Best Failover solution

I send you a private message with my contact information. My firm has a tier 1 data center in Cleveland Ohio that would be ideal for your redundant system.

We can setup the configuration sync between the two trixbox servers.

Depending on how your provider handles registrations the inbound calling failover may require manual intervention.

After a lot of thought, we have decided to create a primary and secondary server, basically mirror images of each other.
Implementing database replication should keep the two mysql databases in sync.
Should be able to keep the two voicemail systems sync using rsync for voicemail files, the same for Announcement files.

Any thoughts? Gotcha's?

Two areas of comments/considerations, from having done this project numerous times:

Before you go too far down any sort of road here, I'd recommend taking a close look at what business problem your customer is trying to solve with a "failover" solution. You need to weigh the risks of downtime against installation and maintenance complexity of the "clustered" environment.

No matter which way you go, you're adding complexity to what at this point is designed to be a single box standalone environment. Your lowest impact (and slowest recovery time) method is going to be to use a Mondo Rescue type solution. You note the obvious deficiencies there.

Beyond that, you need to cleanly delineate what kind of risk are you trying to mitigate with a failover solution. Are you trying to be able to survive a single box failure? Are you trying to survive a whole site failure? Would it be of any business use to have a geographically dispersed telephony cluster if your office was hit with a major natural disaster? Does single box with decent live hardware monitoring mitigate enough risk to make the complexity of cluster management unappealing? Are you carrier and network redundant? If not, does geographic redundancy buy you much?

Now to the technical components:

I'd suggest you do some additional investigation on the architecture of Asterisk / TB to understand what you're getting as you're using mysql clustering or rsync /csync. You need to understand exactly what is contained in the mysql database from a Trixbox perspective and what isn't. The solution you describe, in itself, will not provide you failover in anything more than a data preservation sense. In fact, it's not going to give you much more than having a clean box with a current FreePBX Backup/Restore file is going to.

Understand that Trixbox stores static configurations in the mysql database, but does not store SIP registration information there. You're going to lose all that data when the failover happens. Additionally, if you have the primary and secondary boxes, how are you going to point the phones to the secondary server on failover? Are you going to use a VIP? If so, how do you hold that VIP role? Are you going to force Asterisk reload on failover? You're going to have to set a very low re-registration interval on your endpoints to have anything resembling sub minute failover (especially for inbound calling / SIP trunking) in this scenario.

I would highly recommend you consider Scott's offer - he gets great marks and hosting can produce great results if you design properly. You should check out with Fonality does with TB pro. If those meet your requirements, I don't see why you wouldn't go that route. It'd probably be easier than what you're trying to do.

I can tell you from A LOT of experience on this, if you're looking for on-site cluster with near instantaneous failover, your solution isn't going to be adequate - you'll pay dearly in lost hair and increased alcohol consumption. You'll need to consider Linux HA (Heartbeat + DRBD + a LOT of customization and solid network design) or a form of peering (with Asterisk Realtime) to get you there. The Linux HA has the advantage of being relatively well understood and at least somewhat manageable. The DUNDi Peering / ART solution scales beautifully and can survive lots of kinds of failures, but isn't compatible with Trixbox in the least at this point. If you're looking for help getting over the hump on this, let me know. I've installed mission critical accounts using Linux HA with 2+ years of uptime - with the cluster PBX surviving network core outages and hardware failures. All this requires a lot of careful planning to make worthwhile.

Josh

virtualizing on Linux or on a Windows OS?

If Linux, which flavor?

Thanks!

John

Once Josh has it covered.
We have used the VM ESX the longest and it works great.
OpenVZ is my next choice (just cuz I like the CMD line)

But even the freebie ESXi works fine (I have sen upto 30 extens / 20,000 mins a month on a beefy box.
most host boxes are Centos and FC (I just redid and older Centos 4.7 just cuz I had a spare to move the VM's around on).