[Pacemaker] How to check resource availability after failover

[Marc Wilmots]

Thanks Pavlos!

So it confirms what I thought, it's not possible with pacemaker 0.6.
That's no good news, as I'm not allowed to install anything in this server
that isn't coming from the official CentOS repositories...

Let's say I don't mind whether the resource comes back to it's original
node.
Imagine, that apache fails on node1, then node2 starts apache. Now imagine
that after a few days, apache also fails on node2. In this case, node1 will
failover. The thing is that node1 never tries to start Apache. I suppose
this is because of the fail-count? Is there a way to circumvent this?

2010/12/1 Pavlos Parissis <pavlos.parissis at gmail.com>

> On 1 December 2010 14:53, Marc Wilmots <desjter at gmail.com> wrote:
> > Hi list,
> >
> > I'm kinda new to Pacemaker and have some questions I hope some of you
> could
> > help me with.
> >
> > First of all, I'm using Heartbeat v2.1.3 with CRM still integrated. If
> I'm
> > correct, that must be version 0.6 of Pacemaker.
> > The reason for using this version is because I am not allowed to install
> any
> > applications from source, and version 2.1.3 is the latest version
> available
> > in the CentOS repositories. (Does anyone know why?)
>
> This is not correct. Heartbeat 3.0.3 and pacemaker 1.0.10 is available
> from Clusterlabs.org repository, take a look here
> http://www.clusterlabs.org/rpm/epel-5/i386/?C=M;O=D
> You better use heartbeat 3.0.3 and pacemaker 1.0.10, all the
> information about the installation are in the wiki pages.
> >
> > My setup is as follows:
> >
> > 2 servers with each 1 Apache httpd to have a Active//Passive high
> > availability using a virtual IP (IPaddr2).
> >
> > My problem is the following:
> >
> > When Pacemaker detects the Apache is not available, it stops the Apache
> > process and fails over to the backup node. I would like Pacemaker to try
> to
> > start the Apache again after some time and check whether it's working
> > correctly again.
> Tune the migration-threshold, set it to 3 for instance and it will try
> to restart it 2 times on the node before it fails over to another
> node.
> > If so, then I would like it to move the resources (IP,
> > Apache) back to the first node, and leave the backup node as passive
> again.
> then make sure resource-stickiness is very low, lower than any location
> scores
> >
> > The only way I found is restarting Heartbeat on the first node so it
> > actually tries to start Apache again. This is not what I want as it
> > requieres manual actions.
> >
> > I've been reading the 0.6 documentation and been searching on Google for
> > quite some time now, but I can't find the right solution. People with
> > similar problems were being advised to alter failure_timeout,
> > cluster_recheck_interval, migration_threshold, etc. but I can't find any
> > information about that for version 0.6.
> >
> > That makes me think it's actually not achieve what I need with Pacemaker
> > 0.6, is that so?
>
> Use pacemaker 1.0.10 and you can accomplish what you want, just read
> the documentation
> >
> > Thanks a bunch.
> >
> >
> >
> > _______________________________________________
> > 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://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
> >
> >
>
> _______________________________________________
> 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://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20101201/bd4578f6/attachment-0002.htm>