On Monday 29 August 2011, MÉSZÁROS Mihály wrote:
Your patch didn't work for me. The kamailio
didn't start if i use "if (
_rank != PROC_MAIN)"
So i commented out, both if statement, and now "out of sync problem
disappeared"
That should do the trick, but it might be possible to create a
zombie
connection this way. I don't know how the initialization of modules works
exactly, so I can't really help you with this.
BUT
I am experiencing that in usrloc module, still more then one worker
process share the same sql connection:
That is weird. AFAIK, connections can't be shared between processes, there
seems to be no locking/semaphores/whatever to protect against simultaneous
access.
Maybe you got lucky in your tests and just didn't have enough traffic to hit
the simultaneous access problem (yet).
I'm out of ideas here, sorry.
--
Greetings,
Alex Hermann