[Pacemaker] pacemaker resource constraints

[Alan Jones]

BTW: The order matters in the colocation rule.  When I configure:
colocation colo-master_worker -1: master worker
Then "failback" is blocked by the stickiness.  In my opinion this is a bug,
but others may have an explanation.
This is the default version that installs on FC12 using the GUI software
packages tools.
Alan

On Tue, Mar 23, 2010 at 3:47 PM, Alan Jones <falancluster at gmail.com> wrote:

> The following rules give me the behavior I was looking for:
>
> primitive master ocf:pacemaker:Dummy meta resource-stickiness="INFINITY"
> is-managed="true"
> location l-master_a master 1: fc12-a
> location l-master_b master 1: fc12-b
> primitive master ocf:pacemaker:Dummy
> location l-worker_a worker 1: fc12-a
> location l-worker_b worker 1: fc12-b
> colocation colo-master_worker -1: worker master
>
> To recap, the goal is an active-active two node cluster were "master" is
> sticky and "master" and "worker" with anti-colocate when possible for
> performance.
> Note that I had to add points for each resource on each node to overcome
> the negative colocation value to allow them both to run on one node.
> If there is a more elegant solution, let me know.
> Alan
>
>
> On Tue, Mar 23, 2010 at 8:24 AM, Andrew Beekhof <andrew at beekhof.net>wrote:
>
>> On Mon, Mar 22, 2010 at 9:18 PM, Alan Jones <falancluster at gmail.com>
>> wrote:
>> > Well, I guess my configuration is not as common.
>> > In my case, one of these resources, say resource A, suffers greater
>> > disruption if it is moved.
>> > So, after a failover I would prefer that resource B move, reversing the
>> node
>> > placement.
>> > Is this possible to express?
>>
>> Make A stickier than B.
>>
>> Please google for the following keywords:
>>    site:clusterlabs.org resource-stickiness
>>
>> > Alan
>> >
>> > On Mon, Mar 22, 2010 at 11:10 AM, Dejan Muhamedagic <
>> dejanmm at fastmail.fm>
>> > wrote:
>> >>
>> >> Hi,
>> >>
>> >> On Mon, Mar 22, 2010 at 09:29:50AM -0700, Alan Jones wrote:
>> >> > Friends,
>> >> > I have what should be a simple goal.  Two resources to run on two
>> nodes.
>> >> > I'd like to configure them to run on separate nodes when available,
>> ie.
>> >> > active-active,
>> >> > and provide for them to run together on either node when one fails,
>> ie.
>> >> > failover.
>> >> > Up until this point I have assumed that this would be a base use case
>> >> > for
>> >> > Pacemaker, however, it seems from the discussion on:
>> >> > http://wiki.lustre.org/index.php/Using_Pacemaker_with_Lustre
>> >> > ... that it is not (see below).  Any ideas?
>> >>
>> >> Why not just two location constraints (aka node preferences):
>> >>
>> >> location l1 rsc1 100: node1
>> >> location l2 rsc2 100: node2
>> >>
>> >> Thanks,
>> >>
>> >> Dejan
>> >>
>> >> > Alan
>> >> >
>> >> > *Note:* Use care when setting up your point system. You can use the
>> >> > point system if your cluster has at least three nodes or if the
>> resource
>> >> > can acquire points from other constraints. However, in a system with
>> >> > only two nodes and no way to acquire points, the constraint in the
>> >> > example above will result in an inability to migrate a resource from
>> a
>> >> > failed node.
>> >> >
>> >> > The example they refer to is similar to yours:
>> >> >
>> >> > # crm configure colocation colresOST1resOST2 -100: resOST1 resOST2
>> >>
>> >> > _______________________________________________
>> >> > Pacemaker mailing list
>> >> > Pacemaker at oss.clusterlabs.org
>> >> > http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>> >>
>> >>
>> >> _______________________________________________
>> >> Pacemaker mailing list
>> >> Pacemaker at oss.clusterlabs.org
>> >> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>> >
>> >
>> > _______________________________________________
>> > Pacemaker mailing list
>> > Pacemaker at oss.clusterlabs.org
>> > http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>> >
>> >
>>
>> _______________________________________________
>> Pacemaker mailing list
>> Pacemaker at oss.clusterlabs.org
>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20100323/b5ee69f5/attachment-0001.html>