vicidial.org

I have recently Deployed a Vicibox 3.1.15 system with 3 servers.. 25 seats total for now, with a max of about 100 eventually....
I am running a fronter/closer setup.. with 16 seats in one room all Fronters and an 8 seat room which will eventually be all closers... At the moment I am running 3 closers and 3 fronters (up to 19 fronters in that campaign) in that room until a remodeling project is completed so I can add an additional 24 seats.

My database/web server is a dual 2.0 ghz Quad core Xeon Asus 1U server with 24g of ram. Opensuse 11.3 Installed via Vicibox X64 3.1.15. The hard drives are 4x300gb Seagate 15k Cheetah SAS Drives running Raid 10 on a LSI 9640 MegaRaid controller. I've been running mtop to check for locked tables and cannot find any. I have also setup a slave database to run all the reports. Load average on my database server is less than 10% at this point.

My telephony/agent servers are HP DL360 G3 servers with 4 gb of ram and 36gb U320 SCSI drives in raid 1 with 2 dual core 32 bit 3.2Ghz processors. Opensuse 11.3 Installed via Vicibox 32bit 3.1.15. Each telephony server has a T1 card wired to channel banks using DAHDI to provide analog extensions for each agent using custom entries in the entensions.conf and defining the phones as external (10001-10024 dialed routed to a dahdi channel). Each telephony server has a max of 40 outbound trunks and 24 users. System load average is about 30% on these servers.. Call quality thus far has not been an issue. I have 2 servers setup but at the moment only one is handling agents.

Upon expanding the system next week I will be adding 2 additional telephony servers that will only handle overflow calling and adding a 2nd 24port T1 channel bank on my other telephony/agent server. I plan on setting them with a max of 80 outbound trunks per server and no agents. This is the max I was able to achieve on these servers with my 2nd telephony/agent server with no agents logged in before the CPU usage approached 45%.

All servers were installed using the latest SVN build. All networking is via 1gb switches no more than 3 deep between to the end stations and the servers. Ping times remain <1ms during calling to all servers.

VERSION: 2.4-357a
BUILD: 120125-2107
From my Admin Screen:
Version: 2.4b0.5
DB Schema Version: 1312
DB Schema Update Date: 2012-04-02 20:30:41

2 campaigns are in use. One for outbound and one for my closers. Fronters are using the quick transfer button to transfer calls to the default inbound transfer group.

I currently have about 700,000 numbers TOTAL in the system split into 20k groups (by state) with one group containing 170k (multistate file).

I am having the following issues...

Issue #1:
Occasionally the information entered by the fronter in my custom fields is not coming up on the closers screen and all of the values are set to their defaults. This is causing my closers some issue but we can work with it. Altho I am worried as I expand the system the frequency of this will increase. I have monitored the system while its running and I do not see any tables locking when this happens, at least using mtop.

Issue #2:
Occasionally when a call is transferred to the inbound closer group the agent is not getting any audio. I tracked it today and found that it was only happening on one station. I rebooted all my servers at lunch and nothing changed. We moved the agent to another machine to see if it resolves itself but it is still too early to tell. Is it possible a problem in the browser session could cause this? I did note that the machine was not shutdown from the day before and the user used the login again option and changed the username. I am wondering if it has something to do with reusing and old session id...

Any help would be greatly appreciated. I need to get all the bugs worked out before I roll this system out to all 100 seats at this office. Sorry for the long message but I know the more details the better.

-Andrew

Update: For anyone that may have this issue or a similar issue.. It seems to have be a agent client issue.. After looking at the logs I see that the calls that were coming to the closing agent with no audio where still qued into the hold que. The agents client session had been running for over 2 days and something was glitched. We are using Xubunutu with chrome in Kiosk Mode (fullscreen locked)... I added a nightly shutdown routine to the crontab so any computer left on shuts down at a set time and yesterday we had absolutly no problems with our vicidial system.

I am wondering if there is a memory leak in Chrome or the client app.. Ether way shutting down the systems nightly seems to have fixed it. You could also close and reopen the browser but since I am running my webclients in kiosk mode there is no control to exit the webbrowser available to the agent... (alt-f4 still works tho)