Vicidial on Callweaver ????

[explidous]

Did anyone try or entertain the idea of running Vicidial on Callweaver (previously called OpenPBX)?

I am aware that it would require a number of changes, however since Asterisk is making a number of problems itself and it looks like they addressed a number of our hot performance issues.

We are entertaining to set up a test system with that. Anyone interested in the results?

Frank

[mflorell]

Of course we are interested in the results!

CallWeaver is now a bit different from Asterisk, but if the Manager API has not changed significantly then it should work just fine.

I do believe there are at least 3 different conferencing engines available with Call Weaver right now, not sure if meetme is still the default or not.

[aster1]

Yah reading project's wiki really impresses on changes they made . According to asterisk-user list digium is discontinuing development of 1.2 branch from August 1 and 1.4 is still buggy . Next release of vicidial can be made to work on callweaver ?

[explidous]

Meetme is not the default from what I read... we started a test install... lets see where it goes...

We were wondering that as well especially since I am again into trying to get those deadlocks resolved... some of them being on the call_list_locks, one of the areas they reworked to eliminate that!

Other things that are broken for ever make me wonder as well....

[mflorell]

On a somewhat unrelated subject, could you post that fix you made for the WaitForSilence app in Asterisk 1.2?

[explidous]

It is in your bugtracker for a while...

Bug 94

File: app_waitforsilence.c
http://www.eflo.net/VICIDIALmantis/file ... 6&type=bug

[mflorell]

Here it is.
http://www.eflo.net/VICIDIALmantis/view.php?id=94

Have you tested this on 1.2.18?

Thanks!

[explidous]

Not extensively since we had problems with 1.2.18 skipping Hangups under load... which we partially fixed by changing the hangup fastAGI to perform log and hangup in one step.
however with 8T1s in one box (7 dialing 1 agent) skipping hangup and deadlocks became really annoying in 1.2.18, 1.2.17 was doing a bit better on it, at least on the skipping of the hangups.
BTW
1.) we had to stop putting the wait in the hangup extension because then nothing but the wait would be executed
2.) we put the wait in the dial between the AGI(calllog) and the real dial() that works somewhat wait(0.15) seemed to elimidate a lot of problems. longer waiting did not make things much better, but we still had to limit the number of channels to use... Most likely because of the funky release from their carrier...

PS answering your question.... yes it compiles and works... your milage might vary

[mflorell]

What kind of loadavg are you seeing on the system with 8 T1s?

Any reason you didn't just add another server?

[explidous]

The system is a SMP dual XEON (non core duo) Hyperthreaded system, 4GB Ram, SCSI disks, hardware RAID.

I see a load of 2-3 till the first call gets transfered to the remote agent at which time the load spikes, from 2 or 3 remote agents connected we see deadlock warnings on 1.2.17 as well as 1.2.18 on 1.2.18 they seem more however. unless we see a deadlock warning the percentile load still stays in the 40-70% range.

We took MySQL off the box but it did not make much of a difference...
Box is running a 8 port Sangoma card with echo cancel.
Our customer has a number of ds3s to fill and would appreciate a higher density then 4 T1s per box.

From what we did before I had identified at least two specific deadlocks (one in AMI and one in channel handling) that were just caused by bad programming in asterisk, it was quite some work to get these worked out but by now the code has changed so much in there that I would have to dig again... One of my suggestion for the channel was taken up at the time but the AMI was never fixed in the official release.
I think that some new problems have sneaked in.

[mcargile]

First off in follow up to franks message the system he is talking about is running 2.0.2 Vicidial (though we plan on upgrading to 2.0.3 soon.

Now getting back to the topic at hand. What all do we need to make sure is working right before even trying to use Vicidial on Callweaver.
* I know there is probably some stuff with the CLI. Currently when I do a show channels concise with a single sip channel open this is the output:
SIP/22-6336:default:500:1:Up:Playback:demo-abouttotry:22::3:14:(None)
The : need to be ! for Vicidial to work.
* I know in the dial plan I will have to replace the MeetMes with NConferences, but that goes without saying.
* AGI is now OGI but there is a conversion script to change AGI to OGI.
* AMI I believe still works the same though I have not done much research on it. I do know they moved AMI out of the main process though into what they call an agent. Here is a document on that (which I have not read yet ):
http://trac.openpbx.org/cgi-bin/trac.cg ... format=raw

[mflorell]

AMI will be the biggest hurdle if they have changed significant portions of it.

The CLI concise patch (!) might actually work unaltered if the cli.c still exists. That is a patch that they should integrate into CallWeaver since it is a bug for them as well.

[mcargile]

the cli.c patch applied fine.

[mcargile]

So I submitted the cli patch to callweaver to try and get them to take it on as well and am getting the same general response of "why not use a ',' " Here is the link:
http://www.callweaver.org/ticket/76
Matt if you could weigh in on this it probably will have more merit than me.

[mcargile]

Callweaver accepted the cli patch the way it is and placed it in trunk so it will be in their first official release.

[mflorell]

Cool!

Thanks for doing that, I'm very interested in hearing more about the possibility of cross-PBX compatibility, it brings some interesting options for the project.

[mcargile]

Indeed.

[mcargile]

Okay I was finally able to get a straight answer from someone in #callweaver on freenode about porting to Callweaver. The conversion to Callweaver should be completely possible for the time being. They are not sure if they will be using AMI in the near future. They may switch it to something else, but as of RC4 it is still compatible.

The big thing though in both the AMI and AGI is that Callweaver is much more case sensitive than Asterisk so I might run into some issues on that. And there are some name changes of applications.

[i_magic]

Michael, Didja have luck with running vici on callweaver?
-TJ

[mcargile]

Nope. I am more interested in getting 1.6 Asterisk functional with Vicidial. They are planning on having a version of app_meetme that does not rely on a zaptel timer ready by the 1.6.3 release (when ever that may be). So there really is no point.