-
-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Avoid backfill when we already have messages to return #15737
Avoid backfill when we already have messages to return #15737
Conversation
…already-have-enough-messages
member_event_id | ||
if leave_token.topological < curr_topo: | ||
from_token = from_token.copy_and_replace( | ||
StreamKeyType.ROOM, leave_token |
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.
This logic is just pulled out from the existing condition to de-nest it further up here and have it ready when we initially fetch.
…already-have-enough-messages
# If we did backfill something, refetch the events from the database to | ||
# catch anything new that might have been added since we last fetched. |
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.
In the worst case now when we do need to backfill, it does mean that we call paginate_room_events()
twice but I think the savings in the happy-path is well worth it.
Plus a slow backfill that is 10+ seconds for the federation request and processing doesn't hurt from an extra 250ms paginate_room_events()
call again (it's not the slowest piece about this whole process to worry about here anyway).
# We don't expect a negative depth but we'll just deal with it in | ||
# any case by taking the absolute value to get the true gap between | ||
# any two integers. | ||
depth_gap = abs(event_depth - previous_event_depth) |
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.
Since the list is sorted won't this always be positive even if there are negative depths?
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.
Yes
ex. -5 - -6 = 1
🤷 Perhaps this gets across that we care about the absolute distance, not the difference in any case.
Thanks for the review @erikjohnston 🐨 |
Please note that this will be the last release of Synapse that is compatible with Python 3.7 and earlier. This is due to Python 3.7 now having reached End of Life; see our [deprecation policy](https://matrix-org.github.io/synapse/v1.87/deprecation_policy.html) for more details. - Pin `pydantic` to `^1.7.4` to avoid backwards-incompatible API changes from the 2.0.0 release. Resolves matrix-org#15858. Contributed by @PaarthShah. ([\matrix-org#15862](matrix-org#15862)) - Split out 2022 changes from the changelog so the rendered version in GitHub doesn't timeout as much. ([\matrix-org#15846](matrix-org#15846)) - Improve `/messages` response time by avoiding backfill when we already have messages to return. ([\matrix-org#15737](matrix-org#15737)) - Add spam checker module API for logins. ([\matrix-org#15838](matrix-org#15838)) - Fix a long-standing bug where media files were served in an unsafe manner. Contributed by @joshqou. ([\matrix-org#15680](matrix-org#15680)) - Avoid invalidating a cache that was just prefilled. ([\matrix-org#15758](matrix-org#15758)) - Fix requesting multiple keys at once over federation, related to [MSC3983](matrix-org/matrix-spec-proposals#3983). ([\matrix-org#15770](matrix-org#15770)) - Fix joining rooms through aliases where the alias server isn't a real homeserver. Contributed by @tulir @ Beeper. ([\matrix-org#15776](matrix-org#15776)) - Fix a bug in push rules handling leading to an invalid (per spec) `is_user_mention` rule sent to clients. Also fix wrong rule names for `is_user_mention` and `is_room_mention`. ([\matrix-org#15781](matrix-org#15781)) - Fix a bug introduced in 1.57.0 where the wrong table would be locked on updating database rows when using SQLite as the database backend. ([\matrix-org#15788](matrix-org#15788)) - Fix Sytest environmental variable evaluation in CI. ([\matrix-org#15804](matrix-org#15804)) - Fix forgotten rooms missing from initial sync after rejoining them. Contributed by Nico from Famedly. ([\matrix-org#15815](matrix-org#15815)) - Fix sqlite `user_filters` upgrade introduced in v1.86.0. ([\matrix-org#15817](matrix-org#15817)) - Document `looping_call()` functionality that will wait for the given function to finish before scheduling another. ([\matrix-org#15772](matrix-org#15772)) - Fix a typo in the [Admin API](https://matrix-org.github.io/synapse/latest/usage/administration/admin_api/index.html). ([\matrix-org#15805](matrix-org#15805)) - Fix typo in MSC number in faster remote room join architecture doc. ([\matrix-org#15812](matrix-org#15812)) - Remove experimental [MSC2716](matrix-org/matrix-spec-proposals#2716) implementation to incrementally import history into existing rooms. ([\matrix-org#15748](matrix-org#15748)) - Replace `EventContext` fields `prev_group` and `delta_ids` with field `state_group_deltas`. ([\matrix-org#15233](matrix-org#15233)) - Regularly try to send transactions to other servers after they failed instead of waiting for a new event to be available before trying. ([\matrix-org#15743](matrix-org#15743)) - Fix requesting multiple keys at once over federation, related to [MSC3983](matrix-org/matrix-spec-proposals#3983). ([\matrix-org#15755](matrix-org#15755)) - Allow for the configuration of max request retries and min/max retry delays in the matrix federation client. ([\matrix-org#15783](matrix-org#15783)) - Switch from `matrix://` to `matrix-federation://` scheme for internal Synapse routing of outbound federation traffic. ([\matrix-org#15806](matrix-org#15806)) - Fix harmless exceptions being printed when running the port DB script. ([\matrix-org#15814](matrix-org#15814)) * Bump attrs from 22.2.0 to 23.1.0. ([\matrix-org#15801](matrix-org#15801)) * Bump cryptography from 40.0.2 to 41.0.1. ([\matrix-org#15800](matrix-org#15800)) * Bump ijson from 3.2.0.post0 to 3.2.1. ([\matrix-org#15802](matrix-org#15802)) * Bump phonenumbers from 8.13.13 to 8.13.14. ([\matrix-org#15798](matrix-org#15798)) * Bump ruff from 0.0.265 to 0.0.272. ([\matrix-org#15799](matrix-org#15799)) * Bump ruff from 0.0.272 to 0.0.275. ([\matrix-org#15833](matrix-org#15833)) * Bump serde_json from 1.0.96 to 1.0.97. ([\matrix-org#15797](matrix-org#15797)) * Bump serde_json from 1.0.97 to 1.0.99. ([\matrix-org#15832](matrix-org#15832)) * Bump towncrier from 22.12.0 to 23.6.0. ([\matrix-org#15831](matrix-org#15831)) * Bump types-opentracing from 2.4.10.4 to 2.4.10.5. ([\matrix-org#15830](matrix-org#15830)) * Bump types-setuptools from 67.8.0.0 to 68.0.0.0. ([\matrix-org#15835](matrix-org#15835))
Avoid backfill when we already have messages to return.
We now only block the client to backfill when we see a large gap in the events (more than 2 events missing in a row according to
depth
), more than 3 single-event holes, or not enough messages to fill the response. Otherwise, we return the messages directly to the client and backfill in the background for eventual consistency sake.Fix #15696
Figure out if this change would skip backfill in a real-world scenario
For a real-world example with the trace linked from the issue (from viewing the Matrix Public Archive), it makes the following
/messages
request. If we run the JS snippet over the response, we can see that there are no gaps in response and we didn't really need to backfill at all. Which means this PR is a good change and would allow us not to block the client waiting for backfill which equates to 14s savings out of that 16s request!GET https://matrix-client.matrix.org/_matrix/client/r0/rooms/!OGEhHVWSdvArJzumhm%3Amatrix.org/messages?dir=b&from=t535491-3610842373_0_0_0_0_0_0_0_0_0&limit=101&filter=%7B%22lazy_load_members%22%3Atrue%2C%22event_format%22%3A%22federation%22%7D
Dev notes
Other PR where we did some backfill/processing in the background, #15585
MSC3871: Gappy timelines
Pull Request Checklist
EventStore
toEventWorkerStore
.".code blocks
.Pull request includes a sign off(run the linters)