[Solved] Talk crashes with 502 when starting a call

Hi,

I’m having issue with Talk on new 18.0.6 install.
Call window crashes, Chrome suggests to close it right away.
I see some 503 errors from spreed API in log.

TURN is configured and tested OK.

Conslole:
xhr.js:178 POST https://server/ocs/v2.php/apps/spreed/api/v1/room/nx2572s2/participants/active 502
(anonymous) @ xhr.js:178

Are you running HTTP2 / FPM combines with mpm_prefork?
Or are you using SQLite?

MariaDB and nginx. But FPM is present, yes.
What should I check?

Any hints guys?

Btw mobile version works fine

You must not use mpm_prefork together with FPM.
Switch to mpm_events instead.

We are not using Apache.
nginx + php-fpm

Failed to load resource: the server responded with a status of 502 ()
CallButton.vue:162 Joining call
webrtc.js:803 localMediaStarted
/ocs/v2.php/apps/spreed/api/v1/call/452qe72t:1 Failed to load resource: the server responded with a status of 404 ()
signaling.js:307 Uncaught (in promise) Error
at signaling.js:307
CallButton.vue:173

Well than that is your issue.
Just as an info, with talk in a call, you have 2 long running requests per user all the time, so make sure you have enough works that all other requests are still served in time.

Checked out /var/log/php-fpm/error.log:

[23-Jun-2020 19:31:47] NOTICE: systemd monitor interval set to 10000ms
[23-Jun-2020 19:31:54] WARNING: [pool www] child 6042 exited on signal 11 (SIGSEGV) after 7.046424 seconds from start
[23-Jun-2020 19:31:54] NOTICE: [pool www] child 6044 started
[23-Jun-2020 19:32:51] WARNING: [pool www] child 6041 exited on signal 11 (SIGSEGV) after 64.163499 seconds from start

[23-Jun-2020 19:32:51] NOTICE: [pool www] child 6060 started
[23-Jun-2020 19:40:04] WARNING: [pool www] child 6034 exited on signal 11 (SIGSEGV) after 496.901954 seconds from start

This segfault in fpm causes nginx to fail with 502.

The reason may be anything - to many children, not enough RAM, not writable sessions directory. I had to disable opcache [opcache.enable=0]