[Pacemaker] Preventing resource from becoming inactive

[David Mohr]

Hi!

On Tue, 10 Aug 2010 07:56:47 +0200, Torsten Bronger
<bronger at physik.rwth-aachen.de> wrote:
> Hallöchen!
> 
> Sometimes Pacemaker just switches off my Lighttpd.  It becomes
> inactive and is never reanimated. Only restarting the Heartbeat
> service helps.  How can I tell Pacemaker to retry to restart
> lighttpd without ever giving up?

I'm a newbie with pacemaker, so take my advice with that in mind.
 
> Since I use Lighty as a load balancer, it is tied with my public
> IP.  I suspect that pingd is non-working in the exact instance
> Lighty is set to "inactive".
> 
> primitive Public-IP ocf:heartbeat:IPaddr2 \
>        params ip="134.94.252.127" broadcast="134.94.253.255" \
> 	   nic="eht0" cidr_netmask="23" \
>        op monitor interval="60s" timeout="10s" \
>        meta migration-threshold="10"
> primitive lighty lsb:lighttpd \
>        op monitor interval="60s" timeout="30s" on-fail="restart" \
>        op start interval="0" timeout="60s" \
>        meta migration-threshold="3" failure-timeout="30s" \
> 	   target-role="Started"
> primitive pingd ocf:pacemaker:pingd \
>        params host_list="134.94.111.186" multiplier="100" \
>        op monitor interval="15s" timeout="20s"
> group lighty_group Public-IP lighty
> clone pingclone pingd meta globally-unique="false"
> location lighty-on-connected-node lighty_group \
>        rule $id="lighty-on-connected-node-rule" -inf: not_defined pingd
or
>        pingd lte 0

The last location rule says that lighty_group should never run on a node
that doesn't have connectivity. So if the ping node goes down, then
lighty_group will be shut down everywhere.

I'm not certain on how to improve this situation, since I'm having trouble
with a similar configuration. In principle you should be assigning a
positive score to nodes that can ping, so that you don't assign negative
scores. As far as I know only negative scores can cause pacemaker to shut
down a resource.

~David