> impl Stream<Item = Result<Listing, DownloadError>> {
request_kind: RequestKind,
customize_builder: impl Fn(ListBlobsBuilder) -> ListBlobsBuilder,
) -> impl Stream<Item = Result<T, DownloadError>> {
// get the passed prefix or if it is not set use prefix_in_bucket value
let list_prefix = prefix.map(|p| self.relative_path_to_name(p)).or_else(|| {
self.prefix_in_container.clone().map(|mut s| {
@@ -311,7 +274,7 @@ impl RemoteStorage for AzureBlobStorage {
});
async_stream::stream! {
let _permit = self.permit(RequestKind::List, cancel).await?;
let _permit = self.permit(request_kind, cancel).await?;
let mut builder = self.client.list_blobs();
@@ -327,6 +290,8 @@ impl RemoteStorage for AzureBlobStorage {
builder = builder.max_results(MaxResults::new(limit));
}
builder = customize_builder(builder);
let mut next_marker = None;
let mut timeout_try_cnt = 1;
@@ -382,26 +347,20 @@ impl RemoteStorage for AzureBlobStorage {
break;
};
let mut res = Listing::default();
let mut res = T::default();
next_marker = entry.continuation();
let prefix_iter = entry
.blobs
.prefixes()
.map(|prefix| self.name_to_relative_path(&prefix.name));
res.prefixes.extend(prefix_iter);
res.add_prefixes(self, prefix_iter);
let blob_iter = entry
.blobs
.blobs()
.map(|k| ListingObject{
key: self.name_to_relative_path(&k.name),
last_modified: k.properties.last_modified.into(),
size: k.properties.content_length,
}
);
.blobs();
for key in blob_iter {
res.keys.push(key);
res.add_blob(self, key);
if let Some(mut mk) = max_keys {
assert!(mk > 0);
@@ -423,6 +382,128 @@ impl RemoteStorage for AzureBlobStorage {
}
}
async fn permit(
&self,
kind: RequestKind,
cancel: &CancellationToken,
) -> Result<tokio::sync::SemaphorePermit<'_>, Cancelled> {
let acquire = self.concurrency_limiter.acquire(kind);
tokio::select! {
permit = acquire => Ok(permit.expect("never closed")),
_ = cancel.cancelled() => Err(Cancelled),
}
}
pub fn container_name(&self) -> &str {
&self.container_name
}
}
trait ListingCollector {
fn add_prefixes(&mut self, abs: &AzureBlobStorage, prefix_it: impl Iterator<Item = RemotePath>);
fn add_blob(&mut self, abs: &AzureBlobStorage, blob: &Blob);
}
impl ListingCollector for Listing {
fn add_prefixes(
&mut self,
_abs: &AzureBlobStorage,
prefix_it: impl Iterator<Item = RemotePath>,
) {
self.prefixes.extend(prefix_it);
}
fn add_blob(&mut self, abs: &AzureBlobStorage, blob: &Blob) {
self.keys.push(ListingObject {
key: abs.name_to_relative_path(&blob.name),
last_modified: blob.properties.last_modified.into(),
size: blob.properties.content_length,
});
}
}
impl ListingCollector for crate::VersionListing {
fn add_prefixes(
&mut self,
_abs: &AzureBlobStorage,
_prefix_it: impl Iterator<Item = RemotePath>,
) {
// nothing
}
fn add_blob(&mut self, abs: &AzureBlobStorage, blob: &Blob) {
let id = crate::VersionId(blob.version_id.clone().expect("didn't find version ID"));
self.versions.push(crate::Version {
key: abs.name_to_relative_path(&blob.name),
last_modified: blob.properties.last_modified.into(),
kind: crate::VersionKind::Version(id),
});
}
}
fn to_azure_metadata(metadata: StorageMetadata) -> Metadata {
let mut res = Metadata::new();
for (k, v) in metadata.0.into_iter() {
res.insert(k, v);
}
res
}
fn to_download_error(error: azure_core::Error) -> DownloadError {
if let Some(http_err) = error.as_http_error() {
match http_err.status() {
StatusCode::NotFound => DownloadError::NotFound,
StatusCode::NotModified => DownloadError::Unmodified,
StatusCode::BadRequest => DownloadError::BadInput(anyhow::Error::new(error)),
_ => DownloadError::Other(anyhow::Error::new(error)),
}
} else {
DownloadError::Other(error.into())
}
}
impl RemoteStorage for AzureBlobStorage {
fn list_streaming(
&self,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> impl Stream<Item = Result<Listing, DownloadError>> {
let customize_builder = |builder| builder;
let kind = RequestKind::ListVersions;
self.list_streaming_for_fn(prefix, mode, max_keys, cancel, kind, customize_builder)
}
async fn list_versions(
&self,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> std::result::Result<crate::VersionListing, DownloadError> {
let customize_builder = |mut builder: ListBlobsBuilder| {
builder = builder.include_versions(true);
builder
};
let kind = RequestKind::ListVersions;
let mut stream = std::pin::pin!(self.list_streaming_for_fn(
prefix,
mode,
max_keys,
cancel,
kind,
customize_builder
));
let mut combined: crate::VersionListing =
stream.next().await.expect("At least one item required")?;
while let Some(list) = stream.next().await {
let list = list?;
combined.versions.extend(list.versions.into_iter());
}
Ok(combined)
}
async fn head_object(
&self,
key: &RemotePath,
@@ -532,7 +613,12 @@ impl RemoteStorage for AzureBlobStorage {
let mut builder = blob_client.get();
if let Some(ref etag) = opts.etag {
builder = builder.if_match(IfMatchCondition::NotMatch(etag.to_string()))
builder = builder.if_match(IfMatchCondition::NotMatch(etag.to_string()));
}
if let Some(ref version_id) = opts.version_id {
let version_id = azure_storage_blobs::prelude::VersionId::new(version_id.0.clone());
builder = builder.blob_versioning(version_id);
}
if let Some((start, end)) = opts.byte_range() {

View File

@@ -176,6 +176,32 @@ pub struct Listing {
pub keys: Vec<ListingObject>,
}
#[derive(Default)]
pub struct VersionListing {
pub versions: Vec<Version>,
}
pub struct Version {
pub key: RemotePath,
pub last_modified: SystemTime,
pub kind: VersionKind,
}
impl Version {
pub fn version_id(&self) -> Option<&VersionId> {
match &self.kind {
VersionKind::Version(id) => Some(id),
VersionKind::DeletionMarker => None,
}
}
}
#[derive(Debug)]
pub enum VersionKind {
DeletionMarker,
Version(VersionId),
}
/// Options for downloads. The default value is a plain GET.
pub struct DownloadOpts {
/// If given, returns [`DownloadError::Unmodified`] if the object still has
@@ -186,6 +212,8 @@ pub struct DownloadOpts {
/// The end of the byte range to download, or unbounded. Must be after the
/// start bound.
pub byte_end: Bound<u64>,
/// Optionally request a specific version of a key
pub version_id: Option<VersionId>,
/// Indicate whether we're downloading something small or large: this indirectly controls
/// timeouts: for something like an index/manifest/heatmap, we should time out faster than
/// for layer files
@@ -197,12 +225,16 @@ pub enum DownloadKind {
Small,
}
#[derive(Debug, Clone)]
pub struct VersionId(pub String);
impl Default for DownloadOpts {
fn default() -> Self {
Self {
etag: Default::default(),
byte_start: Bound::Unbounded,
byte_end: Bound::Unbounded,
version_id: None,
kind: DownloadKind::Large,
}
}
@@ -295,6 +327,14 @@ pub trait RemoteStorage: Send + Sync + 'static {
Ok(combined)
}
async fn list_versions(
&self,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> Result<VersionListing, DownloadError>;
/// Obtain metadata information about an object.
async fn head_object(
&self,
@@ -475,6 +515,22 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
}
}
// See [`RemoteStorage::list_versions`].
pub async fn list_versions<'a>(
&'a self,
prefix: Option<&'a RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &'a CancellationToken,
) -> Result<VersionListing, DownloadError> {
match self {
Self::LocalFs(s) => s.list_versions(prefix, mode, max_keys, cancel).await,
Self::AwsS3(s) => s.list_versions(prefix, mode, max_keys, cancel).await,
Self::AzureBlob(s) => s.list_versions(prefix, mode, max_keys, cancel).await,
Self::Unreliable(s) => s.list_versions(prefix, mode, max_keys, cancel).await,
}
}
// See [`RemoteStorage::head_object`].
pub async fn head_object(
&self,
@@ -727,6 +783,7 @@ impl ConcurrencyLimiter {
RequestKind::Copy => &self.write,
RequestKind::TimeTravel => &self.write,
RequestKind::Head => &self.read,
RequestKind::ListVersions => &self.read,
}
}

View File

@@ -445,6 +445,16 @@ impl RemoteStorage for LocalFs {
}
}
async fn list_versions(
&self,
_prefix: Option<&RemotePath>,
_mode: ListingMode,
_max_keys: Option<NonZeroU32>,
_cancel: &CancellationToken,
) -> Result<crate::VersionListing, DownloadError> {
unimplemented!()
}
async fn head_object(
&self,
key: &RemotePath,

View File

@@ -14,6 +14,7 @@ pub(crate) enum RequestKind {
Copy = 4,
TimeTravel = 5,
Head = 6,
ListVersions = 7,
}
use RequestKind::*;
@@ -29,6 +30,7 @@ impl RequestKind {
Copy => "copy_object",
TimeTravel => "time_travel_recover",
Head => "head_object",
ListVersions => "list_versions",
}
}
const fn as_index(&self) -> usize {
@@ -36,7 +38,10 @@ impl RequestKind {
}
}
const REQUEST_KIND_COUNT: usize = 7;
const REQUEST_KIND_LIST: &[RequestKind] =
&[Get, Put, Delete, List, Copy, TimeTravel, Head, ListVersions];
const REQUEST_KIND_COUNT: usize = REQUEST_KIND_LIST.len();
pub(crate) struct RequestTyped<C>([C; REQUEST_KIND_COUNT]);
impl<C> RequestTyped<C> {
@@ -45,12 +50,11 @@ impl<C> RequestTyped<C> {
}
fn build_with(mut f: impl FnMut(RequestKind) -> C) -> Self {
use RequestKind::*;
let mut it = [Get, Put, Delete, List, Copy, TimeTravel, Head].into_iter();
let mut it = REQUEST_KIND_LIST.iter();
let arr = std::array::from_fn::<C, REQUEST_KIND_COUNT, _>(|index| {
let next = it.next().unwrap();
assert_eq!(index, next.as_index());
f(next)
f(*next)
});
if let Some(next) = it.next() {

View File

@@ -21,9 +21,8 @@ use aws_sdk_s3::config::{AsyncSleep, IdentityCache, Region, SharedAsyncSleep};
use aws_sdk_s3::error::SdkError;
use aws_sdk_s3::operation::get_object::GetObjectError;
use aws_sdk_s3::operation::head_object::HeadObjectError;
use aws_sdk_s3::types::{Delete, DeleteMarkerEntry, ObjectIdentifier, ObjectVersion, StorageClass};
use aws_sdk_s3::types::{Delete, ObjectIdentifier, StorageClass};
use aws_smithy_async::rt::sleep::TokioSleep;
use aws_smithy_types::DateTime;
use aws_smithy_types::body::SdkBody;
use aws_smithy_types::byte_stream::ByteStream;
use aws_smithy_types::date_time::ConversionError;
@@ -46,7 +45,7 @@ use crate::support::PermitCarrying;
use crate::{
ConcurrencyLimiter, Download, DownloadError, DownloadOpts, Listing, ListingMode, ListingObject,
MAX_KEYS_PER_DELETE_S3, REMOTE_STORAGE_PREFIX_SEPARATOR, RemotePath, RemoteStorage,
TimeTravelError, TimeoutOrCancel,
TimeTravelError, TimeoutOrCancel, Version, VersionId, VersionKind, VersionListing,
};
/// AWS S3 storage.
@@ -66,6 +65,7 @@ struct GetObjectRequest {
key: String,
etag: Option<String>,
range: Option<String>,
version_id: Option<String>,
}
impl S3Bucket {
/// Creates the S3 storage, errors if incorrect AWS S3 configuration provided.
@@ -251,6 +251,7 @@ impl S3Bucket {
.get_object()
.bucket(request.bucket)
.key(request.key)
.set_version_id(request.version_id)
.set_range(request.range);
if let Some(etag) = request.etag {
@@ -405,6 +406,124 @@ impl S3Bucket {
Ok(())
}
async fn list_versions_with_permit(
&self,
_permit: &tokio::sync::SemaphorePermit<'_>,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> Result<crate::VersionListing, DownloadError> {
// get the passed prefix or if it is not set use prefix_in_bucket value
let prefix = prefix
.map(|p| self.relative_path_to_s3_object(p))
.or_else(|| self.prefix_in_bucket.clone());
let warn_threshold = 3;
let max_retries = 10;
let is_permanent = |e: &_| matches!(e, DownloadError::Cancelled);
let mut key_marker = None;
let mut version_id_marker = None;
let mut versions_and_deletes = Vec::new();
loop {
let response = backoff::retry(
|| async {
let mut request = self
.client
.list_object_versions()
.bucket(self.bucket_name.clone())
.set_prefix(prefix.clone())
.set_key_marker(key_marker.clone())
.set_version_id_marker(version_id_marker.clone());
if let ListingMode::WithDelimiter = mode {
request = request.delimiter(REMOTE_STORAGE_PREFIX_SEPARATOR.to_string());
}
let op = request.send();
tokio::select! {
res = op => res.map_err(|e| DownloadError::Other(e.into())),
_ = cancel.cancelled() => Err(DownloadError::Cancelled),
}
},
is_permanent,
warn_threshold,
max_retries,
"listing object versions",
cancel,
)
.await
.ok_or_else(|| DownloadError::Cancelled)
.and_then(|x| x)?;
tracing::trace!(
" Got List response version_id_marker={:?}, key_marker={:?}",
response.version_id_marker,
response.key_marker
);
let versions = response
.versions
.unwrap_or_default()
.into_iter()
.map(|version| {
let key = version.key.expect("response does not contain a key");
let key = self.s3_object_to_relative_path(&key);
let version_id = VersionId(version.version_id.expect("needing version id"));
let last_modified =
SystemTime::try_from(version.last_modified.expect("no last_modified"))?;
Ok(Version {
key,
last_modified,
kind: crate::VersionKind::Version(version_id),
})
});
let deletes = response
.delete_markers
.unwrap_or_default()
.into_iter()
.map(|version| {
let key = version.key.expect("response does not contain a key");
let key = self.s3_object_to_relative_path(&key);
let last_modified =
SystemTime::try_from(version.last_modified.expect("no last_modified"))?;
Ok(Version {
key,
last_modified,
kind: crate::VersionKind::DeletionMarker,
})
});
itertools::process_results(versions.chain(deletes), |n_vds| {
versions_and_deletes.extend(n_vds)
})
.map_err(DownloadError::Other)?;
fn none_if_empty(v: Option<String>) -> Option<String> {
v.filter(|v| !v.is_empty())
}
version_id_marker = none_if_empty(response.next_version_id_marker);
key_marker = none_if_empty(response.next_key_marker);
if version_id_marker.is_none() {
// The final response is not supposed to be truncated
if response.is_truncated.unwrap_or_default() {
return Err(DownloadError::Other(anyhow::anyhow!(
"Received truncated ListObjectVersions response for prefix={prefix:?}"
)));
}
break;
}
if let Some(max_keys) = max_keys {
if versions_and_deletes.len() >= max_keys.get().try_into().unwrap() {
return Err(DownloadError::Other(anyhow::anyhow!("too many versions")));
}
}
}
Ok(VersionListing {
versions: versions_and_deletes,
})
}
pub fn bucket_name(&self) -> &str {
&self.bucket_name
}
@@ -621,6 +740,19 @@ impl RemoteStorage for S3Bucket {
}
}
async fn list_versions(
&self,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> Result<crate::VersionListing, DownloadError> {
let kind = RequestKind::ListVersions;
let permit = self.permit(kind, cancel).await?;
self.list_versions_with_permit(&permit, prefix, mode, max_keys, cancel)
.await
}
async fn head_object(
&self,
key: &RemotePath,
@@ -801,6 +933,7 @@ impl RemoteStorage for S3Bucket {
key: self.relative_path_to_s3_object(from),
etag: opts.etag.as_ref().map(|e| e.to_string()),
range: opts.byte_range_header(),
version_id: opts.version_id.as_ref().map(|v| v.0.to_owned()),
},
cancel,
)
@@ -845,94 +978,25 @@ impl RemoteStorage for S3Bucket {
let kind = RequestKind::TimeTravel;
let permit = self.permit(kind, cancel).await?;
let timestamp = DateTime::from(timestamp);
let done_if_after = DateTime::from(done_if_after);
tracing::trace!("Target time: {timestamp:?}, done_if_after {done_if_after:?}");
// get the passed prefix or if it is not set use prefix_in_bucket value
let prefix = prefix
.map(|p| self.relative_path_to_s3_object(p))
.or_else(|| self.prefix_in_bucket.clone());
// Limit the number of versions deletions, mostly so that we don't
// keep requesting forever if the list is too long, as we'd put the
// list in RAM.
// Building a list of 100k entries that reaches the limit roughly takes
// 40 seconds, and roughly corresponds to tenants of 2 TiB physical size.
const COMPLEXITY_LIMIT: Option<NonZeroU32> = NonZeroU32::new(100_000);
let warn_threshold = 3;
let max_retries = 10;
let is_permanent = |e: &_| matches!(e, TimeTravelError::Cancelled);
let mut key_marker = None;
let mut version_id_marker = None;
let mut versions_and_deletes = Vec::new();
loop {
let response = backoff::retry(
|| async {
let op = self
.client
.list_object_versions()
.bucket(self.bucket_name.clone())
.set_prefix(prefix.clone())
.set_key_marker(key_marker.clone())
.set_version_id_marker(version_id_marker.clone())
.send();
tokio::select! {
res = op => res.map_err(|e| TimeTravelError::Other(e.into())),
_ = cancel.cancelled() => Err(TimeTravelError::Cancelled),
}
},
is_permanent,
warn_threshold,
max_retries,
"listing object versions for time_travel_recover",
cancel,
)
let mode = ListingMode::NoDelimiter;
let version_listing = self
.list_versions_with_permit(&permit, prefix, mode, COMPLEXITY_LIMIT, cancel)
.await
.ok_or_else(|| TimeTravelError::Cancelled)
.and_then(|x| x)?;
tracing::trace!(
" Got List response version_id_marker={:?}, key_marker={:?}",
response.version_id_marker,
response.key_marker
);
let versions = response
.versions
.unwrap_or_default()
.into_iter()
.map(VerOrDelete::from_version);
let deletes = response
.delete_markers
.unwrap_or_default()
.into_iter()
.map(VerOrDelete::from_delete_marker);
itertools::process_results(versions.chain(deletes), |n_vds| {
versions_and_deletes.extend(n_vds)
})
.map_err(TimeTravelError::Other)?;
fn none_if_empty(v: Option<String>) -> Option<String> {
v.filter(|v| !v.is_empty())
}
version_id_marker = none_if_empty(response.next_version_id_marker);
key_marker = none_if_empty(response.next_key_marker);
if version_id_marker.is_none() {
// The final response is not supposed to be truncated
if response.is_truncated.unwrap_or_default() {
return Err(TimeTravelError::Other(anyhow::anyhow!(
"Received truncated ListObjectVersions response for prefix={prefix:?}"
)));
}
break;
}
// Limit the number of versions deletions, mostly so that we don't
// keep requesting forever if the list is too long, as we'd put the
// list in RAM.
// Building a list of 100k entries that reaches the limit roughly takes
// 40 seconds, and roughly corresponds to tenants of 2 TiB physical size.
const COMPLEXITY_LIMIT: usize = 100_000;
if versions_and_deletes.len() >= COMPLEXITY_LIMIT {
return Err(TimeTravelError::TooManyVersions);
}
}
.map_err(|err| match err {
DownloadError::Other(e) => TimeTravelError::Other(e),
DownloadError::Cancelled => TimeTravelError::Cancelled,
other => TimeTravelError::Other(other.into()),
})?;
let versions_and_deletes = version_listing.versions;
tracing::info!(
"Built list for time travel with {} versions and deletions",
@@ -948,24 +1012,26 @@ impl RemoteStorage for S3Bucket {
let mut vds_for_key = HashMap::<_, Vec<_>>::new();
for vd in &versions_and_deletes {
let VerOrDelete {
version_id, key, ..
} = &vd;
if version_id == "null" {
let Version { key, .. } = &vd;
let version_id = vd.version_id().map(|v| v.0.as_str());
if version_id == Some("null") {
return Err(TimeTravelError::Other(anyhow!(
"Received ListVersions response for key={key} with version_id='null', \
indicating either disabled versioning, or legacy objects with null version id values"
)));
}
tracing::trace!(
"Parsing version key={key} version_id={version_id} kind={:?}",
vd.kind
);
tracing::trace!("Parsing version key={key} kind={:?}", vd.kind);
vds_for_key.entry(key).or_default().push(vd);
}
let warn_threshold = 3;
let max_retries = 10;
let is_permanent = |e: &_| matches!(e, TimeTravelError::Cancelled);
for (key, versions) in vds_for_key {
let last_vd = versions.last().unwrap();
let key = self.relative_path_to_s3_object(key);
if last_vd.last_modified > done_if_after {
tracing::trace!("Key {key} has version later than done_if_after, skipping");
continue;
@@ -990,11 +1056,11 @@ impl RemoteStorage for S3Bucket {
do_delete = true;
} else {
match &versions[version_to_restore_to - 1] {
VerOrDelete {
kind: VerOrDeleteKind::Version,
version_id,
Version {
kind: VersionKind::Version(version_id),
..
} => {
let version_id = &version_id.0;
tracing::trace!("Copying old version {version_id} for {key}...");
// Restore the state to the last version by copying
let source_id =
@@ -1006,7 +1072,7 @@ impl RemoteStorage for S3Bucket {
.client
.copy_object()
.bucket(self.bucket_name.clone())
.key(key)
.key(&key)
.set_storage_class(self.upload_storage_class.clone())
.copy_source(&source_id)
.send();
@@ -1027,8 +1093,8 @@ impl RemoteStorage for S3Bucket {
.and_then(|x| x)?;
tracing::info!(%version_id, %key, "Copied old version in S3");
}
VerOrDelete {
kind: VerOrDeleteKind::DeleteMarker,
Version {
kind: VersionKind::DeletionMarker,
..
} => {
do_delete = true;
@@ -1036,7 +1102,7 @@ impl RemoteStorage for S3Bucket {
}
};
if do_delete {
if matches!(last_vd.kind, VerOrDeleteKind::DeleteMarker) {
if matches!(last_vd.kind, VersionKind::DeletionMarker) {
// Key has since been deleted (but there was some history), no need to do anything
tracing::trace!("Key {key} already deleted, skipping.");
} else {
@@ -1064,62 +1130,6 @@ impl RemoteStorage for S3Bucket {
}
}
// Save RAM and only store the needed data instead of the entire ObjectVersion/DeleteMarkerEntry
struct VerOrDelete {
kind: VerOrDeleteKind,
last_modified: DateTime,
version_id: String,
key: String,
}
#[derive(Debug)]
enum VerOrDeleteKind {
Version,
DeleteMarker,
}
impl VerOrDelete {
fn with_kind(
kind: VerOrDeleteKind,
last_modified: Option<DateTime>,
version_id: Option<String>,
key: Option<String>,
) -> anyhow::Result<Self> {
let lvk = (last_modified, version_id, key);
let (Some(last_modified), Some(version_id), Some(key)) = lvk else {
anyhow::bail!(
"One (or more) of last_modified, key, and id is None. \
Is versioning enabled in the bucket? last_modified={:?}, version_id={:?}, key={:?}",
lvk.0,
lvk.1,
lvk.2,
);
};
Ok(Self {
kind,
last_modified,
version_id,
key,
})
}
fn from_version(v: ObjectVersion) -> anyhow::Result<Self> {
Self::with_kind(
VerOrDeleteKind::Version,
v.last_modified,
v.version_id,
v.key,
)
}
fn from_delete_marker(v: DeleteMarkerEntry) -> anyhow::Result<Self> {
Self::with_kind(
VerOrDeleteKind::DeleteMarker,
v.last_modified,
v.version_id,
v.key,
)
}
}
#[cfg(test)]
mod tests {
use std::num::NonZeroUsize;

View File

@@ -139,6 +139,20 @@ impl RemoteStorage for UnreliableWrapper {
self.inner.list(prefix, mode, max_keys, cancel).await
}
async fn list_versions(
&self,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> Result<crate::VersionListing, DownloadError> {
self.attempt(RemoteOp::ListPrefixes(prefix.cloned()))
.map_err(DownloadError::Other)?;
self.inner
.list_versions(prefix, mode, max_keys, cancel)
.await
}
async fn head_object(
&self,
key: &RemotePath,

View File

@@ -29,6 +29,7 @@ futures = { workspace = true }
jsonwebtoken.workspace = true
nix = { workspace = true, features = ["ioctl"] }
once_cell.workspace = true
pem.workspace = true
pin-project-lite.workspace = true
regex.workspace = true
serde.workspace = true

View File

@@ -11,7 +11,8 @@ use camino::Utf8Path;
use jsonwebtoken::{
Algorithm, DecodingKey, EncodingKey, Header, TokenData, Validation, decode, encode,
};
use serde::{Deserialize, Serialize};
use pem::Pem;
use serde::{Deserialize, Serialize, de::DeserializeOwned};
use crate::id::TenantId;
@@ -73,7 +74,10 @@ impl SwappableJwtAuth {
pub fn swap(&self, jwt_auth: JwtAuth) {
self.0.swap(Arc::new(jwt_auth));
}
pub fn decode(&self, token: &str) -> std::result::Result<TokenData<Claims>, AuthError> {
pub fn decode<D: DeserializeOwned>(
&self,
token: &str,
) -> std::result::Result<TokenData<D>, AuthError> {
self.0.load().decode(token)
}
}
@@ -148,7 +152,10 @@ impl JwtAuth {
/// The function tries the stored decoding keys in succession,
/// and returns the first yielding a successful result.
/// If there is no working decoding key, it returns the last error.
pub fn decode(&self, token: &str) -> std::result::Result<TokenData<Claims>, AuthError> {
pub fn decode<D: DeserializeOwned>(
&self,
token: &str,
) -> std::result::Result<TokenData<D>, AuthError> {
let mut res = None;
for decoding_key in &self.decoding_keys {
res = Some(decode(token, decoding_key, &self.validation));
@@ -173,8 +180,8 @@ impl std::fmt::Debug for JwtAuth {
}
// this function is used only for testing purposes in CLI e g generate tokens during init
pub fn encode_from_key_file<S: Serialize>(claims: &S, key_data: &[u8]) -> Result<String> {
let key = EncodingKey::from_ed_pem(key_data)?;
pub fn encode_from_key_file<S: Serialize>(claims: &S, pem: &Pem) -> Result<String> {
let key = EncodingKey::from_ed_der(pem.contents());
Ok(encode(&Header::new(STORAGE_TOKEN_ALGORITHM), claims, &key)?)
}
@@ -188,13 +195,13 @@ mod tests {
//
// openssl genpkey -algorithm ed25519 -out ed25519-priv.pem
// openssl pkey -in ed25519-priv.pem -pubout -out ed25519-pub.pem
const TEST_PUB_KEY_ED25519: &[u8] = br#"
const TEST_PUB_KEY_ED25519: &str = r#"
-----BEGIN PUBLIC KEY-----
MCowBQYDK2VwAyEARYwaNBayR+eGI0iXB4s3QxE3Nl2g1iWbr6KtLWeVD/w=
-----END PUBLIC KEY-----
"#;
const TEST_PRIV_KEY_ED25519: &[u8] = br#"
const TEST_PRIV_KEY_ED25519: &str = r#"
-----BEGIN PRIVATE KEY-----
MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
-----END PRIVATE KEY-----
@@ -222,9 +229,9 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
// Check it can be validated with the public key
let auth = JwtAuth::new(vec![
DecodingKey::from_ed_pem(TEST_PUB_KEY_ED25519).unwrap(),
DecodingKey::from_ed_pem(TEST_PUB_KEY_ED25519.as_bytes()).unwrap(),
]);
let claims_from_token = auth.decode(encoded_eddsa).unwrap().claims;
let claims_from_token: Claims = auth.decode(encoded_eddsa).unwrap().claims;
assert_eq!(claims_from_token, expected_claims);
}
@@ -235,13 +242,14 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
scope: Scope::Tenant,
};
let encoded = encode_from_key_file(&claims, TEST_PRIV_KEY_ED25519).unwrap();
let pem = pem::parse(TEST_PRIV_KEY_ED25519).unwrap();
let encoded = encode_from_key_file(&claims, &pem).unwrap();
// decode it back
let auth = JwtAuth::new(vec![
DecodingKey::from_ed_pem(TEST_PUB_KEY_ED25519).unwrap(),
DecodingKey::from_ed_pem(TEST_PUB_KEY_ED25519.as_bytes()).unwrap(),
]);
let decoded = auth.decode(&encoded).unwrap();
let decoded: TokenData<Claims> = auth.decode(&encoded).unwrap();
assert_eq!(decoded.claims, claims);
}

View File

@@ -10,6 +10,8 @@ default = []
# which adds some runtime cost to run tests on outage conditions
testing = ["fail/failpoints", "pageserver_api/testing", "wal_decoder/testing", "pageserver_client/testing"]
fuzz-read-path = ["testing"]
[dependencies]
anyhow.workspace = true
arc-swap.workspace = true
@@ -33,6 +35,7 @@ humantime.workspace = true
humantime-serde.workspace = true
hyper0.workspace = true
itertools.workspace = true
jsonwebtoken.workspace = true
md5.workspace = true
nix.workspace = true
# hack to get the number of worker threads tokio uses
@@ -75,6 +78,7 @@ metrics.workspace = true
pageserver_api.workspace = true
pageserver_client.workspace = true # for ResponseErrorMessageExt TOOD refactor that
pageserver_compaction.workspace = true
pem.workspace = true
postgres_connection.workspace = true
postgres_ffi.workspace = true
pq_proto.workspace = true

View File

@@ -11,6 +11,7 @@ use pageserver::task_mgr::TaskKind;
use pageserver::tenant::storage_layer::InMemoryLayer;
use pageserver::{page_cache, virtual_file};
use pageserver_api::key::Key;
use pageserver_api::models::virtual_file::IoMode;
use pageserver_api::shard::TenantShardId;
use pageserver_api::value::Value;
use tokio_util::sync::CancellationToken;
@@ -21,13 +22,14 @@ use wal_decoder::serialized_batch::SerializedValueBatch;
// A very cheap hash for generating non-sequential keys.
fn murmurhash32(mut h: u32) -> u32 {
h ^= h >> 16;
h = h.wrapping_mul(0x85ebca6b);
h h.wrapping_mul(0x85ebca6b);
h ^= h >> 13;
h = h.wrapping_mul(0xc2b2ae35);
h ^= h >> 16;
h
}
#[derive(serde::Serialize, Clone, Copy, Debug)]
enum KeyLayout {
/// Sequential unique keys
Sequential,
@@ -37,6 +39,7 @@ enum KeyLayout {
RandomReuse(u32),
}
#[derive(serde::Serialize, Clone, Copy, Debug)]
enum WriteDelta {
Yes,
No,
@@ -126,7 +129,7 @@ async fn ingest(
max_concurrency: NonZeroUsize::new(1).unwrap(),
});
let (_desc, path) = layer
.write_to_disk(&ctx, None, l0_flush_state.inner())
.write_to_disk(&ctx, None, l0_flush_state.inner(), &gate, cancel.clone())
.await?
.unwrap();
tokio::fs::remove_file(path).await?;
@@ -138,12 +141,15 @@ async fn ingest(
/// Wrapper to instantiate a tokio runtime
fn ingest_main(
conf: &'static PageServerConf,
io_mode: IoMode,
put_size: usize,
put_count: usize,
key_layout: KeyLayout,
write_delta: WriteDelta,
) {
let runtime = tokio::runtime::Builder::new_current_thread()
pageserver::virtual_file::set_io_mode(io_mode);
let runtime = tokio::runtime::Builder::new_multi_thread()
.enable_all()
.build()
.unwrap();
@@ -174,93 +180,207 @@ fn criterion_benchmark(c: &mut Criterion) {
virtual_file::init(
16384,
virtual_file::io_engine_for_bench(),
// immaterial, each `ingest_main` invocation below overrides this
conf.virtual_file_io_mode,
// without actually doing syncs, buffered writes have an unfair advantage over direct IO writes
virtual_file::SyncMode::Sync,
);
page_cache::init(conf.page_cache_size);
{
let mut group = c.benchmark_group("ingest-small-values");
let put_size = 100usize;
let put_count = 128 * 1024 * 1024 / put_size;
group.throughput(criterion::Throughput::Bytes((put_size * put_count) as u64));
group.sample_size(10);
group.bench_function("ingest 128MB/100b seq", |b| {
b.iter(|| {
ingest_main(
conf,
put_size,
put_count,
KeyLayout::Sequential,
WriteDelta::Yes,
)
})
});
group.bench_function("ingest 128MB/100b rand", |b| {
b.iter(|| {
ingest_main(
conf,
put_size,
put_count,
KeyLayout::Random,
WriteDelta::Yes,
)
})
});
group.bench_function("ingest 128MB/100b rand-1024keys", |b| {
b.iter(|| {
ingest_main(
conf,
put_size,
put_count,
KeyLayout::RandomReuse(0x3ff),
WriteDelta::Yes,
)
})
});
group.bench_function("ingest 128MB/100b seq, no delta", |b| {
b.iter(|| {
ingest_main(
conf,
put_size,
put_count,
KeyLayout::Sequential,
WriteDelta::No,
)
})
});
#[derive(serde::Serialize)]
struct ExplodedParameters {
io_mode: IoMode,
volume_mib: usize,
key_size: usize,
key_layout: KeyLayout,
write_delta: WriteDelta,
}
{
let mut group = c.benchmark_group("ingest-big-values");
let put_size = 8192usize;
let put_count = 128 * 1024 * 1024 / put_size;
group.throughput(criterion::Throughput::Bytes((put_size * put_count) as u64));
#[derive(Clone)]
struct HandPickedParameters {
volume_mib: usize,
key_size: usize,
key_layout: KeyLayout,
write_delta: WriteDelta,
}
let expect = vec![
// Small values (100b) tests
HandPickedParameters {
volume_mib: 128,
key_size: 100,
key_layout: KeyLayout::Sequential,
write_delta: WriteDelta::Yes,
},
HandPickedParameters {
volume_mib: 128,
key_size: 100,
key_layout: KeyLayout::Random,
write_delta: WriteDelta::Yes,
},
HandPickedParameters {
volume_mib: 128,
key_size: 100,
key_layout: KeyLayout::RandomReuse(0x3ff),
write_delta: WriteDelta::Yes,
},
HandPickedParameters {
volume_mib: 128,
key_size: 100,
key_layout: KeyLayout::Sequential,
write_delta: WriteDelta::No,
},
// Large values (8k) tests
HandPickedParameters {
volume_mib: 128,
key_size: 8192,
key_layout: KeyLayout::Sequential,
write_delta: WriteDelta::Yes,
},
HandPickedParameters {
volume_mib: 128,
key_size: 8192,
key_layout: KeyLayout::Sequential,
write_delta: WriteDelta::No,
},
];
let exploded_parameters = {
let mut out = Vec::new();
for io_mode in [
IoMode::Buffered,
#[cfg(target_os = "linux")]
IoMode::Direct,
] {
for param in expect.clone() {
let HandPickedParameters {
volume_mib,
key_size,
key_layout,
write_delta,
} = param;
out.push(ExplodedParameters {
io_mode,
volume_mib,
key_size,
key_layout,
write_delta,
});
}
}
out
};
impl ExplodedParameters {
fn benchmark_id(&self) -> String {
let ExplodedParameters {
io_mode,
volume_mib,
key_size,
key_layout,
write_delta,
} = self;
format!(
"io_mode={io_mode:?} volume_mib={volume_mib:?} key_size_bytes={key_size:?} key_layout={key_layout:?} write_delta={write_delta:?}"
)
}
}
let mut group = c.benchmark_group("ingest");
for params in exploded_parameters {
let id = params.benchmark_id();
let ExplodedParameters {
io_mode,
volume_mib,
key_size,
key_layout,
write_delta,
} = params;
let put_count = volume_mib * 1024 * 1024 / key_size;
group.throughput(criterion::Throughput::Bytes((key_size * put_count) as u64));
group.sample_size(10);
group.bench_function("ingest 128MB/8k seq", |b| {
b.iter(|| {
ingest_main(
conf,
put_size,
put_count,
KeyLayout::Sequential,
WriteDelta::Yes,
)
})
});
group.bench_function("ingest 128MB/8k seq, no delta", |b| {
b.iter(|| {
ingest_main(
conf,
put_size,
put_count,
KeyLayout::Sequential,
WriteDelta::No,
)
})
group.bench_function(id, |b| {
b.iter(|| ingest_main(conf, io_mode, key_size, put_count, key_layout, write_delta))
});
}
}
criterion_group!(benches, criterion_benchmark);
criterion_main!(benches);
/*
cargo bench --bench bench_ingest
im4gn.2xlarge:
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=Yes
time: [1.8491 s 1.8540 s 1.8592 s]
thrpt: [68.847 MiB/s 69.039 MiB/s 69.222 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=Random write_delta=Yes
time: [2.6976 s 2.7123 s 2.7286 s]
thrpt: [46.911 MiB/s 47.193 MiB/s 47.450 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=RandomReuse(1023) write_delta=Y...
time: [1.7433 s 1.7510 s 1.7600 s]
thrpt: [72.729 MiB/s 73.099 MiB/s 73.423 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=No
time: [499.63 ms 500.07 ms 500.46 ms]
thrpt: [255.77 MiB/s 255.96 MiB/s 256.19 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=Yes
time: [456.97 ms 459.61 ms 461.92 ms]
thrpt: [277.11 MiB/s 278.50 MiB/s 280.11 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=No
time: [158.82 ms 159.16 ms 159.56 ms]
thrpt: [802.22 MiB/s 804.24 MiB/s 805.93 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=Yes
time: [1.8856 s 1.8997 s 1.9179 s]
thrpt: [66.740 MiB/s 67.380 MiB/s 67.882 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=Random write_delta=Yes
time: [2.7468 s 2.7625 s 2.7785 s]
thrpt: [46.068 MiB/s 46.335 MiB/s 46.600 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=RandomReuse(1023) write_delta=Yes
time: [1.7689 s 1.7726 s 1.7767 s]
thrpt: [72.045 MiB/s 72.208 MiB/s 72.363 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=No
time: [497.64 ms 498.60 ms 499.67 ms]
thrpt: [256.17 MiB/s 256.72 MiB/s 257.21 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=Yes
time: [493.72 ms 505.07 ms 518.03 ms]
thrpt: [247.09 MiB/s 253.43 MiB/s 259.26 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=No
time: [267.76 ms 267.85 ms 267.96 ms]
thrpt: [477.69 MiB/s 477.88 MiB/s 478.03 MiB/s]
Hetzner AX102:
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=Yes
time: [1.0683 s 1.1006 s 1.1386 s]
thrpt: [112.42 MiB/s 116.30 MiB/s 119.82 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=Random write_delta=Yes
time: [1.5719 s 1.6012 s 1.6228 s]
thrpt: [78.877 MiB/s 79.938 MiB/s 81.430 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=RandomReuse(1023) write_delta=Y...
time: [1.1095 s 1.1331 s 1.1580 s]
thrpt: [110.53 MiB/s 112.97 MiB/s 115.37 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=No
time: [303.20 ms 307.83 ms 311.90 ms]
thrpt: [410.39 MiB/s 415.81 MiB/s 422.16 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=Yes
time: [406.34 ms 429.37 ms 451.63 ms]
thrpt: [283.42 MiB/s 298.11 MiB/s 315.00 MiB/s]
ingest/io_mode=Buffered volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=No
time: [134.01 ms 135.78 ms 137.48 ms]
thrpt: [931.03 MiB/s 942.68 MiB/s 955.12 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=Yes
time: [1.0406 s 1.0580 s 1.0772 s]
thrpt: [118.83 MiB/s 120.98 MiB/s 123.00 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=Random write_delta=Yes
time: [1.5059 s 1.5339 s 1.5625 s]
thrpt: [81.920 MiB/s 83.448 MiB/s 84.999 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=RandomReuse(1023) write_delta=Yes
time: [1.0714 s 1.0934 s 1.1161 s]
thrpt: [114.69 MiB/s 117.06 MiB/s 119.47 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=100 key_layout=Sequential write_delta=No
time: [262.68 ms 265.14 ms 267.71 ms]
thrpt: [478.13 MiB/s 482.76 MiB/s 487.29 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=Yes
time: [375.19 ms 393.80 ms 411.40 ms]
thrpt: [311.14 MiB/s 325.04 MiB/s 341.16 MiB/s]
ingest/io_mode=Direct volume_mib=128 key_size_bytes=8192 key_layout=Sequential write_delta=No
time: [123.02 ms 123.85 ms 124.66 ms]
thrpt: [1.0027 GiB/s 1.0093 GiB/s 1.0161 GiB/s]
*/

View File

@@ -65,7 +65,7 @@ use bytes::{Buf, Bytes};
use criterion::{BenchmarkId, Criterion};
use once_cell::sync::Lazy;
use pageserver::config::PageServerConf;
use pageserver::walredo::PostgresRedoManager;
use pageserver::walredo::{PostgresRedoManager, RedoAttemptType};
use pageserver_api::key::Key;
use pageserver_api::record::NeonWalRecord;
use pageserver_api::shard::TenantShardId;
@@ -223,7 +223,14 @@ impl Request {
// TODO: avoid these clones
manager
.request_redo(*key, *lsn, base_img.clone(), records.clone(), *pg_version)
.request_redo(
*key,
*lsn,
base_img.clone(),
records.clone(),
*pg_version,
RedoAttemptType::ReadPage,
)
.await
.context("request_redo")
}

View File

@@ -419,6 +419,23 @@ impl Client {
}
}
pub async fn timeline_detail(
&self,
tenant_shard_id: TenantShardId,
timeline_id: TimelineId,
) -> Result<TimelineInfo> {
let uri = format!(
"{}/v1/tenant/{tenant_shard_id}/timeline/{timeline_id}",
self.mgmt_api_endpoint
);
self.request(Method::GET, &uri, ())
.await?
.json()
.await
.map_err(Error::ReceiveBody)
}
pub async fn timeline_archival_config(
&self,
tenant_shard_id: TenantShardId,

View File

@@ -68,6 +68,13 @@ pub(crate) struct Args {
targets: Option<Vec<TenantTimelineId>>,
}
/// State shared by all clients
#[derive(Debug)]
struct SharedState {
start_work_barrier: tokio::sync::Barrier,
live_stats: LiveStats,
}
#[derive(Debug, Default)]
struct LiveStats {
completed_requests: AtomicU64,
@@ -240,24 +247,26 @@ async fn main_impl(
all_ranges
};
let live_stats = Arc::new(LiveStats::default());
let num_live_stats_dump = 1;
let num_work_sender_tasks = args.num_clients.get() * timelines.len();
let num_main_impl = 1;
let start_work_barrier = Arc::new(tokio::sync::Barrier::new(
num_live_stats_dump + num_work_sender_tasks + num_main_impl,
));
let shared_state = Arc::new(SharedState {
start_work_barrier: tokio::sync::Barrier::new(
num_live_stats_dump + num_work_sender_tasks + num_main_impl,
),
live_stats: LiveStats::default(),
});
let cancel = CancellationToken::new();
let ss = shared_state.clone();
tokio::spawn({
let stats = Arc::clone(&live_stats);
let start_work_barrier = Arc::clone(&start_work_barrier);
async move {
start_work_barrier.wait().await;
ss.start_work_barrier.wait().await;
loop {
let start = std::time::Instant::now();
tokio::time::sleep(std::time::Duration::from_secs(1)).await;
let stats = &ss.live_stats;
let completed_requests = stats.completed_requests.swap(0, Ordering::Relaxed);
let missed = stats.missed.swap(0, Ordering::Relaxed);
let elapsed = start.elapsed();
@@ -270,14 +279,12 @@ async fn main_impl(
}
});
let cancel = CancellationToken::new();
let rps_period = args
.per_client_rate
.map(|rps_limit| Duration::from_secs_f64(1.0 / (rps_limit as f64)));
let make_worker: &dyn Fn(WorkerId) -> Pin<Box<dyn Send + Future<Output = ()>>> = &|worker_id| {
let live_stats = live_stats.clone();
let start_work_barrier = start_work_barrier.clone();
let ss = shared_state.clone();
let cancel = cancel.clone();
let ranges: Vec<KeyRange> = all_ranges
.iter()
.filter(|r| r.timeline == worker_id.timeline)
@@ -287,85 +294,8 @@ async fn main_impl(
rand::distributions::weighted::WeightedIndex::new(ranges.iter().map(|v| v.len()))
.unwrap();
let cancel = cancel.clone();
Box::pin(async move {
let client =
pageserver_client::page_service::Client::new(args.page_service_connstring.clone())
.await
.unwrap();
let mut client = client
.pagestream(worker_id.timeline.tenant_id, worker_id.timeline.timeline_id)
.await
.unwrap();
start_work_barrier.wait().await;
let client_start = Instant::now();
let mut ticks_processed = 0;
let mut inflight = VecDeque::new();
while !cancel.is_cancelled() {
// Detect if a request took longer than the RPS rate
if let Some(period) = &rps_period {
let periods_passed_until_now =
usize::try_from(client_start.elapsed().as_micros() / period.as_micros())
.unwrap();
if periods_passed_until_now > ticks_processed {
live_stats.missed((periods_passed_until_now - ticks_processed) as u64);
}
ticks_processed = periods_passed_until_now;
}
while inflight.len() < args.queue_depth.get() {
let start = Instant::now();
let req = {
let mut rng = rand::thread_rng();
let r = &ranges[weights.sample(&mut rng)];
let key: i128 = rng.gen_range(r.start..r.end);
let key = Key::from_i128(key);
assert!(key.is_rel_block_key());
let (rel_tag, block_no) = key
.to_rel_block()
.expect("we filter non-rel-block keys out above");
PagestreamGetPageRequest {
hdr: PagestreamRequest {
reqid: 0,
request_lsn: if rng.gen_bool(args.req_latest_probability) {
Lsn::MAX
} else {
r.timeline_lsn
},
not_modified_since: r.timeline_lsn,
},
rel: rel_tag,
blkno: block_no,
}
};
client.getpage_send(req).await.unwrap();
inflight.push_back(start);
}
let start = inflight.pop_front().unwrap();
client.getpage_recv().await.unwrap();
let end = Instant::now();
live_stats.request_done();
ticks_processed += 1;
STATS.with(|stats| {
stats
.borrow()
.lock()
.unwrap()
.observe(end.duration_since(start))
.unwrap();
});
if let Some(period) = &rps_period {
let next_at = client_start
+ Duration::from_micros(
(ticks_processed) as u64 * u64::try_from(period.as_micros()).unwrap(),
);
tokio::time::sleep_until(next_at.into()).await;
}
}
client_libpq(args, worker_id, ss, cancel, rps_period, ranges, weights).await
})
};
@@ -387,7 +317,7 @@ async fn main_impl(
};
info!("waiting for everything to become ready");
start_work_barrier.wait().await;
shared_state.start_work_barrier.wait().await;
info!("work started");
if let Some(runtime) = args.runtime {
tokio::time::sleep(runtime.into()).await;
@@ -416,3 +346,91 @@ async fn main_impl(
anyhow::Ok(())
}
async fn client_libpq(
args: &Args,
worker_id: WorkerId,
shared_state: Arc<SharedState>,
cancel: CancellationToken,
rps_period: Option<Duration>,
ranges: Vec<KeyRange>,
weights: rand::distributions::weighted::WeightedIndex<i128>,
) {
let client = pageserver_client::page_service::Client::new(args.page_service_connstring.clone())
.await
.unwrap();
let mut client = client
.pagestream(worker_id.timeline.tenant_id, worker_id.timeline.timeline_id)
.await
.unwrap();
shared_state.start_work_barrier.wait().await;
let client_start = Instant::now();
let mut ticks_processed = 0;
let mut inflight = VecDeque::new();
while !cancel.is_cancelled() {
// Detect if a request took longer than the RPS rate
if let Some(period) = &rps_period {
let periods_passed_until_now =
usize::try_from(client_start.elapsed().as_micros() / period.as_micros()).unwrap();
if periods_passed_until_now > ticks_processed {
shared_state
.live_stats
.missed((periods_passed_until_now - ticks_processed) as u64);
}
ticks_processed = periods_passed_until_now;
}
while inflight.len() < args.queue_depth.get() {
let start = Instant::now();
let req = {
let mut rng = rand::thread_rng();
let r = &ranges[weights.sample(&mut rng)];
let key: i128 = rng.gen_range(r.start..r.end);
let key = Key::from_i128(key);
assert!(key.is_rel_block_key());
let (rel_tag, block_no) = key
.to_rel_block()
.expect("we filter non-rel-block keys out above");
PagestreamGetPageRequest {
hdr: PagestreamRequest {
reqid: 0,
request_lsn: if rng.gen_bool(args.req_latest_probability) {
Lsn::MAX
} else {
r.timeline_lsn
},
not_modified_since: r.timeline_lsn,
},
rel: rel_tag,
blkno: block_no,
}
};
client.getpage_send(req).await.unwrap();
inflight.push_back(start);
}
let start = inflight.pop_front().unwrap();
client.getpage_recv().await.unwrap();
let end = Instant::now();
shared_state.live_stats.request_done();
ticks_processed += 1;
STATS.with(|stats| {
stats
.borrow()
.lock()
.unwrap()
.observe(end.duration_since(start))
.unwrap();
});
if let Some(period) = &rps_period {
let next_at = client_start
+ Duration::from_micros(
(ticks_processed) as u64 * u64::try_from(period.as_micros()).unwrap(),
);
tokio::time::sleep_until(next_at.into()).await;
}
}
}

View File

@@ -34,7 +34,7 @@ use utils::lsn::Lsn;
use crate::context::RequestContext;
use crate::pgdatadir_mapping::Version;
use crate::tenant::storage_layer::IoConcurrency;
use crate::tenant::timeline::GetVectoredError;
use crate::tenant::timeline::{GetVectoredError, VersionedKeySpaceQuery};
use crate::tenant::{PageReconstructError, Timeline};
#[derive(Debug, thiserror::Error)]
@@ -353,9 +353,10 @@ where
let mut slru_builder = SlruSegmentsBuilder::new(&mut self.ar);
for part in slru_partitions.parts {
let query = VersionedKeySpaceQuery::uniform(part, self.lsn);
let blocks = self
.timeline
.get_vectored(part, self.lsn, self.io_concurrency.clone(), self.ctx)
.get_vectored(query, self.io_concurrency.clone(), self.ctx)
.await?;
for (key, block) in blocks {

View File

@@ -416,8 +416,18 @@ fn start_pageserver(
// The storage_broker::connect call needs to happen inside a tokio runtime thread.
let broker_client = WALRECEIVER_RUNTIME
.block_on(async {
let tls_config = storage_broker::ClientTlsConfig::new().ca_certificates(
conf.ssl_ca_certs
.iter()
.map(pem::encode)
.map(storage_broker::Certificate::from_pem),
);
// Note: we do not attempt connecting here (but validate endpoints sanity).
storage_broker::connect(conf.broker_endpoint.clone(), conf.broker_keepalive_interval)
storage_broker::connect(
conf.broker_endpoint.clone(),
conf.broker_keepalive_interval,
tls_config,
)
})
.with_context(|| {
format!(
@@ -452,6 +462,24 @@ fn start_pageserver(
info!("Using auth for http API: {:#?}", conf.http_auth_type);
info!("Using auth for pg connections: {:#?}", conf.pg_auth_type);
let tls_server_config = if conf.listen_https_addr.is_some() || conf.enable_tls_page_service_api
{
let resolver = BACKGROUND_RUNTIME.block_on(ReloadingCertificateResolver::new(
"main",
&conf.ssl_key_file,
&conf.ssl_cert_file,
conf.ssl_cert_reload_period,
))?;
let server_config = rustls::ServerConfig::builder()
.with_no_client_auth()
.with_cert_resolver(resolver);
Some(Arc::new(server_config))
} else {
None
};
match var("NEON_AUTH_TOKEN") {
Ok(v) => {
info!("Loaded JWT token for authentication with Safekeeper");
@@ -670,17 +698,11 @@ fn start_pageserver(
let https_task = match https_listener {
Some(https_listener) => {
let resolver = MGMT_REQUEST_RUNTIME.block_on(ReloadingCertificateResolver::new(
&conf.ssl_key_file,
&conf.ssl_cert_file,
conf.ssl_cert_reload_period,
))?;
let tls_server_config = tls_server_config
.clone()
.expect("tls_server_config is set earlier if https is enabled");
let server_config = rustls::ServerConfig::builder()
.with_no_client_auth()
.with_cert_resolver(resolver);
let tls_acceptor = tokio_rustls::TlsAcceptor::from(Arc::new(server_config));
let tls_acceptor = tokio_rustls::TlsAcceptor::from(tls_server_config);
let server =
http_utils::server::Server::new(service, https_listener, Some(tls_acceptor))?;
@@ -736,6 +758,11 @@ fn start_pageserver(
tokio::net::TcpListener::from_std(pageserver_listener)
.context("create tokio listener")?
},
if conf.enable_tls_page_service_api {
tls_server_config
} else {
None
},
);
// All started up! Now just sit and wait for shutdown signal.

View File

@@ -17,9 +17,10 @@ use once_cell::sync::OnceCell;
use pageserver_api::config::{DiskUsageEvictionTaskConfig, MaxVectoredReadBytes};
use pageserver_api::models::ImageCompressionAlgorithm;
use pageserver_api::shard::TenantShardId;
use pem::Pem;
use postgres_backend::AuthType;
use remote_storage::{RemotePath, RemoteStorageConfig};
use reqwest::{Certificate, Url};
use reqwest::Url;
use storage_broker::Uri;
use utils::id::{NodeId, TimelineId};
use utils::logging::{LogFormat, SecretString};
@@ -67,8 +68,8 @@ pub struct PageServerConf {
/// Period to reload certificate and private key from files.
/// Default: 60s.
pub ssl_cert_reload_period: Duration,
/// Trusted root CA certificates to use in https APIs.
pub ssl_ca_certs: Vec<Certificate>,
/// Trusted root CA certificates to use in https APIs in PEM format.
pub ssl_ca_certs: Vec<Pem>,
/// Current availability zone. Used for traffic metrics.
pub availability_zone: Option<String>,
@@ -118,13 +119,13 @@ pub struct PageServerConf {
/// A lower value implicitly deprioritizes loading such tenants, vs. other work in the system.
pub concurrent_tenant_warmup: ConfigurableSemaphore,
/// Number of concurrent [`Tenant::gather_size_inputs`](crate::tenant::Tenant::gather_size_inputs) allowed.
/// Number of concurrent [`TenantShard::gather_size_inputs`](crate::tenant::TenantShard::gather_size_inputs) allowed.
pub concurrent_tenant_size_logical_size_queries: ConfigurableSemaphore,
/// Limit of concurrent [`Tenant::gather_size_inputs`] issued by module `eviction_task`.
/// Limit of concurrent [`TenantShard::gather_size_inputs`] issued by module `eviction_task`.
/// The number of permits is the same as `concurrent_tenant_size_logical_size_queries`.
/// See the comment in `eviction_task` for details.
///
/// [`Tenant::gather_size_inputs`]: crate::tenant::Tenant::gather_size_inputs
/// [`TenantShard::gather_size_inputs`]: crate::tenant::TenantShard::gather_size_inputs
pub eviction_task_immitated_concurrent_logical_size_queries: ConfigurableSemaphore,
// How often to collect metrics and send them to the metrics endpoint.
@@ -219,6 +220,16 @@ pub struct PageServerConf {
pub generate_unarchival_heatmap: bool,
pub tracing: Option<pageserver_api::config::Tracing>,
/// Enable TLS in page service API.
/// Does not force TLS: the client negotiates TLS usage during the handshake.
/// Uses key and certificate from ssl_key_file/ssl_cert_file.
pub enable_tls_page_service_api: bool,
/// Run in development mode, which disables certain safety checks
/// such as authentication requirements for HTTP and PostgreSQL APIs.
/// This is insecure and should only be used in development environments.
pub dev_mode: bool,
}
/// Token for authentication to safekeepers
@@ -391,6 +402,8 @@ impl PageServerConf {
load_previous_heatmap,
generate_unarchival_heatmap,
tracing,
enable_tls_page_service_api,
dev_mode,
} = config_toml;
let mut conf = PageServerConf {
@@ -441,6 +454,8 @@ impl PageServerConf {
page_service_pipelining,
get_vectored_concurrent_io,
tracing,
enable_tls_page_service_api,
dev_mode,
// ------------------------------------------------------------
// fields that require additional validation or custom handling
@@ -490,7 +505,10 @@ impl PageServerConf {
ssl_ca_certs: match ssl_ca_file {
Some(ssl_ca_file) => {
let buf = std::fs::read(ssl_ca_file)?;
Certificate::from_pem_bundle(&buf)?
pem::parse_many(&buf)?
.into_iter()
.filter(|pem| pem.tag() == "CERTIFICATE")
.collect()
}
None => Vec::new(),
},
@@ -581,10 +599,10 @@ impl ConfigurableSemaphore {
/// Initializse using a non-zero amount of permits.
///
/// Require a non-zero initial permits, because using permits == 0 is a crude way to disable a
/// feature such as [`Tenant::gather_size_inputs`]. Otherwise any semaphore using future will
/// feature such as [`TenantShard::gather_size_inputs`]. Otherwise any semaphore using future will
/// behave like [`futures::future::pending`], just waiting until new permits are added.
///
/// [`Tenant::gather_size_inputs`]: crate::tenant::Tenant::gather_size_inputs
/// [`TenantShard::gather_size_inputs`]: crate::tenant::TenantShard::gather_size_inputs
pub fn new(initial_permits: NonZeroUsize) -> Self {
ConfigurableSemaphore {
initial_permits,

View File

@@ -24,7 +24,7 @@ use crate::task_mgr::{self, BACKGROUND_RUNTIME, TaskKind};
use crate::tenant::mgr::TenantManager;
use crate::tenant::size::CalculateSyntheticSizeError;
use crate::tenant::tasks::BackgroundLoopKind;
use crate::tenant::{LogicalSizeCalculationCause, Tenant};
use crate::tenant::{LogicalSizeCalculationCause, TenantShard};
mod disk_cache;
mod metrics;
@@ -428,7 +428,7 @@ async fn calculate_synthetic_size_worker(
}
}
async fn calculate_and_log(tenant: &Tenant, cancel: &CancellationToken, ctx: &RequestContext) {
async fn calculate_and_log(tenant: &TenantShard, cancel: &CancellationToken, ctx: &RequestContext) {
const CAUSE: LogicalSizeCalculationCause =
LogicalSizeCalculationCause::ConsumptionMetricsSyntheticSize;

View File

@@ -175,9 +175,9 @@ impl MetricsKey {
.absolute_values()
}
/// [`Tenant::remote_size`]
/// [`TenantShard::remote_size`]
///
/// [`Tenant::remote_size`]: crate::tenant::Tenant::remote_size
/// [`TenantShard::remote_size`]: crate::tenant::TenantShard::remote_size
const fn remote_storage_size(tenant_id: TenantId) -> AbsoluteValueFactory {
MetricsKey {
tenant_id,
@@ -199,9 +199,9 @@ impl MetricsKey {
.absolute_values()
}
/// [`Tenant::cached_synthetic_size`] as refreshed by [`calculate_synthetic_size_worker`].
/// [`TenantShard::cached_synthetic_size`] as refreshed by [`calculate_synthetic_size_worker`].
///
/// [`Tenant::cached_synthetic_size`]: crate::tenant::Tenant::cached_synthetic_size
/// [`TenantShard::cached_synthetic_size`]: crate::tenant::TenantShard::cached_synthetic_size
/// [`calculate_synthetic_size_worker`]: super::calculate_synthetic_size_worker
const fn synthetic_size(tenant_id: TenantId) -> AbsoluteValueFactory {
MetricsKey {
@@ -254,7 +254,7 @@ pub(super) async fn collect_all_metrics(
async fn collect<S>(tenants: S, cache: &Cache, ctx: &RequestContext) -> Vec<NewRawMetric>
where
S: futures::stream::Stream<Item = (TenantId, Arc<crate::tenant::Tenant>)>,
S: futures::stream::Stream<Item = (TenantId, Arc<crate::tenant::TenantShard>)>,
{
let mut current_metrics: Vec<NewRawMetric> = Vec::new();
@@ -263,7 +263,9 @@ where
while let Some((tenant_id, tenant)) = tenants.next().await {
let mut tenant_resident_size = 0;
for timeline in tenant.list_timelines() {
let timelines = tenant.list_timelines();
let timelines_len = timelines.len();
for timeline in timelines {
let timeline_id = timeline.timeline_id;
match TimelineSnapshot::collect(&timeline, ctx) {
@@ -289,6 +291,11 @@ where
tenant_resident_size += timeline.resident_physical_size();
}
if timelines_len == 0 {
// Force set it to 1 byte to avoid not being reported -- all timelines are offloaded.
tenant_resident_size = 1;
}
let snap = TenantSnapshot::collect(&tenant, tenant_resident_size);
snap.to_metrics(tenant_id, Utc::now(), cache, &mut current_metrics);
}
@@ -308,7 +315,7 @@ impl TenantSnapshot {
///
/// `resident_size` is calculated of the timelines we had access to for other metrics, so we
/// cannot just list timelines here.
fn collect(t: &Arc<crate::tenant::Tenant>, resident_size: u64) -> Self {
fn collect(t: &Arc<crate::tenant::TenantShard>, resident_size: u64) -> Self {
TenantSnapshot {
resident_size,
remote_size: t.remote_size(),

View File

@@ -3,17 +3,19 @@ use std::collections::HashMap;
use futures::Future;
use pageserver_api::config::NodeMetadata;
use pageserver_api::controller_api::{AvailabilityZone, NodeRegisterRequest};
use pageserver_api::models::ShardImportStatus;
use pageserver_api::shard::TenantShardId;
use pageserver_api::upcall_api::{
ReAttachRequest, ReAttachResponse, ReAttachResponseTenant, ValidateRequest,
ValidateRequestTenant, ValidateResponse,
PutTimelineImportStatusRequest, ReAttachRequest, ReAttachResponse, ReAttachResponseTenant,
ValidateRequest, ValidateRequestTenant, ValidateResponse,
};
use reqwest::Certificate;
use serde::Serialize;
use serde::de::DeserializeOwned;
use tokio_util::sync::CancellationToken;
use url::Url;
use utils::generation::Generation;
use utils::id::NodeId;
use utils::id::{NodeId, TimelineId};
use utils::{backoff, failpoint_support};
use crate::config::PageServerConf;
@@ -45,6 +47,12 @@ pub trait StorageControllerUpcallApi {
&self,
tenants: Vec<(TenantShardId, Generation)>,
) -> impl Future<Output = Result<HashMap<TenantShardId, bool>, RetryForeverError>> + Send;
fn put_timeline_import_status(
&self,
tenant_shard_id: TenantShardId,
timeline_id: TimelineId,
status: ShardImportStatus,
) -> impl Future<Output = Result<(), RetryForeverError>> + Send;
}
impl StorageControllerUpcallClient {
@@ -76,8 +84,8 @@ impl StorageControllerUpcallClient {
client = client.default_headers(headers);
}
for ssl_ca_cert in &conf.ssl_ca_certs {
client = client.add_root_certificate(ssl_ca_cert.clone());
for cert in &conf.ssl_ca_certs {
client = client.add_root_certificate(Certificate::from_der(cert.contents())?);
}
Ok(Some(Self {
@@ -272,4 +280,30 @@ impl StorageControllerUpcallApi for StorageControllerUpcallClient {
Ok(result.into_iter().collect())
}
/// Send a shard import status to the storage controller
///
/// The implementation must have at-least-once delivery semantics.
/// To this end, we retry the request until it succeeds. If the pageserver
/// restarts or crashes, the shard import will start again from the beggining.
#[tracing::instrument(skip_all)] // so that warning logs from retry_http_forever have context
async fn put_timeline_import_status(
&self,
tenant_shard_id: TenantShardId,
timeline_id: TimelineId,
status: ShardImportStatus,
) -> Result<(), RetryForeverError> {
let url = self
.base_url
.join("timeline_import_status")
.expect("Failed to build path");
let request = PutTimelineImportStatusRequest {
tenant_shard_id,
timeline_id,
status,
};
self.retry_http_forever(&url, request).await
}
}

View File

@@ -787,6 +787,15 @@ mod test {
Ok(result)
}
async fn put_timeline_import_status(
&self,
_tenant_shard_id: TenantShardId,
_timeline_id: TimelineId,
_status: pageserver_api::models::ShardImportStatus,
) -> Result<(), RetryForeverError> {
unimplemented!()
}
}
async fn setup(test_name: &str) -> anyhow::Result<TestSetup> {

View File

@@ -1873,7 +1873,7 @@ async fn update_tenant_config_handler(
&ShardParameters::default(),
);
crate::tenant::Tenant::persist_tenant_config(state.conf, &tenant_shard_id, &location_conf)
crate::tenant::TenantShard::persist_tenant_config(state.conf, &tenant_shard_id, &location_conf)
.await
.map_err(|e| ApiError::InternalServerError(anyhow::anyhow!(e)))?;
@@ -1917,7 +1917,7 @@ async fn patch_tenant_config_handler(
&ShardParameters::default(),
);
crate::tenant::Tenant::persist_tenant_config(state.conf, &tenant_shard_id, &location_conf)
crate::tenant::TenantShard::persist_tenant_config(state.conf, &tenant_shard_id, &location_conf)
.await
.map_err(|e| ApiError::InternalServerError(anyhow::anyhow!(e)))?;
@@ -2274,6 +2274,7 @@ async fn timeline_compact_handler(
if Some(true) == parse_query_param::<_, bool>(&request, "dry_run")? {
flags |= CompactFlags::DryRun;
}
// Manual compaction does not yield for L0.
let wait_until_uploaded =
parse_query_param::<_, bool>(&request, "wait_until_uploaded")?.unwrap_or(false);
@@ -3252,7 +3253,7 @@ async fn ingest_aux_files(
modification
.put_file(&fname, content.as_bytes(), &ctx)
.await
.map_err(ApiError::InternalServerError)?;
.map_err(|e| ApiError::InternalServerError(e.into()))?;
}
modification
.commit(&ctx)

View File

@@ -27,7 +27,7 @@ use crate::context::RequestContext;
use crate::metrics::WAL_INGEST;
use crate::pgdatadir_mapping::*;
use crate::tenant::Timeline;
use crate::walingest::WalIngest;
use crate::walingest::{WalIngest, WalIngestErrorKind};
// Returns checkpoint LSN from controlfile
pub fn get_lsn_from_controlfile(path: &Utf8Path) -> Result<Lsn> {
@@ -157,9 +157,9 @@ async fn import_rel(
.put_rel_creation(rel, nblocks as u32, ctx)
.await
{
match e {
RelationError::AlreadyExists => {
debug!("Relation {} already exist. We must be extending it.", rel)
match e.kind {
WalIngestErrorKind::RelationAlreadyExists(rel) => {
debug!("Relation {rel} already exists. We must be extending it.")
}
_ => return Err(e.into()),
}

View File

@@ -49,7 +49,7 @@ use tracing::{info, info_span};
/// backwards-compatible changes to the metadata format.
pub const STORAGE_FORMAT_VERSION: u16 = 3;
pub const DEFAULT_PG_VERSION: u32 = 16;
pub const DEFAULT_PG_VERSION: u32 = 17;
// Magic constants used to identify different kinds of files
pub const IMAGE_FILE_MAGIC: u16 = 0x5A60;

Some files were not shown because too many files have changed in this diff Show More