-
-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Complement flake(?): GET_/rooms/:room_id/aliases_lists_aliases
#14183
Comments
Why are we suddenly seeing this today? https://github.com/matrix-org/synapse/actions/runs/3251269844/jobs/5336107983 |
https://github.com/matrix-org/synapse/actions/runs/3252498613/jobs/5338825082 seems to be a worker-specific failure(?) |
Test from matrix-org/complement#266. Sytest still exists here: https://github.com/matrix-org/sytest/blame/da1c9e221f998b1d50214b9dd8ab18eb3804d57a/tests/10apidoc/32room-alias.pl#L52 and was deflaked matrix-org/sytest#1056. The deflaking was not ported to complement. |
Indeed, as Rich notes in matrix-org/sytest#1055 (comment):
|
(I wonder if #14165 might have exposed this?) |
I've worked around this with a retry loop, but as noted here, the underlying problem is that Synapse (as an aggregation of workers) isn't offering read-after-write consistency here. |
https://github.com/matrix-org/synapse/actions/runs/3249032216/jobs/5331241845
Edit(dmr): I am trying to reproduce this with
WORKERS=1 POSTGRES=1 COMPLEMENT_DIR=../complement scripts-dev/complement.sh -run 'TestRoomAlias/.*aliases_lists_aliases'
and have manually upped the replication torture setting in Synapse's source.
The text was updated successfully, but these errors were encountered: