VD Stops Dialling When Many Agents Log Off

[duncanb]

Hi All,

I have not been able to find a resolution for this by searching the forums, Mantis or Google, and I'm now completely stumped. Any assistance will be hugely appreciated.

The problem:
Around times when there's a shift change, or when lots of people log off (to go into a new campaign perhaps), ViciDial just stops dialling calls for the remaining agents. There are no errors I can see in the Asterisk, system or ViciDial logs. When I look on the Asterisk console, nothing is happening. There are diallable leads in the hopper, and the dropped call ratio is within limits. Load very low on all servers (incl MySQL server). After 5-15 minutes, ViciDial just seems to wake up again and everything returns to normal.

The forum wouldn't let me post a direct URL as a new user, but a screenshot of the realtime report while we experience this problem can be found at a site called postimage (dot) org. and whack a "(forwardslash) image (forwardslash) s7uta05on" on the end of that. If you're seeing this my sneakiness worked.

Complete setup includes a PBX server (where all the agent extensions are registered), and 2 ViciDial servers in load balanced mode. We have a stand-alone MySQL server. All campaigns are on ratio dial. ViciDial generally runs with 50-60 agents. Load average with this amount of agents is 0.25 - 0.30. When we have this problem, the load average drops to basically zero. We have load balanced ViciDial servers, but we experienced this problem when we were still only on 1 VD server. Agents registered on PBX server connecting to VD servers via SIP/alaw, outbound calls made directly from VD servers to SIP/g729 trunk. We have no shifts defined in ViciDIal.

ViciDial version:
VERSION: 2.4-338a
BUILD: 110923-2043

Asterisk version on VD servers: 1.4.24.1
Asterisk version on PBX server: 1.6.2.23

OS:
SUSE Linux Enterprise Server 11 (x86_64)
VERSION = 11
PATCHLEVEL = 1

Perl: v5.10.0 built for x86_64-linux-thread-multi
MySQL: 5.0.67 SUSE MySQL RPM
PHP: 5.2.6

[williamconley]

Please post your Installer version (or a link to your scratch install instructions if you installed that way).

You will find on your system that "screen -list" will show many screens running. "screen -r XXXXXX" with one of those running screens will allow you to view them (one at a time). "Control A" then "D" will exit the screen again (without damage!). Theses screens also generate output in the /var/log/astguiclient folder. One of the things visible in these screens is the calculation of agents/calls and "how many should I make for this campaign".

You can also check your auto_calls table to see if there are calls in there that are not "real" (ie: asterisk -rx "show channels" shows empty, but auto_calls has 10 calls in it ...).

Happy Hunting

[duncanb]

Thanks for the response William. It was the auto_calls table. I had made it into a memory table to try and reduce disk load on the server, and it just wasn't big enough to hold all the data.

[williamconley]

good post back. although i'm not sure i understand. when everyone logs off ... the auto calls table should be just about empty. and it should hold almost 100k records when full (auto calls should never be that full anyway?)

[duncanb]

There was some limit affecting the allowed size of the memory table. There's a Vicidial process that runs every hour and cleans out the auto calls table. On a busy day the system would hit the limit before the process ran. As soon as the cleanup ran on the hour, calls would start flowing again.

[williamconley]

Change the cleaner to run every 5 minutes.

Or (better advice), stop tweaking Vicidial like that and stick with settings in the GUI.