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This repository has been archived by the owner on Apr 26, 2024.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

Fix caching devices for remote servers in worker. #6332

Merged
merged 3 commits into from
Nov 26, 2019
Merged

Fix caching devices for remote servers in worker. #6332

merged 3 commits into from
Nov 26, 2019

Conversation

erikjohnston
Copy link
Member

When the /keys/query API is hit on client_reader worker Synapse may
decide that it needs to resync some remote deivces. Usually this happens
on master, and then gets cached. However, that fails on workers and so
it falls back to fetching devices from remotes directly, which may in
turn fail if the remote is down.

This was picked up as part of matrix-org/sytest#732

@erikjohnston
Fix caching devices for remote servers in worker.
c16e192 
When the `/keys/query` API is hit on client_reader worker Synapse may
decide that it needs to resync some remote deivces. Usually this happens
on master, and then gets cached. However, that fails on workers and so
it falls back to fetching devices from remotes directly, which may in
turn fail if the remote is down.
@erikjohnston
Newsfile
248111b
@erikjohnston erikjohnston requested a review from a team November 5, 2019 16:00
richvdh
richvdh approved these changes Nov 26, 2019
View reviewed changes
Copy link
Member

@richvdh richvdh left a comment

Choose a reason for hiding this comment

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

lgtm otherwise

synapse/replication/http/devices.py Outdated


class ReplicationUserDevicesResyncRestServlet(ReplicationEndpoint):
"""Notifies that a user has joined or left the room
Copy link
Member

Choose a reason for hiding this comment

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

it doesn't seem to be about a notification? It seems to be a request to fetch the user's devices from the remote server?

We should probably also record here (or at least somewhere in the code comments, rather than just the PR where it will get lost) why it is important for this to happen on the master.

Copy link
Member Author

Choose a reason for hiding this comment

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

Woops, failed to delete the docstring when c+p'ed and then was like "yay i've already written a docstring" 🤦‍♂️

richvdh
richvdh reviewed Nov 26, 2019
View reviewed changes
changelog.d/6332.bugfix Outdated
@@ -0,0 +1 @@
Fix caching devices for remote users when using workers.
Copy link
Member

Choose a reason for hiding this comment

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

this is a bit cryptic. Perhaps you can explain what the user-visible symptoms were?

@erikjohnston erikjohnston merged commit f9f1c8a into develop Nov 26, 2019
@erikjohnston erikjohnston mentioned this pull request Nov 28, 2019
Closed
neilisfragile added a commit that referenced this pull request Dec 9, 2019
@neilisfragile
Merge tag 'v1.7.0rc1' into develop
0f3614f 
Synapse 1.7.0rc1 (2019-12-09)
=============================

Features
--------

- Implement per-room message retention policies. ([\#5815](#5815), [\#6436](#6436))
- Add etag and count fields to key backup endpoints to help clients guess if there are new keys. ([\#5858](#5858))
- Add `/admin/v2/users` endpoint with pagination. Contributed by Awesome Technologies Innovationslabor GmbH. ([\#5925](#5925))
- Require User-Interactive Authentication for `/account/3pid/add`, meaning the user's password will be required to add a third-party ID to their account. ([\#6119](#6119))
- Implement the `/_matrix/federation/unstable/net.atleastfornow/state/<context>` API as drafted in MSC2314. ([\#6176](#6176))
- Configure privacy-preserving settings by default for the room directory. ([\#6354](#6354))
- Add ephemeral messages support by partially implementing [MSC2228](matrix-org/matrix-spec-proposals#2228). ([\#6409](#6409))
- Add support for [MSC 2367](matrix-org/matrix-spec-proposals#2367), which allows specifying a reason on all membership events. ([\#6434](#6434))

Bugfixes
--------

- Transfer non-standard power levels on room upgrade. ([\#6237](#6237))
- Fix error from the Pillow library when uploading RGBA images. ([\#6241](#6241))
- Correctly apply the event filter to the `state`, `events_before` and `events_after` fields in the response to `/context` requests. ([\#6329](#6329))
- Fix caching devices for remote users when using workers, so that we don't attempt to refetch (and potentially fail) each time a user requests devices. ([\#6332](#6332))
- Prevent account data syncs getting lost across TCP replication. ([\#6333](#6333))
- Fix bug: TypeError in `register_user()` while using LDAP auth module. ([\#6406](#6406))
- Fix an intermittent exception when handling read-receipts. ([\#6408](#6408))
- Fix broken guest registration when there are existing blocks of numeric user IDs. ([\#6420](#6420))
- Fix startup error when http proxy is defined. ([\#6421](#6421))
- Fix error when using synapse_port_db on a vanilla synapse db. ([\#6449](#6449))
- Fix uploading multiple cross signing signatures for the same user. ([\#6451](#6451))
- Fix bug which lead to exceptions being thrown in a loop when a cross-signed device is deleted. ([\#6462](#6462))
- Fix `synapse_port_db` not exiting with a 0 code if something went wrong during the port process. ([\#6470](#6470))
- Improve sanity-checking when receiving events over federation. ([\#6472](#6472))
- Fix inaccurate per-block Prometheus metrics. ([\#6491](#6491))
- Fix small performance regression for sending invites. ([\#6493](#6493))
- Back out cross-signing code added in Synapse 1.5.0, which caused a performance regression. ([\#6494](#6494))

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

- Update documentation and variables in user contributed systemd reference file. ([\#6369](#6369), [\#6490](#6490))
- Fix link in the user directory documentation. ([\#6388](#6388))
- Add build instructions to the docker readme. ([\#6390](#6390))
- Switch Ubuntu package install recommendation to use python3 packages in INSTALL.md. ([\#6443](#6443))
- Write some docs for the quarantine_media api. ([\#6458](#6458))
- Convert CONTRIBUTING.rst to markdown (among other small fixes). ([\#6461](#6461))

Deprecations and Removals
-------------------------

- Remove admin/v1/users_paginate endpoint. Contributed by Awesome Technologies Innovationslabor GmbH. ([\#5925](#5925))
- Remove fallback for federation with old servers which lack the /federation/v1/state_ids API. ([\#6488](#6488))

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

- Add benchmarks for structured logging and improve output performance. ([\#6266](#6266))
- Improve the performance of outputting structured logging. ([\#6322](#6322))
- Refactor some code in the event authentication path for clarity. ([\#6343](#6343), [\#6468](#6468), [\#6480](#6480))
- Clean up some unnecessary quotation marks around the codebase. ([\#6362](#6362))
- Complain on startup instead of 500'ing during runtime when `public_baseurl` isn't set when necessary. ([\#6379](#6379))
- Add a test scenario to make sure room history purges don't break `/messages` in the future. ([\#6392](#6392))
- Clarifications for the email configuration settings. ([\#6423](#6423))
- Add more tests to the blacklist when running in worker mode. ([\#6429](#6429))
- Refactor data store layer to support multiple databases in the future. ([\#6454](#6454), [\#6464](#6464), [\#6469](#6469), [\#6487](#6487))
- Port synapse.rest.client.v1 to async/await. ([\#6482](#6482))
- Port synapse.rest.client.v2_alpha to async/await. ([\#6483](#6483))
- Port SyncHandler to async/await. ([\#6484](#6484))
erikjohnston added a commit that referenced this pull request Dec 13, 2019
@erikjohnston
Merge tag 'v1.7.0'
bee1982 
Synapse 1.7.0 (2019-12-13)
==========================

This release changes the default settings so that only local authenticated users can query the server's room directory. See the [upgrade notes](UPGRADE.rst#upgrading-to-v170) for details.

Support for SQLite versions before 3.11 is now deprecated. A future release will refuse to start if used with an SQLite version before 3.11.

Administrators are reminded that SQLite should not be used for production instances. Instructions for migrating to Postgres are available [here](docs/postgres.md). A future release of synapse will, by default, disable federation for servers using SQLite.

No significant changes since 1.7.0rc2.

Synapse 1.7.0rc2 (2019-12-11)
=============================

Bugfixes
--------

- Fix incorrect error message for invalid requests when setting user's avatar URL. ([\#6497](#6497))
- Fix support for SQLite 3.7. ([\#6499](#6499))
- Fix regression where sending email push would not work when using a pusher worker. ([\#6507](#6507), [\#6509](#6509))

Synapse 1.7.0rc1 (2019-12-09)
=============================

Features
--------

- Implement per-room message retention policies. ([\#5815](#5815), [\#6436](#6436))
- Add etag and count fields to key backup endpoints to help clients guess if there are new keys. ([\#5858](#5858))
- Add `/admin/v2/users` endpoint with pagination. Contributed by Awesome Technologies Innovationslabor GmbH. ([\#5925](#5925))
- Require User-Interactive Authentication for `/account/3pid/add`, meaning the user's password will be required to add a third-party ID to their account. ([\#6119](#6119))
- Implement the `/_matrix/federation/unstable/net.atleastfornow/state/<context>` API as drafted in MSC2314. ([\#6176](#6176))
- Configure privacy-preserving settings by default for the room directory. ([\#6355](#6355))
- Add ephemeral messages support by partially implementing [MSC2228](matrix-org/matrix-spec-proposals#2228). ([\#6409](#6409))
- Add support for [MSC 2367](matrix-org/matrix-spec-proposals#2367), which allows specifying a reason on all membership events. ([\#6434](#6434))

Bugfixes
--------

- Transfer non-standard power levels on room upgrade. ([\#6237](#6237))
- Fix error from the Pillow library when uploading RGBA images. ([\#6241](#6241))
- Correctly apply the event filter to the `state`, `events_before` and `events_after` fields in the response to `/context` requests. ([\#6329](#6329))
- Fix caching devices for remote users when using workers, so that we don't attempt to refetch (and potentially fail) each time a user requests devices. ([\#6332](#6332))
- Prevent account data syncs getting lost across TCP replication. ([\#6333](#6333))
- Fix bug: TypeError in `register_user()` while using LDAP auth module. ([\#6406](#6406))
- Fix an intermittent exception when handling read-receipts. ([\#6408](#6408))
- Fix broken guest registration when there are existing blocks of numeric user IDs. ([\#6420](#6420))
- Fix startup error when http proxy is defined. ([\#6421](#6421))
- Fix error when using synapse_port_db on a vanilla synapse db. ([\#6449](#6449))
- Fix uploading multiple cross signing signatures for the same user. ([\#6451](#6451))
- Fix bug which lead to exceptions being thrown in a loop when a cross-signed device is deleted. ([\#6462](#6462))
- Fix `synapse_port_db` not exiting with a 0 code if something went wrong during the port process. ([\#6470](#6470))
- Improve sanity-checking when receiving events over federation. ([\#6472](#6472))
- Fix inaccurate per-block Prometheus metrics. ([\#6491](#6491))
- Fix small performance regression for sending invites. ([\#6493](#6493))
- Back out cross-signing code added in Synapse 1.5.0, which caused a performance regression. ([\#6494](#6494))

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

- Update documentation and variables in user contributed systemd reference file. ([\#6369](#6369), [\#6490](#6490))
- Fix link in the user directory documentation. ([\#6388](#6388))
- Add build instructions to the docker readme. ([\#6390](#6390))
- Switch Ubuntu package install recommendation to use python3 packages in INSTALL.md. ([\#6443](#6443))
- Write some docs for the quarantine_media api. ([\#6458](#6458))
- Convert CONTRIBUTING.rst to markdown (among other small fixes). ([\#6461](#6461))

Deprecations and Removals
-------------------------

- Remove admin/v1/users_paginate endpoint. Contributed by Awesome Technologies Innovationslabor GmbH. ([\#5925](#5925))
- Remove fallback for federation with old servers which lack the /federation/v1/state_ids API. ([\#6488](#6488))

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

- Add benchmarks for structured logging and improve output performance. ([\#6266](#6266))
- Improve the performance of outputting structured logging. ([\#6322](#6322))
- Refactor some code in the event authentication path for clarity. ([\#6343](#6343), [\#6468](#6468), [\#6480](#6480))
- Clean up some unnecessary quotation marks around the codebase. ([\#6362](#6362))
- Complain on startup instead of 500'ing during runtime when `public_baseurl` isn't set when necessary. ([\#6379](#6379))
- Add a test scenario to make sure room history purges don't break `/messages` in the future. ([\#6392](#6392))
- Clarifications for the email configuration settings. ([\#6423](#6423))
- Add more tests to the blacklist when running in worker mode. ([\#6429](#6429))
- Refactor data store layer to support multiple databases in the future. ([\#6454](#6454), [\#6464](#6464), [\#6469](#6469), [\#6487](#6487))
- Port synapse.rest.client.v1 to async/await. ([\#6482](#6482))
- Port synapse.rest.client.v2_alpha to async/await. ([\#6483](#6483))
- Port SyncHandler to async/await. ([\#6484](#6484))
@erikjohnston erikjohnston deleted the erikj/query_devices_fix branch January 9, 2020 15:47
babolivier pushed a commit that referenced this pull request Sep 1, 2021
@anoadragon453
Merge pull request #6332 from matrix-org/erikj/query_devices_fix
5c881dd 
* commit 'f9f1c8acb':
  Fixup docs
  Newsfile
  Fix caching devices for remote servers in worker.
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Reviewers

@richvdh richvdh richvdh approved these changes

Assignees

@erikjohnston erikjohnston

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@erikjohnston @richvdh