Agents on random 3 way

[PierreDeux]

Hi everybody.
I'm running a multi-server setup of vicidial VERSION: 2.4-306a BUILD: 110310-0311 (vicibox install) and very often I see several agents (usually 2, but sometimes more) talking on a 3 way with the same customer. The problem is that they did not do anything to initiate a 3 way call. This happens on random servers with random agents and in different campaigns for many times per shift. Just boom and we have 2 agents with the same customer, boom - another 2. Most often these are outbound calls, but I'm not sure if they are the only ones.
Any ideas how to solve the issue are highly appreciated.

[williamconley]

Thanks for your vicidial version, very useful. How about the Vicibox Version as well (also important!).

1) Server Load when this happens? (or at least the number of Live Calls/Ringing Calls ... although I do prefer the Load figures from "uptime" or "htop")

2) Are agents sharing logins? (Very bad idea ...)

[PierreDeux]

Hi, William
1. Vicibox Redux i686 3.1.15 preload
2. it's completely different from time to time - this happens both when 90 calls and when 900
3. Of course every login is dedicated

[williamconley]

Thanks for your vicidial version, very useful. How about the Vicibox Version as well (also important!).

1) Server Load when this happens? (or at least the number of Live Calls/Ringing Calls ... although I do prefer the Load figures from "uptime" or "htop")

You did not mention your installation version. (important!)
You did not mention the server load. Although the number of calls was listed I did mention that I prefer output from uptime or htop. The server being overloaded would explain your problem.

Is this installation on Atom processors, virtual machines, or 16 processor monster boxes? (Or perhaps somewhere in between?)

Last but not least: How did you "cluster" your installation? What processes are running on which machines if you did this manually?

[PierreDeux]

Thanks for your vicidial version, very useful. How about the Vicibox Version as well (also important!).

1) Server Load when this happens? (or at least the number of Live Calls/Ringing Calls ... although I do prefer the Load figures from "uptime" or "htop")

You did not mention your installation version. (important!)
You did not mention the server load. Although the number of calls was listed I did mention that I prefer output from uptime or htop. The server being overloaded would explain your problem.

Is this installation on Atom processors, virtual machines, or 16 processor monster boxes? (Or perhaps somewhere in between?)

Last but not least: How did you "cluster" your installation? What processes are running on which machines if you did this manually?

It's been a while, but still trying to find where this issue comes from.
Load is fine (there are more servers in a cluser, but none of them is overloaded):
top - 16:57:06 up 93 days, 8:43, 2 users, load average: 0.28, 0.73, 0.79
top - 16:57:21 up 4 days, 10:36, 2 users, load average: 0.33, 0.18, 0.15

My asterisk servers are Dual 5520 Quad-Core Xeon 2.26GHz with HT & Turbo, 72GB RAM, 2 x 500GB Sata, RAID 1 (I know that it's too much for a regular asterisk server, but got them for a good price, so why not . Master DB server is:

Code: Select all

Dual 2.4GHz E5-2609 Intel Quad-Core Sandy Bridge Romley Xeon
96GB DDR3 Ram
240GB Intel Solid State Drive 520 series
240 GB Intel Solid State Drive 520 series
1000GB SATAIII
1000GB SATAIII
SSD RAID 1


So I believe it's definitely not any kind of overload issue.

I've updated my entire system to VERSION: 2.8-417a BUILD: 131019-0849 several months ago, all was installed with Vicibox 5.0.3 ISO (asterisk 1.8 ) and the problem is still there: multiple 3-ways happen pretty often. I don't see any logic in them:
1. they can initiate both at the beginning of the call and in a while
2. agents can be on any server in the cluser
3. agents can be logged in in different campaigns

I have no ideas where to look for the issue since Any suggestions?

[geoff3dmg]

I've seen this happen after the database tables have been locked for too long. Monitor your database with 'mtop' or similar.

[PierreDeux]

Thx for your suggestion, do you know what tables can cause such behavior?

[geoff3dmg]

vicidial_list usually.

[PierreDeux]

vicidial_list works fine for me, it had wait issues before, but I ended up converting it to innodb and now it's working much faster (yeah, I know that vicidial guys don't like innodb, but my experience tells me that it works better at least for my setup (5m leads, 110-120 agent online, dialing ratio varies from 2 to 7)). I had this 3way issues before converting list table (when it was innodb), so obviously it's something else, but vicidial_list.