Multiple previously working client instances are unable to connect(hang) following upgrade of homeserver and pgsql

[derzahla]

Description

Element instances that were previously configured on various devices(iOS, android and linux) no longer connect to the homeserver. The client will just hang forever trying to load. Stopping the client(updating), clearing all cache, and then attempting to login fresh has the same results. This seems to coincide with a recent homeserver upgrade a few weeks ago that also moved from pgsql 12->13.

additionall: On my primary linux client which is still functional, I am noticing element/electron pegging the CPU at 100% very often and needing to be manually killed/restarted

Steps to reproduce

• launch element on devices that used to work without error

There are no obvious errors in the client or server logs that have proven to be relevent, but I have been chasing various around presumably unrelated errors for the last week or so, ie. element-hq/element-web#21727 and element-hq/element-desktop#871.

Version information

• Version: matrix-synapse-1.55.2

• Install method: pkg

• Platform: FreeBSD 13

TrueNAS 13 iocage jail

[DMRobertson]

Do you have server logs? Can you track down /login requests which correspond to the clients that cannot login?

[derzahla]

Yes. Here is the latest example from today. Only two lines from the client match '/login' though so not sure how useful it will be

2022-04-20 11:47:35,984 - synapse.access.http.8008 - 427 - INFO - GET-32133- 10.10.27.1 - 8008 - {None} Processed request: 0.002sec/0.004sec (0.005sec, 0.000sec ) (0.000sec/0.000sec/0) 136B 200 "GET /_matrix/client/r0/login HTTP/1.0" "Mozilla/5.0 (X11; Linux aarch64) AppleWebKit/537.36 (KHTML, like Gecko) Element/1.10.3 Chrome/98.0.4758.82 Electron/17.0.1 Safari/537.36" [0 dbevts] 2022-04-20 11:47:46,022 - synapse.access.http.8008 - 427 - INFO - POST-32149- 10.10.27.1 - 8008 - {None} Processed request: 0.341sec/0.000sec (0.311sec, 0.000se c) (0.001sec/0.028sec/5) 204B 200 "POST /_matrix/client/r0/login HTTP/1.0" "Mozilla/5.0 (X11; Linux aarch64) AppleWebKit/537.36 (KHTML, like Gecko) Element/1.10 .3 Chrome/98.0.4758.82 Electron/17.0.1 Safari/537.36" [0 dbevts] 2022-04-20 11:47:46,258 - synapse.access.http.8008 - 427 - INFO -

[DMRobertson]

If you register a completely new user and try to login, do you see the same delay?

[reivilibre]

Please could you be a bit clearer about what you mean when the client 'hangs'... is it showing a loading screen? Frozen completely (UI lock-up)? A screenshot would be appreciated if easy to do.

I think it would be nice to have more server logs to see what's going on here, as well. The /login request itself looks fine and fast.

[CalebFenton]

I also have a problem with hanging after login, but my setup is brand new with only a single admin created. But others are complaining about this problem after an upgrade:

• Logging in a new device hangs the client and eventually crashes the server #12071
• The state_group_edges table doesn't have a unique index; can severely impact WITH RECURSIVE state(state_group) AS ... query performance if duplicates are somehow introduced #11779.

[derzahla]

Sorry, I dont have any further information to provide. I ended up just rolling back everything to a 3 week old snapshot. I did the postgres upgrade to 13 and no problems this time.