Dialer Clogging

[shinigami]

We've used the new version of ViciDial (VERSION: 2.9-441a BUILD: 140612-162) with PBX or Asterisk server version of 2.6.32-26-pve (root@lola) (gcc version 4.7.2 (Debian 4.7.2-5)). My bad, I'm not that really familiar where to locate the exact version of Linux servers. Anyways, we've tested those servers trying to dial a 100+ seaters. Observing the dialer with only 80+ representatives, it ran smoothly and database load is so low but when all those 100+ reps logged in database or dialer clogging occurred.

Message errors such as the following happened on the Agents' dialer interface (agc/vicidial.php)--

[1] "Another live agent session was open using your user ID. It has been disabled. Click OK to continue to the agent screen."
[2] "Customer has hung up: SIP/Carrier-0000491d." - even the reps are still engaged on a call.

On the dialer side, sudden "Dead" status happened and/or sudden "Pause" status even the reps are still talking to their customers. Sometimes, hard to modify the campaigns when the said clogging experienced.

Please give advise masters. Thanks!

[dspaan]

Of how many servers does your cluster setup consist?

[shinigami]

Hi dspaan,

We got the following components on our system (I don't know if this is the cluster):

Database Server -> 12 Cores, 1TB SSD (HDD1) and 250GB SSD (HDD2), 126GB Memory (stand-alone server)
Asterisk 4, 5 -> 4 Cores, 1TB HDD, 8GB Memory (on Proxmox)
Asterisk 6, 7 -> 4 Cores, 1TB HDD, 8GB Memory (on Proxmox)
Asterisk 1, 8 and Agent Web 1, 2 -> 4 Cores, 1TB HDD, 8GB Memory (on Proxmox)
Asterisk 2, Agent Web 3, 4, 5 -> 4 Cores, 1TB HDD, 8GB Memory (on Proxmox)
Asterisk 3 and Recording Server - 4 Cores, 3TB HDD (HDD1) and 3TB HDD (HDD2), 8GB Memory (on Proxmox)

To sum up everything - we got 1 database server, 8 asterisk server, 5 agent web and 1 recording server...

[dspaan]

Since this is only your 3rd post on the forums how did you make such a big setup with which experience? I suggest testing it with 30 agents on a single server, from there you can expand to 2 extra dialers and you should be able to do 80 agents without too many problems.

8 asterisk servers seems like overkill....
5 agent web?

Are we talking about physical servers here? What do you mean with components on a 'system'?

[shinigami]

We've used ViciDial for more than 4 years in our center with the same agent capacity of more than a hundred. We're currently using "VERSION: 2.2.1-237 BUILD: 100510-2015" of the ViciDial with single database server and 7 asterisk servers with no clogging except when showing all the list which is normal for millions of leads on the system. Agent web means the agent dialer interface (agc/vicidial.php).

Yes sir, I honestly don't know what's cluster so I've stated all the physical components of our servers. Those stated above were all new servers aside from the servers we're currently using with 2.2.1 version.

[shinigami]

No problem experienced on asterisk servers, just on database which clogging suddenly occurred when 100+ agents are logged in.

[dspaan]

Since an asterisk server can do generally 25 agents why would you use 8 of them and also 5 apache servers? If you have so many servers i can imagine it's more work for the database server to keep up with them all. Are you using voicetimers? Are all servers using the same NTP source?

[shinigami]

We've already allocated 25 agents on each asterisk servers, I mean 25 SIP phone every server. Yes sir, 5 apache servers for agent dialer interface. What's voicetimers sir, my bad I'm not familiar with the terminology. Honestly, I'm basically on the leads management of our center and don't have enough knowledge on Linux world.

On the NTP source, we're using time.nist.gov or ntp.ubuntu.com

[dspaan]

When you have such a big cluster it's important all machines are in time sync with each other. The database should be the source for all the other servers. So the database syncs with time.nist.gov and the others sync with the database server. I never used a voicetimer either but i've heard it can help with timing issues:
viewtopic.php?f=4&t=33679

You first have to check if that's even a problem, go to the Reports page and press the + next to SERVER (bottom) and see if all times match. Who is the person managing your cluster? He should post here.

[shinigami]

There's no problem on the time of each servers sir, all were synchronized with the DB Time. We're checking the feature you've stated every time synchronization problem experienced. I will ask him to post here when to help me on clusters thingy.

It is just weird and confusing sir that on version 2.2 of the ViciDial, we haven't experienced such issue and problem of clogging even without the use of a voicetimer.

[dspaan]

I would suggest contacting vicidial support, you have such a big setup, a lot of stuff could go wrong here. There is no easy answer i can give you here.