Pamela.Schmidt-1 at ksc.nasa.gov
Fri Oct 4 20:21:22 MDT 2002
Is there a way to modify ipfail so that it performs a "heartbeat stop" when
it detects that the ping node is down?
----- Original Message -----
From: "Kevin Dwyer" <kevin at pheared.net>
To: <linux-ha at muc.de>
Sent: Friday, October 04, 2002 9:58 AM
Subject: Re: Ipfail.c question
> On Fri, 4 Oct 2002, Schmidt-1, Pamela grunted something like:
> > Can someone explain how ipfail.c works.
> > I am trying to test the ipfail.c setup on my cluster and when the ping
> > node is disconnected from the server; the server recognizes that the
> > ping node is down, the secondary server recognizes that the ping node is
> > down but the resources are never taken over from the primary to the
> > secondary or vice versa.
> If a ping node actually goes down, you've got a situation that you can't
> deal with. The point at which ipfail becomes useful is if the ping nodes
> stay up, but one of the machines suddenly can't reach one. That way, the
> two machines can make a decision about who is still up (one will see the
> ping node, one will not) and do something about it. Hopefully ;)
> > When ipfail detects a node down, does it send a "heartbeat stop"
> > command to signal the secondary that it should take over the services?
> > How does hb_standby work? and what is the nice_failover feature do in
> > the ha.cf file?
> IIRC, ipfail sends a T_ASKRESOURCES message just like hb_standby, so it
> won't cause the other node to shutdown, just give up the resources.
> nice_failover is something you want to turn on if you are going to use
> ipfail. It makes it so heartbeat won't failback to the 'primary' since as
> far as it knows, everything is still cool.
> /* kevin at pheared.net http://devel.pheared.net/ */
> /* Network Security Engineer http://pheared.net/~kevin */
> /* Sabotage will set us free. Throw a rock in the machine. */
> /* >++++++++++[<++++++++++>-]<.+++++.----.[-]++++++++++. */
More information about the Linux-HA