[ClusterLabs] About globally unique resource instances distribution per node

[Daniel Hernández]

> On 1/4/16, Ferenc Wagner <wferi at niif.hu> wrote:
> Please check how http://bugs.clusterlabs.org/show_bug.cgi?id=5221
> applies to your case.  Comment on the issue if you've got something to
> add.
> --
> Regards,
> Feri.
>

I found that http://bugs.clusterlabs.org/show_bug.cgi?id=5221 applies
to my case when I have free capacity (utilization value greater than
total resource value). In my example I have capacity: on node1 6, on
node2 9, on node3 8. If I change that capacity to: on node1 4, on
node2 3, on node3 5 (utilization value equal to total resource value),
there the cluster start the instances as I want(on node1 4, on node2
3, on node3 5).

 Clone Set: clone_example1 [example1] (unique)
     example1:0	(ocf::heartbeat:Dummy):	Started node3
     example1:1	(ocf::heartbeat:Dummy):	Started node1
     example1:2	(ocf::heartbeat:Dummy):	Started node3
     example1:3	(ocf::heartbeat:Dummy):	Started node2
     example1:4	(ocf::heartbeat:Dummy):	Started node1
     example1:5	(ocf::heartbeat:Dummy):	Started node3
     example1:6	(ocf::heartbeat:Dummy):	Started node2
     example1:7	(ocf::heartbeat:Dummy):	Started node1
     example1:8	(ocf::heartbeat:Dummy):	Started node3
     example1:9	(ocf::heartbeat:Dummy):	Started node2
     example1:10	(ocf::heartbeat:Dummy):	Started node1
     example1:11	(ocf::heartbeat:Dummy):	Started node3


I find that maybe if I use different names for utilization on nodes
for different resources, that could work when having multiple
resources. Example:

crm node utilization node1 set cpuexample1 4
crm node utilization node2 set cpuexample1 3
crm node utilization node3 set cpuexample1 5

crm node utilization node1 set cpuexample2 2
crm node utilization node2 set cpuexample2 6
crm node utilization node3 set cpuexample2 3

crm resource utilization example1 set cpuexample1 1
crm resource utilization example2 set cpuexample2 1

Using balanced placement strategy the cluster state was:

Clone Set: clone_example1 [example1] (unique)
     example1:0	(ocf::heartbeat:Dummy):	Started node3
     example1:1	(ocf::heartbeat:Dummy):	Started node1
     example1:2	(ocf::heartbeat:Dummy):	Started node3
     example1:3	(ocf::heartbeat:Dummy):	Started node2
     example1:4	(ocf::heartbeat:Dummy):	Started node1
     example1:5	(ocf::heartbeat:Dummy):	Started node3
     example1:6	(ocf::heartbeat:Dummy):	Started node2
     example1:7	(ocf::heartbeat:Dummy):	Started node1
     example1:8	(ocf::heartbeat:Dummy):	Started node3
     example1:9	(ocf::heartbeat:Dummy):	Started node2
     example1:10	(ocf::heartbeat:Dummy):	Started node1
     example1:11	(ocf::heartbeat:Dummy):	Started node3
 Clone Set: clone_example2 [example2] (unique)
     example2:0	(ocf::heartbeat:Dummy):	Started node2
     example2:1	(ocf::heartbeat:Dummy):	Started node2
     example2:2	(ocf::heartbeat:Dummy):	Started node2
     example2:3	(ocf::heartbeat:Dummy):	Started node2
     example2:4	(ocf::heartbeat:Dummy):	Started node2
     example2:5	(ocf::heartbeat:Dummy):	Started node3
     example2:6	(ocf::heartbeat:Dummy):	Started node1
     example2:7	(ocf::heartbeat:Dummy):	Started node3
     example2:8	(ocf::heartbeat:Dummy):	Started node1
     example2:9	(ocf::heartbeat:Dummy):	Started node3
     example2:10	(ocf::heartbeat:Dummy):	Started node2

I am not sure if in all the scenarios that work the same way but I
will consider using that approach.
Thanks for your time.
Daniel