Agents auto pausing after call disposition

[ccabrera]

Hello,

I´ve been having problems with a recent server install. These are the specs:

Dual Intel(R) Xeon(R) CPU E5645 @ 2.40GHz (24 cores)
60GB RAM
Dual SAS 15KRPM in RAID
Vicibox 4.01 with Vicidial version 2.8-403a
BUILD: 130510-1350
Installed on 2013-05-19
4 Digium cards for a total of 8 spans (4 PRI, 4 R2)
About 60 agents using Mozila Firefox 19 on Windows 7

My load as of now: 0.60, 0.99, 0.96. Very, very rarely server goes above 1.40 for load.

I have a random agent auto pausing after call disposition. Sometimes, an agent gets auto paused after every call, sometimes after the 3rd or 4th, some agents never get this error. It occurs very randomly, but it still does. When that happens, the pause code status just suddenly pops up (it happens immediately after agent double clicks on the call disposition code).

The server is acting as a VoIP gateway for another Asterisk system. That´s why it has 8 E1s but only about 60 agents.

I´ve tried the following:

- Increasing the Serverlimit and MaxClients in Apache to 350, KeepAlive timeout to 60 and MaxKeepAliveRequest to 1000
- Setting $conf_check_attempts=5 in agc/options.php
- Replacing the my.cnf for the my-bigvici.cnf
- Switching agent PCs
- Switching agent IPs
- I can confirm agents are NOT clicking back/forward on their browsers

As far as I´ve found by searching the forums, this error happens with one of these 3:

- Sudden server load spike (which in this case, all processes are default and there are no new crons added into it)
- Using older IE versions (all agents use Firefox)
- Network load (installation is using Cisco SB gigabit switches, and the network is certified)

There are some other posts I´ve found regarding this problem but I haven´t been able to solve it by going through those steps:

http://www.eflo.net/VICIDIALforum/viewtopic.php?t=24027
http://vicidial.org/VICIDIALforum/viewtopic.php?t=10942

The problems seems to be more noticeable when agents use the ALT-DIAL option, but it is still random.

I have seen bigger installations running smoothly in smaller servers, but this one has that little something which won´t let agents go for more than an hour without auto pausing. Server specs look too good for me. Any ideas on what could be causing this?

[ccabrera]

After some more testing, I can confirm that the PAUSE is triggering this code in agc/vicidial.php:

if ( (VLAStatuS == 'PAUSED') && (AutoDialWaiting == 1) )
{
if (PausENotifYCounTer > 10)
{
alert_box('Your session has been paused');
AutoDial_ReSume_PauSe('VDADpause',"CAUSA5");
PausENotifYCounTer=0;
}
else {PausENotifYCounTer++;}
}
else {PausENotifYCounTer=0;}

The thing is, I don´t know which part of Vicidial is triggering the pause. The Agent Interface is just reacting to whatever Vicidial says.

Any ideas?

[ccabrera]

Some more thoughts onto the matter:

- Asterisk is installed by RPM. Is there any room for improvement if I recompile it manually? (I know there is, but is it worth it?)

- Vicidial is acting as a PSTN gateway for another Asterisk, and the connection to the other server is established as a combination of IAX2 and SIP trunks. However, IAX2 is not using trunk=yes. In my lab testing, load decreases significantly after turning it on for high call load. Would you recommend doing this?

- After looking at the output of the top command, the top 2 CPU processes are these:

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3046 root 20 0 12.5g 64m 8012 S 107 0.1 549:30.81 asterisk
2745 mysql 20 0 6704m 411m 6820 S 36 0.8 176:33.86 mysqld

Under about half a dozen other servers I have access to (all of them have database + Asterisk on them), the CPU usage of the two is about the same. However, in this case Asterisk is going almost 3x the CPU usage as of MySQL. Has anyone seen this before on 1 server Vicidial installs?

Sorry to keep going and going about this topic. I´m doing my homework and all of these seem to be related, I´m just trying to see the light at the end of the tunnel to know I´m going in the right direction.

Regards,

[mflorell]

Recompiling Asterisk won't help much at all on Vicibox.

Using IAX trunking is fine if you are not going above about 30 concurrent channels on a single trunk, beyond that the packets can get too big.

Moving MySQL to a separate server is always a good idea if your system has a lot of load.

[PremBalwani]

Hi -
Were there any updates to this post, or problem? We are experiencing similar issue with Agent sessions automatically going into pause randomly. Server avg load doesn't tick above 10 at any point? Looks like for some reason Vicidial is thinking the agent is on pause. Does any kind of delay in updates to the database cause the problem.

Thanks,