Page 1 of 1

Phone Login Load Balancing Call to Agent

PostPosted: Wed Dec 12, 2018 2:27 pm
by adymeblack
As stated in my signature, we have a 12 dialer server setup and we are utilizing phone login load balancing across them with Jitsi and a provisioning script. At the beginning it was working great, but now we seem to have an issue popping up where the agent will go to login, but they will never get the call.

I'm not seeing anything strange in the asterisk logs, but i'm not sure if it's something with the software.

Anyone have a direction for me to look in to get this figured out?

Thanks.

Re: Phone Login Load Balancing Call to Agent

PostPosted: Wed Dec 12, 2018 2:33 pm
by williamconley
Jitsi and a provisioning script

Since this isn't part of Vicidial ... I'd have to say you're asking the wrong people.

Vicidial does have a load balancing system that has nothing to do with jitsi, built in to Vicidial. If you were using that, which is covered in the Vicidial Manager's Manual, we could be of some assistance. But you're using nonOEM parts on this car and one of those parts stopped working. Not a lot the manufacturer can do for you, and those with competing parts companies will generally just say "you shoulda bought ours, or the OEM, everyone else sucks". 8-)

Re: Phone Login Load Balancing Call to Agent

PostPosted: Wed Dec 12, 2018 2:41 pm
by adymeblack
The actual load balancing is handled by Vicidial, the provisioning script for Jitsi just registers the softphone to all of the dialers (passes the IP's, usernames, passwords, etc). Which it does look like all of the servers get connected and registered to the vicidial servers. Jitsi is the only one we were able to find that we could provision to 12 dialers with a great (free) price.

Apologies for the confusion on that.

Re: Phone Login Load Balancing Call to Agent

PostPosted: Wed Dec 12, 2018 3:34 pm
by williamconley
The agent screen login initiates the process. Begin tracing there and follow the request/response until you get to "dial". Once you see how it works from beginning to end, you can begin tracing a "fail" and see if you can find where it died (now that you know how it *should* work).

May be a good idea to test this on a sandbox server first so you can indiscriminately add debug code and test the heck out of it without (a) danger to the live system and (b) any other traffic. VMware (vSphere), Virtualbox, etc make great "single agent sandboxes" for development purposes.

Agent screen "submit" is where you start.

System Settings: Agent Screen Debug Logging (this may not actually start until after the submit button, though, lol, so you may end up using the console and network activity viewers to watch and debug)