Open thefinn93 opened 1 week ago
We had similar errors, but resolved them by restarting MAS and running provision-all-users
.
The whole thing felt a bit buggy for such a central authentication software. Maybe MAS can automatically try to reprovision the user in Synapse if it receives a 404 on device creation?
This is basically due to the job queue being unreliable.
See #2625 and #1490
We've started rewriting the job queue completely to make it reliable and properly retry lost/errored jobs, which should solve this issue
You should subscribe to #2785 to check with the progress
Describe the bug On our MAS instance, every time we attempt to onboard a new user, they are prevented from completing login and shown error
request failed with status 404 Not Found: M_NOT_FOUND: Unknown user
. A sample of the MAS logs when this happens:To Reproduce Steps to reproduce the behavior:
Expected behavior User should be able to login to matrix normally.
Screenshots
Desktop (please complete the following information): This seems to happen to people on many platforms.
Smartphone (please complete the following information): This seems to happen to people on many platforms.
Additional context
#matrix-auth:matrix.org
the first time it happened, some workarounds were suggested. I triedmanage provision-all-users
at the time it was believed that the worker just got stuck. That seemed plausible, but it has since happened several times.cleanup-expired-tokens
jobs continuing to complete, suggesting the jobs are not just stuck: According to the logs, the errors happened at 19:20:50, 19:21:38, 19:22:08 and 19:24:20. I restarted the process shortly after that.