-
-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Use supervisord to supervise Postgres and Caddy in the Complement image. #12480
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
it looks like supervisord still isn't ending up as PID 1 within the container.
Try replacing the entrypoint with an explicit shell invocation:
ENTRYPOINT ["/bin/bash", "/start-complement-synapse-workers.sh"]
from https://docs.docker.com/engine/reference/builder/#entrypoint:
The shell form ... has the disadvantage that your
ENTRYPOINT
will be started as a subcommand of/bin/sh -c
, which does not pass signals. This means that the executable will not be the container’sPID 1
- and will not receive Unix signals - so your executable will not receive aSIGTERM
fromdocker stop <container>
.
(TIL. We probably ought to do something similar in our other Dockerfiles)
It looks like supervisord is now PID 1 and the container stops within 4 seconds. |
Do we actually need postgres to be a higher prority than synapse? It's not like there's any guarantee that doing so will mean that postgres is ready before synapse starts.
Is it just synapse that is slow? I could be persuaded that SIGKILLing it might be good. Think I'd make it a separate PR though. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm otherwise
No, though Synapse's configuration already defined it at priority Supervisor by default will check that a process is up for 1 second before considering it 'up' (see
I couldn't remember, so I tried it again. This is the timeline of stopping:
It seems faster than I was thinking, so maybe it's not a problem. (That's 2.1 seconds from stop signal to stopped.) Footnotes
|
Co-authored-by: Richard van der Hoff <1389908+richvdh@users.noreply.github.com>
# Lower priority number = starts first | ||
priority=100 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
following up on #12480 (comment):
Do we actually need postgres to be a higher prority than synapse? It's not like there's any guarantee that doing so will mean that postgres is ready before synapse starts.
No, though Synapse's configuration already defined it at priority
500
so it seemed like this was the intention.
I'm not sure the priority was particularly carefully chosen, tbh. (ftr, this bit came from #9162).
(I just realised: The main Synapse process' priority is actually defined as
10
; Redis' is1
. Postgres should probably be on the same level as Redis if we're going to have priority ordering?)
If we're going to have priority ordering, then yes it ought to be lower than Synapse.
I would have thought that Postgres being up first might avoid some noisy errors in Synapse's logs.
If it actually manages to start up in 1 second, then yes maybe. TBH though I would optimise for fastest runtime rather than lack of noise in the logs.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
let's set this to be the same as Redis for now, and get this PR merged.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
oh you're way ahead of me
Re SIGKILL:
In this case it looks like it's stopping postgres before synapse stops? Had you already changed the priorities? I can imagine that stopping postgres will encourage synapse to stop quickly. (I can also imagine the opposite...) Let's leave this for now, either way. |
This was added in matrix-org/complement#62. My understanding is that the basic synapse-workers image uses an |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm!
This was taken from Postgres not having an explicit priority (so priority 999), in which case it starts after Synapse and stops before it. With priority = 1 (same as Redis):
2 sec, so no likely difference to the other way around. I'll go with this for now since it makes most sense.
Good point. |
The tests still fail:
so there looks to be more work to do in further PRs. |
Synapse 1.59.0rc1 (2022-05-10) ============================== This release makes several changes that server administrators should be aware of: - Device name lookup over federation is now disabled by default. ([\#12616](#12616)) - The `synapse.app.appservice` and `synapse.app.user_dir` worker application types are now deprecated. ([\#12452](#12452), [\#12654](#12654)) See [the upgrade notes](https://github.com/matrix-org/synapse/blob/develop/docs/upgrade.md#upgrading-to-v1590) for more details. Additionally, this release removes the non-standard `m.login.jwt` login type from Synapse. It can be replaced with `org.matrix.login.jwt` for identical behaviour. This is only used if `jwt_config.enabled` is set to `true` in the configuration. ([\#12597](#12597)) Features -------- - Support [MSC3266](matrix-org/matrix-spec-proposals#3266) room summaries over federation. ([\#11507](#11507)) - Implement [changes](matrix-org/matrix-spec-proposals@4a77139) to [MSC2285 (hidden read receipts)](matrix-org/matrix-spec-proposals#2285). Contributed by @SimonBrandner. ([\#12168](#12168), [\#12635](#12635), [\#12636](#12636), [\#12670](#12670)) - Extend the [module API](https://github.com/matrix-org/synapse/blob/release-v1.59/synapse/module_api/__init__.py) to allow modules to change actions for existing push rules of local users. ([\#12406](#12406)) - Add the `notify_appservices_from_worker` configuration option (superseding `notify_appservices`) to allow a generic worker to be designated as the worker to send traffic to Application Services. ([\#12452](#12452)) - Add the `update_user_directory_from_worker` configuration option (superseding `update_user_directory`) to allow a generic worker to be designated as the worker to update the user directory. ([\#12654](#12654)) - Add new `enable_registration_token_3pid_bypass` configuration option to allow registrations via token as an alternative to verifying a 3pid. ([\#12526](#12526)) - Implement [MSC3786](matrix-org/matrix-spec-proposals#3786): Add a default push rule to ignore `m.room.server_acl` events. ([\#12601](#12601)) - Add new `mau_appservice_trial_days` configuration option to specify a different trial period for users registered via an appservice. ([\#12619](#12619)) Bugfixes -------- - Fix a bug introduced in Synapse 1.48.0 where the latest thread reply provided failed to include the proper bundled aggregations. ([\#12273](#12273)) - Fix a bug introduced in Synapse 1.22.0 where attempting to send a large amount of read receipts to an application service all at once would result in duplicate content and abnormally high memory usage. Contributed by Brad & Nick @ Beeper. ([\#12544](#12544)) - Fix a bug introduced in Synapse 1.57.0 which could cause `Failed to calculate hosts in room` errors to be logged for outbound federation. ([\#12570](#12570)) - Fix a long-standing bug where status codes would almost always get logged as `200!`, irrespective of the actual status code, when clients disconnect before a request has finished processing. ([\#12580](#12580)) - Fix race when persisting an event and deleting a room that could lead to outbound federation breaking. ([\#12594](#12594)) - Fix a bug introduced in Synapse 1.53.0 where bundled aggregations for annotations/edits were incorrectly calculated. ([\#12633](#12633)) - Fix a long-standing bug where rooms containing power levels with string values could not be upgraded. ([\#12657](#12657)) - Prevent memory leak from reoccurring when presence is disabled. ([\#12656](#12656)) Updates to the Docker image --------------------------- - Explicitly opt-in to using [BuildKit-specific features](https://github.com/moby/buildkit/blob/master/frontend/dockerfile/docs/syntax.md) in the Dockerfile. This fixes issues with building images in some GitLab CI environments. ([\#12541](#12541)) - Update the "Build docker images" GitHub Actions workflow to use `docker/metadata-action` to generate docker image tags, instead of a custom shell script. Contributed by @henryclw. ([\#12573](#12573)) Improved Documentation ---------------------- - Update SQL statements and replace use of old table `user_stats_historical` in docs for Synapse Admins. ([\#12536](#12536)) - Add missing linebreak to `pipx` install instructions. ([\#12579](#12579)) - Add information about the TCP replication module to docs. ([\#12621](#12621)) - Fixes to the formatting of `README.rst`. ([\#12627](#12627)) - Fix docs on how to run specific Complement tests using the `complement.sh` test runner. ([\#12664](#12664)) Deprecations and Removals ------------------------- - Remove unstable identifiers from [MSC3069](matrix-org/matrix-spec-proposals#3069). ([\#12596](#12596)) - Remove the unspecified `m.login.jwt` login type and the unstable `uk.half-shot.msc2778.login.application_service` from [MSC2778](matrix-org/matrix-spec-proposals#2778). ([\#12597](#12597)) - Synapse now requires at least Python 3.7.1 (up from 3.7.0), for compatibility with the latest Twisted trunk. ([\#12613](#12613)) Internal Changes ---------------- - Use supervisord to supervise Postgres and Caddy in the Complement image to reduce restart time. ([\#12480](#12480)) - Immediately retry any requests that have backed off when a server comes back online. ([\#12500](#12500)) - Use `make_awaitable` instead of `defer.succeed` for return values of mocks in tests. ([\#12505](#12505)) - Consistently check if an object is a `frozendict`. ([\#12564](#12564)) - Protect module callbacks with read semantics against cancellation. ([\#12568](#12568)) - Improve comments and error messages around access tokens. ([\#12577](#12577)) - Improve docstrings for the receipts store. ([\#12581](#12581)) - Use constants for read-receipts in tests. ([\#12582](#12582)) - Log status code of cancelled requests as 499 and avoid logging stack traces for them. ([\#12587](#12587), [\#12663](#12663)) - Remove special-case for `twisted` logger from default log config. ([\#12589](#12589)) - Use `getClientAddress` instead of the deprecated `getClientIP`. ([\#12599](#12599)) - Add link to documentation in Grafana Dashboard. ([\#12602](#12602)) - Reduce log spam when running multiple event persisters. ([\#12610](#12610)) - Add extra debug logging to federation sender. ([\#12614](#12614)) - Prevent remote homeservers from requesting local user device names by default. ([\#12616](#12616)) - Add a consistency check on events which we read from the database. ([\#12620](#12620)) - Remove use of the `constantly` library and switch to enums for `EventRedactBehaviour`. Contributed by @andrewdoh. ([\#12624](#12624)) - Remove unused code related to receipts. ([\#12632](#12632)) - Minor improvements to the scripts for running Synapse in worker mode under Complement. ([\#12637](#12637)) - Move `pympler` back in to the `all` extras. ([\#12652](#12652)) - Fix spelling of `M_UNRECOGNIZED` in comments. ([\#12665](#12665)) - Release script: confirm the commit to be tagged before tagging. ([\#12556](#12556)) - Fix a typo in the announcement text generated by the Synapse release development script. ([\#12612](#12612)) - Fix scripts-dev to pass typechecking. ([\#12356](#12356)) - Add some type hints to datastore. ([\#12485](#12485)) - Remove unused `# type: ignore`s. ([\#12531](#12531)) - Allow unused `# type: ignore` comments in bleeding edge CI jobs. ([\#12576](#12576)) - Remove redundant lines of config from `mypy.ini`. ([\#12608](#12608)) - Update to mypy 0.950. ([\#12650](#12650)) - Use `Concatenate` to better annotate `_do_execute`. ([\#12666](#12666)) - Use `ParamSpec` to refine type hints. ([\#12667](#12667)) - Fix mypy against latest pillow stubs. ([\#12671](#12671))
This is still leading to corrupt Postgres databases, need to investigate.