[Asterisk-Users] High(er) availability

[Rich Adamson]

> > >>Which app do you use for monitoring the primary box and if it fails
> > >>taking over the IP address by the backup one? I haven't found a suitable
> > >>(active-active) app so far.
> > > 
> > > Thinking of using heartbeat or something.
> > 
> > VRRP, Virtual Redundancy Router Protocol, an option?
> > 
> 
> Cisco claims that VRRP falls under one of their patents, so it could
> become an expensive option.  There are several options out there at this
> point though that may be able to handle the needs for pre-empting the IP
> address.  
> 
> About 1 year ago the OpenBSD project wrote a patent-free alternative for
> VRRP called CARP.  It allows for sharing of and automatic failover on an
> IP address.  I have used it to build redundant firewalls that don't lose
> any state information when the connection drops. CARP is of course built
> into OpenBSD however I did find what looks to be a userland
> implementation for Linux.  See www.ucarp.org for more information.
> 
> There are other possible solutions as well, unfortunately I have not
> used any of these solutions they are just from brief google search.  LVS
> (Linux Virtual Server) mentions "VoIP services" however I do not know if
> Asterisk would run in a cluster environment.  There are also several
> sites that deal with high availibity from linux, the first one I noticed
> that looked like it had some really valuable information is
> www.linux-ha.org.
> 
> Unfortunately this is all the easy part.  The difficult part will be
> getting Asterisk to handle the failover gracefully.  You probably don't
> want to lose all the SIP registration data and I have no idea if it will
> be possible to prevent you from losing the calls.  You haven't named
> that as one of your goals, but it is always something to think about.

There is a lot of interest from lots of folks in how one handles
failure overs, etc. I've got to believe that a fair number would be
very happy with a primary-secondary arrangement where calls that were
in-flight might be dropped, but recovery in terms of displacing the
failed * box happens within several seconds (or possibly even a minute).
Reading between the lines from the original poster's question, it would
sound like that would be an acceptable aproach.

If the failover time for a primary-secondary approach was short,
keeping the registration data and other somewhat dynamic data in
sync between boxes is relatively easy. It would seem the only remaining
issues involve MAC addresses, redundant physical pstn-type interfaces
(and probably something as simple as a relay flopping T1's or fxo's
over).

Has anyone truly implemented such a pri-sec failure, and if so, care
to offer up some specific configuration data that is usable?