Conversation

Notices

1. Sorokin Alexei (xrevan86@loadaverage.org)'s status on Friday, 14-Aug-2020 04:30:48 JST Sorokin Alexei

   • VegOs ✔♋
   @vegos Not sure what causes this – TwitterBridge has always been a mystery to me.
   But I'm looking into it.
   • Sorokin Alexei (xrevan86@loadaverage.org)'s status on Friday, 14-Aug-2020 08:52:10 JST Sorokin Alexei

     • VegOs ✔♋
     @vegos It seems that this was caused by a flaw in the RedisCache plug-in, which shared connections between daemon threads.
     Something about TwitterBridge made it more apparent.
   • Sorokin Alexei (xrevan86@loadaverage.org)'s status on Sunday, 16-Aug-2020 22:46:20 JST Sorokin Alexei

     • LoadAverage.org instance
     • VegOs ✔♋
     > Thanks, I had to read first, what that Redis is. Never ate it, but I'm pleased, you care about it.
     @vegos That was a technicality. It's a performant key-value database that can be used for caching and queues in GNU social.
     And the plug-in in GNU social for caching via Redis didn't account for that when the GNU social daemons fork the new processes need to reconnect to Redis.
     Otherwise different commands from different processes that share the same Redis connection will start receiving replies from Redis that could have been meant for a different process.
     Sometimes that meant a PHP crash, which is what stopped some posts from being sent to Twitter.
     
     !loadaverage had the plug-in enabled for about month, I'm not sure why this issue got more evident now.