We were `json.Unmarshal`ing the EDU and `json.Marshal`ing right before
sending the EDU to the stream. Those are now removed and the consumer
does `json.Unmarshal` once.
`If a device list update goes missing, the server resyncs on the next
one` was failing because a previous test would receive a `waitTime` of
1h, resulting in the test timing out.
This now tries to handle the returned errors differently, e.g. by using
the default `waitTime` of 2s. Also doesn't try further users in the
list, if one of the errors would cause a longer `waitTime`.
This makes the following changes:
* The various `Defaults` functions are now responsible for setting sane defaults if `generate` is specified, rather than hiding them in `generate-config`
* Some configuration options have been marked as `omitempty` so that they don't appear in generated configs unnecessarily (monolith-specific vs. polylith-specific options)
* A new option `-polylith` has been added to `generate-config` to create a config that makes sense for polylith deployments (i.e. including the internal/external API listeners and per-component database sections)
* A new option `-normalise` has been added to `generate-config` to take an existing file and add any missing options and/or defaults
This PR refactors the app services component. It makes the following changes:
* Each appservice now gets its own NATS JetStream consumer
* The appservice database is now removed entirely, since we just use JetStream as a data source instead
* The entire component is now much simpler and we deleted lots of lines of code 💅
The result is that it should be much lighter and hopefully much more performant.
This refactors the `dendrite-demo-pinecone` executable so that it:
1. Converts the old `.key` file into a standard `.pem` file
2. Allows passing in the `--config` option to supply a normal Dendrite configuration file, so that you can configure PostgreSQL instead of SQLite, appservices and all the other usual stuff
Introduced index improves select query performance. Example execution time of `selectSendToDeviceMessagesSQL` query dropped from 80 ms to 15 ms. No sytest modifications are required.
### Pull Request Checklist
* [x] I have added added tests for PR _or_ I have justified why this PR doesn't need tests.
* [x] Pull request includes a [sign off](https://github.com/matrix-org/dendrite/blob/main/docs/CONTRIBUTING.md#sign-off)
Signed-off-by: `Piotr Kozimor <p1996k@gmail.com>`
This should fix an issue where we return less than the expected membership events, when doing an initial sync.
When doing an initial sync, the state limit is set to `math.MaxInt32`, while the default filter is set to 20.
- Reverts 9dc57122d9 as it was causing issues https://github.com/matrix-org/dendrite/issues/2660
- Updates the GMSL `DefaultStateFilter` to use a limit of 20 events
- Uses the timeline events to determine the new position instead of the state events
This PR does the following:
- adds a `keysize` parameter to `generate-keys`, so we can use lower sized keys when running in CI
- updates the Complement docker files to use BuildKit (requires Docker >18.09)
- uses `exec` when executing `dendrite-monotlith-server`, making it PID 1 inside docker, which results in Dendrite actually receiving the `SIGTERM` signal send by Docker. (Making it faster when running tests with Complement, as we don't take 10 seconds to timeout)
A timeout of 10 seconds could cause issues with servers having a high `bcrypt_cost` configured in the config.
This adds a parameter to manually configure the timeout, defaults to 30 seconds.
commit 1929b688e31987c46e0c8a546f0f9cb0a46bf9a3
Author: Neil Alexander <neilalexander@users.noreply.github.com>
Date: Mon Aug 22 10:09:44 2022 +0100
Still process state-before for soft-failed events
commit e83c0b701d40d78b92072c4643f6bc6f71b72800
Author: Neil Alexander <neilalexander@users.noreply.github.com>
Date: Mon Aug 22 10:06:50 2022 +0100
Improve logging
commit 29e26124bc27cb83d449de2a4214b253c594aa93
Author: Neil Alexander <neilalexander@users.noreply.github.com>
Date: Mon Aug 22 09:58:13 2022 +0100
Don't store soft-failed events as rejected
### Fixes
* A bug in the roomserver around handling rejected outliers has been fixed
* Backfilled events will now use the correct history visibility where possible
* The device list updater backoff has been fixed, which should reduce the number of outbound HTTP requests and `Failed to query device keys for some users` log entries for dead servers
* The `/sync` endpoint will no longer incorrectly return room entries for retired invites which could cause some rooms to show up in the client "Historical" section
* The `/createRoom` endpoint will now correctly populate `is_direct` in invite membership events, which may help clients to classify direct messages correctly
* The `create-account` tool will now log an error if the shared secret is not set in the Dendrite config
* A couple of minor bugs have been fixed in the membership lazy-loading
* Queued EDUs in the federation API are now cached properly
This ensures that if the device list updater is already backing off a node, we don't try to call processServer again anyway for server just because the server name arrived in the channel. Otherwise we can keep trying to hit a remote server that is offline or not behaving every second and that spams the logs too.
This should hopefully deflake Backfill works correctly with history visibility set to joined as we were using the default shared visibility, even if the events are set to joined (or something else)
In some conditions (fast CPUs), this test would race the clock for EDU expiration when all we want to make sure of is that the expired EDUs are properly deleted. Given this, we set the expiry time to 0 so the specified EDUs are always deleted when DeleteExpiredEDUs is called.
Fixes#2650.
Signed-off-by: Winter <winter@winter.cafe>
