vicidial.org

My thinking this might be acceptable for a 30-40 agent call center 1/2 on inbound campaigns
Dual 8core 4ghz AMD servers identical except one acts as a slave database server.

Scenario 1 -> slave goes down
The orphaned uses would just login using their credentaisl on the surviving server?


Scenario 2 -> master goes down
Reconfig slave as master
then have users login?

Is it that simple?
Am I missing any steps?
I've never run a cluster

That could work, we have a few clients that have set up redundancy in that way. You will need to either manually change the MySQL slave to master manually or write a script to change them automatically under certain circumstances.

I'd probably have a database/web server, two telephony servers, and one floating spare that acts as the slave database/web server but can be switched to a telephony role should a telephony server fail. This way you spread out your resources so you aren't over-loading the machine and the impact of asterisk crashing (which is mostly what will happen on any given week) only effects half your call center. You also get the benefit of a more graceful fail-over should Asterisk or a Telephony Server crash if you use Phone Alias for the agent's phones.

Here's what I would look to do for redundancy:
1) 3-server cluster + 1 spare/slave (1 Db/Web, 2 Telephony)
2) Redundant power supplies, going to two different UPS, each UPS on a different power circuit/leg in the building (you'll need an electrician here)
3) RAID-1 with some sort of monitoring so you know when it's failed (RAID cards beep, Linux RAID can send emails, etc)
4) Some useful scripts to make your fail-over quicker and more foolproof

You should also test your fail-over strategy to make sure it works. All the planning in the world doesn't matter if you never tested it. Chances are something will go bump you didn't think of.