[Pacemaker] Make resources more important than others

[Andrew Beekhof]

There is resource priorities, but I don;t think that will help in this case.

On Thu, Jun 2, 2011 at 5:44 PM, Luke Bigum <Luke.Bigum at lmax.com> wrote:
> Hi list,
>
> I have a conceptual question about making Pacemaker treat some resources a
> lot more important than others. I'm afraid the answer is going to come back
> "sorry you can't", but I just want to confirm.
>
> I have a simple 2 node cluster that runs two services (MySQL and a Java
> daemon) on top of network attached storage that can float between both
> nodes. CRM config looks a bit like:
>
> node node1
> node node2
> primitive lsb:mysql mysql
> primitive lsb:java java
> primitve lsb:storage storage
> colocation mysql_with_storage inf: mysql storage
> colocation java_with_storage inf: java storage
> order mysql_after_storage: storage:start mysql:start symmetrical=true
> order java_after_storage: storage:start java:start symmetrical=true
>
> I've got the interesting requirement that mysql is vastly more important
> than the java resource. The java has to run on the same server as mysql.
> Also, if mysql is stopped/unmanaged, the java should still be running on
> what would be the correct node for mysql. I've mostly achieved this by
> colocating the java and mysql to the underlying storage.
>
> It gets tricky when we start simulating hard failures. If we simulate a hard
> error with mysql, the storage moves to another node, then mysql and java,
> and everything's great. What I don't want is the opposite to occur. I don't
> want any hard error with java to make the mysql move, as I consider the
> mysql more important than java and I don't want to outage mysql if the java
> has a problem. I'd love for the mysql to stay running  where it is and have
> an administrator come along and cleanup what's wrong with the java.
>
> I guess I want to somehow describe the java resource as "not as important"
> as mysql, and this is what I don't think is possible. I know I'm talking
> about very edge cases by simulating hard errors (eg: monitor return code 5),
> but it'd be nice to achieve.
>
> Any thoughts?
>
> Thanks in advance,
>
> -Luke
>
> --
> Luke Bigum
>
> Information Systems
> Luke.Bigum at lmax.com | http://www.lmax.com
> LMAX, Yellow Building, 1A Nicholas Road, London W11 4AN
>
>
> The information in this e-mail and any attachment is confidential and is
> intended only for the named recipient(s). The e-mail may not be disclosed or
> used by any person other than the addressee, nor may it be copied in any
> way. If you are not a named recipient please notify the sender immediately
> and delete any copies of this message. Any unauthorized copying, disclosure
> or distribution of the material in this e-mail is strictly forbidden. Any
> view or opinions presented are solely those of the author and do not
> necessarily represent those of the company.
>
> _______________________________________________
> 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
>