HA Asterisk Installation (Help Please)


I am knee deep into an install and have come into some problems. This is what I have:

RanchNetworks RN-40 - For HA IP failover

2 Intel Blade servers - both with dual xeon 3.4ghz processors, and 4gigs RAM, Raid 1 mirrored SAS 73gig HDD

I will be using a SIP Gateway for PSTN connections. Either Vega400 or Patton SmartNode.

Here are my issues:

How do I keep the databases on both servers in sync and HA?

Using ISCSI and a small SAN keep the database and FreePBX on the
SAN and run asterisk on the blades with all .conf files pointed to the

Using ISCSI and a small SAN run everthing on the SAN and use the
blades as nodes with no HDDs. One fails the second picksup using the
exact same config as the first.

How will I keep all phones registered if the primary server fails? Is phone registry kept on the HDD or in the RAM of the server?

Any info would be much appreciated.

Thank you,
Randy Quick


I can’t really comment on your first question, other than to say option 2 sounds better to me from an asterisk standpoint. Ideally, you’d want a linux distro that can run across multiple boxes but appear to the software to be a single “entity”. Option 2 sounds more like this than option 1, but they’re both viable.

As for the registrations, those are kept in the AstDB, which is a variant of the Berkeley DB - I BELIEVE that is actually stored in memory except for when the system shuts down, as asterisk failing will result in your DB values being lost. I am not 100% sure on this, so somebody else with more experience may have a different answer - defer to them.

You may look at using the Realtime architecture to store your dialplan and SIP settings in a Mysql or ODBC-compliant database. This might be slightly more ‘realtime’ in terms of a failover and may be your best option. I can’t tell you more than that as I simply do not know, but PLEASE post back any results or progress you make as our next major Asterisk project will involve much the same thing - making our systems redundant and offering some sort of failover.

Thanks, and best of luck.