[08:57:41] this reminded me to check apc.shm_size for my php 8 installation and sure enough it was the default 32M instead of 1G that I had for 7.4 [13:17:02] …wow, shm_size affects load averages a whole lot, the values dropped from 1.2 to 0.4 even when I use a very agressively caching Varnish in front [20:59:14] Hi, how can I start to investigate why users are getting 'session_fail_preview' msg? [21:11:53] Guest85: typically this is an issue if things fall out of cache too quickly [21:12:36] Guest85: you can try setting $wgSessionCacheType = CACHE_DB; to tell it to store user sessions in db rather than the more ephermal cache, which is slightly slower, but won't have the issue of things getting deleted too soon [21:13:11] If using apcu (CACHE_ACCEL), check what apcu.shm_size is set to in php.ini [21:13:13] Actually I'm with CACHE_DB option [21:13:27] Ah ok. In that case its probably a different cause [21:13:34] There is any way to check the log and identify why it happened? [21:14:13] MediaWiki debug log might have some information about why the user session failed. It might not be the easiest to interpret [21:14:39] This is part of 'happy flow' of the user or it's an exception? [21:14:59] In theory its not supposed to happen, or at least if it does its supposed to be very rare [21:17:10] Its basically saying that it looks like the save page came from a different user session than the one that started editing. Which really should only happen if the user logs in/logs out between starting the edit and finishing the edit [21:17:50] Sometimes though, it can happen if the user deletes their cookie or if for some reason the user session was prematurely deleted [21:18:30] If you have any authentication extensions they can potentially affect things [21:21:07] Got you, thanks for help!