18 Jul
2016
18 Jul
'16
12:58 p.m.
On 18/07/16 11:23, Dirk Teurlings - Signet B.V. wrote:
I see -- can you describe a bit the changes you did from old config to
the new one (e.g., new modules and routing actions)? If you want to keep
that private, email those details directly to me.
If you didn't get other crashes, I can't think of anything else that can
help from the backtrace.
Cheers,
Daniel
--
Daniel-Constantin Mierla
http://www.asipto.com - http://www.kamailio.org
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
On 07/18/2016 10:49 AM, Daniel-Constantin Mierla
wrote:
OK -- just to have full picture -- did you experienced more crashes than
the two ones you reported?
have you run with -x qm before reverting to 4.2?
No we didn't, as I didn't know what the effects would be, it's a live
server so we don't want to try changes before testing. Is the same config you run with 4.2 and 4.4?
Not at the moment, but we developed the new routes in 4.2 first, then
upgraded the testplatform to 4.4 without issues. Though the testplatform
is not identical in hardware to the live system, this looked promising. The version 4.2.8 has qm as default memory
manager. In 4.4, fm is the
default one but qm can be selected at startup with -x. The main
difference is that qm has more safety checks for detecting double free
and buffer overflows...
Right, that sounds very plausible then... Upgrading
"back" to 4.4 now
will have to wait a bit, we don't want to experience the same issues
again. That being said, it's difficult to track what causes this exactly
if we can't reproduce it on the testplatform. Unfortunately I'm not at
liberty to try these things on the running servers.
Anyway, can I get any more information out of the backtraces to help
resolve this issue with the new memory manager as well? Or do you need
more info for that?