mirror of https://github.com/asterisk/asterisk
Prior to this patch, updating the device state cache was done by the thread that originated the event. It would update the cache and then queue the event up for another thread to dispatch. This thread moves the cache updating part to be in the same thread as event dispatching. I was working with someone on a heavily loaded Asterisk system and while reviewing backtraces of the system while it was having problems, I noticed that there were a lot of threads contending for the lock on the event cache. By simply moving this into a single thread, this helped performance *a lot* and alleviated some deadlock-like symptoms. Review: https://reviewboard.asterisk.org/r/2066/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@370664 65c4cc65-6c06-0410-ace0-fbb531ad65f3certified/11.2
parent
9b16c8b0f6
commit
733b46022b
Loading…
Reference in new issue