-
-
Notifications
You must be signed in to change notification settings - Fork 2.1k
Repeated exceptions in background_process_metrics #7840
Comments
According to the logs, this seems to be happening due to stream orderings getting out of sync in the calculation of the room directory. A temporary fix would be to switch Additionally interesting that the error seems to be choking on a transaction containing user presence EDUs. I assume presence (config option |
|
@kpfleming Ok, I had a hunch on whether #7508 ignoring incoming updates had been breaking things if presence had been toggled on and off, but that doesn't sound like the case here. |
If it helps, the only 'unusual' activity recently on this homeserver was an attempt to create a direct-chat room with a user on matrix.org, but while the federation configuration for km6g.us was broken (the .well-known file was not reachable). |
Do you know how long ago that was? I could see that causing matrix.org to stop sending events to you for a little bit, but I don't think it would cause event streams to become unordered. This seems like a bug internal to the homeserver. |
It would have been on July 11 or 12. |
to help with searching, the exception is:
|
I saw this previously, and it was caused by somebody migrating their database from one postgres instance to another without correctly copying the |
Yes! I did indeed do that, and did not think about copying the sequence generators. I ran a full dump of the database and imported it on the new machine. |
well, I'd venture to suggest you did not run a full dump of the database :). Try this:
and then restart synapse. However, a warning: your database may well now be somewhat corrupt. If you see other odd behaviour, it could be caused by this. |
raised opened #7938 to add a check for this |
Thanks; this homeserver only services two users and we don't have much data in it, so in the worst case I can rebuild from an empty database. I used In any case, I know the cause now and can address it, so thanks! |
I dropped the database, created a new one, and let synapse set up the schema, and re-registered my user account. It's all working as expected so this confirms that the problem was an improper transfer of the database. |
Description
Looks like the same issue as #6405, but this is happening with 1.17.0. Frequent (every ~10 seconds) exceptions thrown from
notify_new_event
operations.Steps to reproduce
Not sure; this just started happening recently.
Log is at https://gist.github.com/kpfleming/b2fdf66e545b33c18200361689e0341e
Version information
If not matrix.org:
Version: 1.17.0
Install method: pip into a virtual environment
Platform: Debian Buster on an amd64 machine, running in a systemd-nspawn container
The text was updated successfully, but these errors were encountered: