| 11 Nov 2024 |
 Hank | 
Download LocalStatsChart.png | 02:24:32 |
|  @vangelis369:matrix.org joined the room. | 05:30:02 |
| @vangelis369:matrix.org joined the room. | 05:30:15 |
| @vangelis369:matrix.org left the room. | 05:31:07 |
 bkil | @pepecyb What's the timeout for nginx? Have you found anything interesting in the logs of Friendica itself? Have you tried to connect your Friendica instance of the Hubzilla instance of third parties as a test? You may also look into your Hubzilla logs also.You may be impacted by https://github.com/friendica/friendica/pull/14486 https://github.com/friendica/friendica/issues/14145#event-14767981426 | 07:34:09 |
 pepecyb | bkil: Yes, I have been able to connect to channels on other (third-party) hubs from my Friendica channel. No problems.
Trying to connect by prepending [activitypub] to the handle failed with the message ‘The network type was not recognised. The contact cannot be added.’
The fact is, however, that the channel has ActivityPub switched on (and can otherwise connect to all possible AP services without any problems and federates reliably). | 11:41:16 |
| pepecyb | After another connection attempt at 11:40:
Nginx error-log:
| 11:50:03 |
| pepecyb | 2024/11/11 11:40:44 [error] 3234024#3234024: *11392818 upstream timed out (110: Connection timed out) while reading response header from upstream, client: 5.204.231.9, server: barat.fedi-verse.hu, request: "GET /contact/follow?url=pepecyb%40hub.hubzilla.hu HTTP/2.0", upstream: "fastcgi://unix:/var/run/php/php8.2-fpm-friendica.sock", host: "barat.fedi-verse.hu", referrer: "https://barat.fedi-verse.hu/contact" | 11:50:10 |
| pepecyb | The Friendica log is growing so fast that I can hardly find the right lines. But I haven't found anything that would make me any clearer either. | 11:51:05 |
| bkil | Redacted or Malformed Event | 11:53:54 |
| bkil | I see HTTP/2.0 here. Could you try with HTTP 1.1 perhaps? It may be relevant with regards to the configuration of the web server. | 11:54:44 |
| pepecyb | bkil: Where would I have to change something in order to switch to HTTP 1.1? | 12:08:30 |
| bkil | https://stackoverflow.com/questions/39347054/how-can-i-prevent-nginx-from-serving-http-2 | 12:19:10 |
| pepecyb | Puuuh... | 12:22:54 |
| bkil | How about into the Hubzilla logs and the nginx serving Hubzilla as well? | 12:24:26 |
| bkil | * How about looking into the Hubzilla logs and the nginx serving Hubzilla as well? | 12:24:32 |
| pepecyb | bkil: The Hubzilla hubs also run with HTTP/2.0
I can't find any anomalies in the logs. It seems that the connection request is not reaching the hubs at all.
Which surprises me, because they are two different hubs running on different computers and they were not installed absolutely identically.
| 12:35:20 |
| bkil | Firewalls come to mind after you have checked with HTTP/1.1 | 13:18:59 |
| pepecyb | Ok... | 13:28:16 |
| pepecyb | Addendum: The few (older) Hubzilla connections I had are ok. But now I am no longer able to add any Hubzilla channels to my connections. Not even those on hubs where the existing connections are also located. Nothing works with Hubzilla any more.
So it must be the Friendica instance. What surprises me: I haven't changed anything since the installation. | 14:25:37 |
| Hank | Like never even upgraded versions or anything? | 14:28:38 |
| pepecyb | No, I installed the instance at the beginning of October. Then I added the connections (including the Hubzilla connections). In the meantime, there have been no updates from Friendica. | 14:35:08 |
| 12 Nov 2024 |
|  @kayden45:matrix.org joined the room. | 10:21:22 |
| @kayden45:matrix.org | Redacted or Malformed Event | 11:04:17 |
| bkil banned @kayden45:matrix.org (spam). | 11:41:08 |
| pepecyb | So... just now, all of a sudden, connection requests (from connection attempts I made yesterday) are coming in from my Hubzilla channels... and it works. And vice versa, I was able to connect to Hubzilla channels from my Friendica instance... without time-out.
That's nice! But that doesn't increase my confidence in the stability of Friendica.
Currently (emphasis on ‘currently’) the error is probably fixed and no longer acute.
| 21:19:58 |
| bkil | Something must be up with processing your queue on either side. You may be seeing jobs failing, their error trace and them getting repeatedly readded to the queue until they are finally let go perhaps. | 21:24:52 |
| pepecyb | bkil: Then the problem with the queue must suddenly be gone, because I can now make consecutive connections to any Hubzilla channel without running into a time-out and without the connection failing. | 21:47:32 |
| pepecyb | BTW: When I try to connect to Hubzilla channels by prepending [activitypub], it breaks with the error message ‘The network type was not recognised. The contact cannot be added.’.
If I connect to the channel without the prefix, the connection is established normally and it appears in the contact list as ‘hubzilla (AP)’. | 21:53:29 |
 heluecht | This [activitypub] is some Hubzilla functionality. | 22:25:17 |
