Still on it, but I have some servers deployed without rtp-timer-name set to soft ( blocking read is something I'd like to avoid, but needs to find out what changed in FS code and also find out the impact on our users ), and so far they have load avg divided by 10 with the same amount of calls and the same trafic as those with rtp-timer-name to soft. Had no issue with 1.4.12 as said previously, but 1.4.14 has some interesting fixes that justify staying in production. Le 08/12/2014 14:33, Dmitry Sytchev a écrit :
So, have you solved your problem? Did you change rtp-timer-name? After we moved to 64-bit system, 1.4 branch shows huge LA spikes up to 35 on 4-core (or 4 single core virtual processor) with only 200-300 concurrent channels and very low CPS (1-5). On the same call load LA of 1.2 server is significantly lower. Setting 1.4 64bit to normal priority with -np helps to lower LA, but not completely solves a problem. It shows very big values in "ni" top fields (20-30%), but changing nice value of FS process to zero or positive value doesn't help.
2014-12-09 2:23 GMT+04:00 Tristan Mahé <t.mahe@b-and-c.net <mailto:t.mahe@b-and-c.net>>:
Hi everyone,
Just a little followup, this is caused by <param name="rtp-timer-name" value="soft"/> in a sip_profile. Happens on baremetal servers as well as Xen guest.
Still investigating on the exact root of this behaviour change, but at least now we know what is causing the trouble.
Best,
Tristan.
Le 05/12/2014 19:46, Tristan Mahé a écrit : > Hi guys, > > While trying to find out the root cause of this, I just want to know if > someone is also seeing a huge cpu load with freeswitch 1.4.14, which was > not present in 1.4.12. > > OS is debian wheezy, packages are official ones. > > Same config, same trafic pattern, just an upgrade from 1.4.12 ( no other > packages updated ). > > We were previously seeing load around 0.5, with pikes up to around 2. > > We now are seeing a constant load of around 2, and pikes up to 20... > > If someone is also seeing this, that would be interesting to exchange, > helping find out what is causing this ! > > I'll file a jira once I find the root of this problem ( and no, this is > production servers, we can't put master there, we need stable ). > > Best, > > > > _________________________________________________________________________ > Professional FreeSWITCH Consulting Services: > consulting@freeswitch.org <mailto:consulting@freeswitch.org> > http://www.freeswitchsolutions.com > > Official FreeSWITCH Sites > http://www.freeswitch.org > http://confluence.freeswitch.org > http://www.cluecon.com > > FreeSWITCH-users mailing list > FreeSWITCH-users@lists.freeswitch.org <mailto:FreeSWITCH-users@lists.freeswitch.org> > http://lists.freeswitch.org/mailman/listinfo/freeswitch-users > UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users > http://www.freeswitch.org >
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org <mailto:consulting@freeswitch.org> http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org <mailto:FreeSWITCH-users@lists.freeswitch.org> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Best regards,
Dmitry Sytchev, IT Engineer
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
Official FreeSWITCH Sites http://www.freeswitch.org http://confluence.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org