Skip to content
This repository has been archived by the owner on Apr 26, 2024. It is now read-only.

Improve performance of _get_state_groups_from_groups_txn #7567

Conversation

ilmari
Copy link
Contributor

@ilmari ilmari commented May 23, 2020

The query keeps showing up in my slow query log.

This changes the plan under the top-level Sort node from

WindowAgg  (cost=280335.88..292963.15 rows=561212 width=80) (actual time=138.651..160.562 rows=27112 loops=1)
  ->  Sort  (cost=280335.88..281738.91 rows=561212 width=84) (actual time=138.597..140.622 rows=27112 loops=1)
        Sort Key: state_groups_state.type, state_groups_state.state_key, state_groups_state.state_group
        Sort Method: quicksort  Memory: 4581kB
        ->  Nested Loop  (cost=2.83..226745.22 rows=561212 width=84) (actual time=21.548..47.657 rows=27112 loops=1)
              ->  HashAggregate  (cost=2.27..3.28 rows=101 width=8) (actual time=21.526..21.535 rows=20 loops=1)
                    Group Key: state.state_group
                    ->  CTE Scan on state  (cost=0.00..2.02 rows=101 width=8) (actual time=21.280..21.493 rows=20 loops=1)
              ->  Index Scan using state_groups_state_type_idx on state_groups_state  (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.005..0.991 rows=1356 loops=20)
                    Index Cond: (state_group = state.state_group)

to

Nested Loop  (cost=2.83..226745.22 rows=561212 width=84) (actual time=24.194..52.834 rows=27112 loops=1)
  ->  HashAggregate  (cost=2.27..3.28 rows=101 width=8) (actual time=24.130..24.138 rows=20 loops=1)
        Group Key: state.state_group
        ->  CTE Scan on state  (cost=0.00..2.02 rows=101 width=8) (actual time=23.887..24.113 rows=20 loops=1)
  ->  Index Scan using state_groups_state_type_idx on state_groups_state  (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.016..1.159 rows=1356 loops=20)
        Index Cond: (state_group = state.state_group)

This cuts the execution time from ~190ms to ~130ms, i.e. a reduction
of ~30%.

The full plans are visualised at https://explain.depesz.com/s/WpbT and
https://explain.depesz.com/s/KlEk

Pull Request Checklist

  • Pull request is based on the develop branch
  • Pull request includes a changelog file. The entry should:
    • Be a short description of your change which makes sense to users. "Fixed a bug that prevented receiving messages from other servers." instead of "Moved X method from EventStore to EventWorkerStore.".
    • Use markdown where necessary, mostly for code blocks.
    • End with either a period (.) or an exclamation mark (!).
    • Start with a capital letter.
  • Pull request includes a sign off
  • Code style is correct (run the linters)

@ilmari ilmari force-pushed the improve-get-state-groups-from-groups-sql branch from d600189 to b53790f Compare May 23, 2020 23:06
@ilmari ilmari force-pushed the improve-get-state-groups-from-groups-sql branch from b53790f to 914be9f Compare May 23, 2020 23:35
@anoadragon453 anoadragon453 requested a review from a team May 24, 2020 20:23
@anoadragon453
Copy link
Member

Since this is failing on all worker tests but not monolith tests, I have a horrible feeling that SyTest was relying on this SQL to be slow.

I know @erikjohnston did some race fixes recently. I'll try rebasing this PR to see if CI fares any better a second time.

@anoadragon453 anoadragon453 force-pushed the improve-get-state-groups-from-groups-sql branch from 914be9f to 6af29d4 Compare May 27, 2020 15:49
@anoadragon453 anoadragon453 removed the request for review from a team May 27, 2020 15:50
@ilmari ilmari force-pushed the improve-get-state-groups-from-groups-sql branch from 6af29d4 to 277ff60 Compare May 30, 2020 21:48
…7567)

The query keeps showing up in my slow query log.

This changes the plan under the top-level Sort node from

    WindowAgg  (cost=280335.88..292963.15 rows=561212 width=80) (actual time=138.651..160.562 rows=27112 loops=1)
      ->  Sort  (cost=280335.88..281738.91 rows=561212 width=84) (actual time=138.597..140.622 rows=27112 loops=1)
            Sort Key: state_groups_state.type, state_groups_state.state_key, state_groups_state.state_group
            Sort Method: quicksort  Memory: 4581kB
            ->  Nested Loop  (cost=2.83..226745.22 rows=561212 width=84) (actual time=21.548..47.657 rows=27112 loops=1)
                  ->  HashAggregate  (cost=2.27..3.28 rows=101 width=8) (actual time=21.526..21.535 rows=20 loops=1)
                        Group Key: state.state_group
                        ->  CTE Scan on state  (cost=0.00..2.02 rows=101 width=8) (actual time=21.280..21.493 rows=20 loops=1)
                  ->  Index Scan using state_groups_state_type_idx on state_groups_state  (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.005..0.991 rows=1356 loops=20)
                        Index Cond: (state_group = state.state_group)

to

    Nested Loop  (cost=2.83..226745.22 rows=561212 width=84) (actual time=24.194..52.834 rows=27112 loops=1)
      ->  HashAggregate  (cost=2.27..3.28 rows=101 width=8) (actual time=24.130..24.138 rows=20 loops=1)
            Group Key: state.state_group
            ->  CTE Scan on state  (cost=0.00..2.02 rows=101 width=8) (actual time=23.887..24.113 rows=20 loops=1)
      ->  Index Scan using state_groups_state_type_idx on state_groups_state  (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.016..1.159 rows=1356 loops=20)
            Index Cond: (state_group = state.state_group)

This cuts the execution time from ~190ms to ~130ms, i.e. a reduction
of ~30%.

The full plans are visualised at https://explain.depesz.com/s/WpbT and
https://explain.depesz.com/s/KlEk

Signed-off-by: Dagfinn Ilmari Mannsåker <[email protected]>
@ilmari ilmari force-pushed the improve-get-state-groups-from-groups-sql branch from 277ff60 to f16584c Compare May 30, 2020 23:27
@ilmari
Copy link
Contributor Author

ilmari commented May 31, 2020

@anoadragon453 Turns out I messed up when adapting the query from my experimentation to the code, and was sorting the state_groups_state rows by event_id, not by state_group.

I've fixed it in the last force push

@clokep clokep requested a review from a team June 1, 2020 11:31
@erikjohnston
Copy link
Member

Oooh, interesting. I didn't realise DISTINCT ON had a defined ordering for postgres, the relevant part of the documentation explaining that is here.

changelog.d/7567.misc Outdated Show resolved Hide resolved
Copy link
Member

@erikjohnston erikjohnston left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I do indeed believe this works! Awesome, thank you!

@erikjohnston erikjohnston merged commit df8a3ce into matrix-org:develop Jun 1, 2020
@ilmari ilmari deleted the improve-get-state-groups-from-groups-sql branch June 1, 2020 14:52
babolivier added a commit that referenced this pull request Jun 10, 2020
…rg-hotfixes

Synapse 1.15.0rc1 (2020-06-09)
==============================

Features
--------

- Advertise support for Client-Server API r0.6.0 and remove related unstable feature flags. ([\#6585](#6585))
- Add an option to disable autojoining rooms for guest accounts. ([\#6637](#6637))
- For SAML authentication, add the ability to pass email addresses to be added to new users' accounts via SAML attributes. Contributed by Christopher Cooper. ([\#7385](#7385))
- Add admin APIs to allow server admins to manage users' devices. Contributed by @dklimpel. ([\#7481](#7481))
- Add support for generating thumbnails for WebP images. Previously, users would see an empty box instead of preview image. ([\#7586](#7586))
- Support the standardized `m.login.sso` user-interactive authentication flow. ([\#7630](#7630))

Bugfixes
--------

- Allow new users to be registered via the admin API even if the monthly active user limit has been reached. Contributed by @dkimpel. ([\#7263](#7263))
- Fix email notifications not being enabled for new users when created via the Admin API. ([\#7267](#7267))
- Fix str placeholders in an instance of `PrepareDatabaseException`. Introduced in Synapse v1.8.0. ([\#7575](#7575))
- Fix a bug in automatic user creation during first time login with `m.login.jwt`. Regression in v1.6.0. Contributed by @olof. ([\#7585](#7585))
- Fix a bug causing the cross-signing keys to be ignored when resyncing a device list. ([\#7594](#7594))
- Fix metrics failing when there is a large number of active background processes. ([\#7597](#7597))
- Fix bug where returning rooms for a group would fail if it included a room that the server was not in. ([\#7599](#7599))
- Fix duplicate key violation when persisting read markers. ([\#7607](#7607))
- Prevent an entire iteration of the device list resync loop from failing if one server responds with a malformed result. ([\#7609](#7609))
- Fix exceptions when fetching events from a remote host fails. ([\#7622](#7622))
- Make `synctl restart` start synapse if it wasn't running. ([\#7624](#7624))
- Pass device information through to the login endpoint when using the login fallback. ([\#7629](#7629))
- Advertise the `m.login.token` login flow when OpenID Connect is enabled. ([\#7631](#7631))
- Fix bug in account data replication stream. ([\#7656](#7656))

Improved Documentation
----------------------

- Update the OpenBSD installation instructions. ([\#7587](#7587))
- Advertise Python 3.8 support in `setup.py`. ([\#7602](#7602))
- Add a link to `#synapse:matrix.org` in the troubleshooting section of the README. ([\#7603](#7603))
- Clarifications to the admin api documentation. ([\#7647](#7647))

Internal Changes
----------------

- Convert the identity handler to async/await. ([\#7561](#7561))
- Improve query performance for fetching state from a PostgreSQL database. ([\#7567](#7567))
- Speed up processing of federation stream RDATA rows. ([\#7584](#7584))
- Add comment to systemd example to show postgresql dependency. ([\#7591](#7591))
- Refactor `Ratelimiter` to limit the amount of expensive config value accesses. ([\#7595](#7595))
- Convert groups handlers to async/await. ([\#7600](#7600))
- Clean up exception handling in `SAML2ResponseResource`. ([\#7614](#7614))
- Check that all asynchronous tasks succeed and general cleanup of `MonthlyActiveUsersTestCase` and `TestMauLimit`. ([\#7619](#7619))
- Convert `get_user_id_by_threepid` to async/await. ([\#7620](#7620))
- Switch to upstream `dh-virtualenv` rather than our fork for Debian package builds. ([\#7621](#7621))
- Update CI scripts to check the number in the newsfile fragment. ([\#7623](#7623))
- Check if the localpart of a Matrix ID is reserved for guest users earlier in the registration flow, as well as when responding to requests to `/register/available`. ([\#7625](#7625))
- Minor cleanups to OpenID Connect integration. ([\#7628](#7628))
- Attempt to fix flaky test: `PhoneHomeStatsTestCase.test_performance_100`. ([\#7634](#7634))
- Fix typos of `m.olm.curve25519-aes-sha2` and `m.megolm.v1.aes-sha2` in comments, test files. ([\#7637](#7637))
- Convert user directory, state deltas, and stats handlers to async/await. ([\#7640](#7640))
- Remove some unused constants. ([\#7644](#7644))
- Fix type information on `assert_*_is_admin` methods. ([\#7645](#7645))
- Convert registration handler to async/await. ([\#7649](#7649))
phil-flex pushed a commit to phil-flex/synapse that referenced this pull request Jun 16, 2020
…7567)

The query keeps showing up in my slow query log.

This changes the plan under the top-level Sort node from

```
    WindowAgg  (cost=280335.88..292963.15 rows=561212 width=80) (actual time=138.651..160.562 rows=27112 loops=1)
      ->  Sort  (cost=280335.88..281738.91 rows=561212 width=84) (actual time=138.597..140.622 rows=27112 loops=1)
            Sort Key: state_groups_state.type, state_groups_state.state_key, state_groups_state.state_group
            Sort Method: quicksort  Memory: 4581kB
            ->  Nested Loop  (cost=2.83..226745.22 rows=561212 width=84) (actual time=21.548..47.657 rows=27112 loops=1)
                  ->  HashAggregate  (cost=2.27..3.28 rows=101 width=8) (actual time=21.526..21.535 rows=20 loops=1)
                        Group Key: state.state_group
                        ->  CTE Scan on state  (cost=0.00..2.02 rows=101 width=8) (actual time=21.280..21.493 rows=20 loops=1)
                  ->  Index Scan using state_groups_state_type_idx on state_groups_state  (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.005..0.991 rows=1356 loops=20)
                        Index Cond: (state_group = state.state_group)
```

to

```
    Nested Loop  (cost=2.83..226745.22 rows=561212 width=84) (actual time=24.194..52.834 rows=27112 loops=1)
      ->  HashAggregate  (cost=2.27..3.28 rows=101 width=8) (actual time=24.130..24.138 rows=20 loops=1)
            Group Key: state.state_group
            ->  CTE Scan on state  (cost=0.00..2.02 rows=101 width=8) (actual time=23.887..24.113 rows=20 loops=1)
      ->  Index Scan using state_groups_state_type_idx on state_groups_state  (cost=0.56..2189.40 rows=5557 width=84) (actual time=0.016..1.159 rows=1356 loops=20)
            Index Cond: (state_group = state.state_group)
```

This cuts the execution time from ~190ms to ~130ms, i.e. a reduction
of ~30%.

The full plans are visualised at https://explain.depesz.com/s/WpbT and
https://explain.depesz.com/s/KlEk

Signed-off-by: Dagfinn Ilmari Mannsåker <[email protected]>
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants