[Pacemaker] Upgrade of Pacemaker on CentOS 6.4 to 1.1.10 - Delay RA missing ...

[Andrew Beekhof]

On 7 Nov 2013, at 9:30 pm, Robert H. <pacemaker at elconas.de> wrote:

>> This does a reasonable job of explaining:
>>   http://blog.clusterlabs.org/blog/2013/pacemaker-and-rhel-6-dot-4/
> 
> I see, thanks for the hint ... small step for man, huge step for mankind .. (or something like this :))
> 
>> I would be interested to know what for.
> 
> We have a setup of load balancing via OSPF (ECMP). We must make sure that some service is stopped not before 30 seconds (or better OSPF heartbeat timeout) after the stop of the OSPF deamon (quagga). This allows seamless failover without any downtime.

Congratulations!  This is the first valid sounding use of Delay I've ever heard :)

You should file a bug.
You're on CentOS, so it may not carry much weight for now, but it does provide a place for other people to "pile on".

I'd suggest the wget option for now.

> 
> - OSPF session active and traffic is routed to our load balancer
> - crm node standby
> - now OSPFD should be stopped
> (now ~<OSPF dead detection timeout> seconds, traffic is still routed to our node, which we want to service, in our case ~10sec)
> - we wait 30 seconds (with Delay)
> (after ~<OSPF dead detection timeout> seconds, no more traffic to our "standby node")
> - now the load balanced service should be stopped
> 
> For this we need Delay :)
> 
> Any other suggestion (besides writing a custom OSPF resource agent that waits itself for the node to disappear on the router) ?
> 
> Regards,
> Robert
> 
> -- 
> --
> Robert
> 
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org