[asterisk-users] fail-over server

[Jonathan Thurman]

On Tue, Feb 8, 2011 at 8:07 AM, Vieri <rentorbuy at yahoo.com> wrote:
> Suppose you have 2 identical Asterisk servers and 1 alias IP address that you assign to either one, according to system failures, etc.
> Also suppose that all SIP clients register requests go to the alias IP address.

This is a typical setup for two node HA.  Just be careful when
clustering only two servers.

> Imagine server1 fails and server2 gets the alias IP address.
> Correct me if I'm wrong but I would have to wait at least 60 seconds before
> most SIP clients re-register to server2 and that server2 knows that they are
> actually "on-line" so calls can be routed to them.

It depends on your configuration.  If you use Asterisk Realtime to
store SIP registrations, then the database will contain information on
how to contact the device (fullcontact, ipaddr, and port fields).
Then on a failover, Asterisk will do a lookup for the peer in the
database, find the needed information and dial the device.

Of course any registrations that happen before being written right
before the server fails may not work.  Also make sure to use the
latest version of Asterisk as there was a bug where fullcontact wasn't
saved correctly.

> How can I minimize this time lapse? Can Asterisk "notify" all SIP
> clients in its sip.conf that they need to acknowledge being on-line
> or not (thus forcing re-registration in my scenario)?

In the above scenario, I can kill Asterisk, start it again, and place
a call from two devices that have not registered again.  So, the best
timeout is your dead time detection and failover startup time.

-Jonathan