asterisk users - May 2005 - high availibilty (heartbeats) - a good way to ensure automatic redundency?

being from a telecoms background, the thought of a single asterisk box
solution (even in a low production environment of say <10 phones)
worries me slightly!

starting from say a base of asterisk@home, you would have several
MySQL databases, in addition to numerous config files.

I have looked at high availiblity solutions, and from a hardware
monitoring point of view, its relitivly straight forward, you have 2
(identical?) boxes, each with 2 network interfaces.  One of the
network interface cards on each box has the same IP address, there is
another cable that is sending a heartbeat message between the two
boxes, heart beat fails the other box brings up automatically the
interface.

I have used HA on firewalls, and as the equipment is propriertary you
talk to the master side, which pushes config and state to the slave,
in the same way telephone exchanges run 1 micro-instruction behind the
other.  Obviously howver if you have a PRI on the box, it will lose
its calls, IP could be more resilient.

asterisk as it stands isn't geared up for this push of state, so
leaving that to one side there are a few obvious questions, but
firstly my assumptions.

MySQL has some sort of master/slave database system built in, so that
config is ok.  AMP self generates the dynamic config, so a cron job to
reload the slave every few minuites is possible to keep that part in
sync

to my questions (sorry its dragged slightly).  How can the astdb (the
one that you type show database at the asterisk cli) be kept in sync??

is this the right way to design a warm standby system or is there an
already established method.

The wiki suggests HA, but doesn't specify how.  Googling doesn't seem
to find anything (I have been trying on and off for a couple of days
now)


Thanks for any comments

On 11 May 2005, at 21:20, David John Walsh wrote:
> being from a telecoms background, the thought of a single asterisk box
> solution (even in a low production environment of say <10 phones)
> worries me slightly!
Sure, and so it should, but not too much.
There is a complex trade off here, and you as you add
hardware and complexity you soon hit diminishing returns.

You really need to establish what an acceptable downtime would be
before you start down the HA route.

We are in a small office, we buy solid hardware, keep it
in a decent environment and keep backups, documentation and spares.

Our strategy in case of hardware failure is :
     1) get the phone co to redirect calls to our mobile phones.
     2) get the spare out of the cupboard, load the last known
good backup
     3) swap systems.
     4) test
     5) cancel the redirect.
Total down-time of ~ 30 mins probably.

For a big telesales office, this would be unacceptable, but
for us, frankly, the internet is more indispensable than the phones
these days.


Tim.