[Pacemaker] crm_resource - migrating/halt a cloned resource

[Andrew Beekhof]

On Dec 6, 2008, at 4:45 PM, Raoul Bhatia [IPAX] wrote:

> Andrew Beekhof wrote:
>> but not the clone name (i took a guess before because i didn't know  
>> what
>> it was called):
>>
>>  crm_resource -M -r clone_webservice
>
> indeed - but in my opinion, i want to tell the cluster the exact
> resource that i want to migrate and the cluster should work the rest
> out itself.

nope

this doesn't make sense for clones because such a location constraint  
does nothing to prevent any of its peers from running there.
so you've caused a heap of churn and the clone is still running there  
- just with a different name.

bottom line, you don't get to chose where specific clone instances get  
placed.

> as far as i remember, the clones/groups are just an easy way to
> configure resources and not having to specify all the rules/ 
> constraints
> manually.

groups yes, clones definitely not.
there is no combination of primitives and constraints that results in  
a clone.

> so, imho, it could be valid that parts of the group remain running.
>
> a.g. when i have a group with A B C D and C is not able to start,
> then A B will still be able to run - right?

possibly, but that is a failure state and being able to "configure"  
every possible failure state isn't one of the design goals :-)