[Pacemaker] Migration atomicity

[Andrew Beekhof]

On Tue, Mar 20, 2012 at 9:26 PM, Lars Marowsky-Bree <lmb at suse.com> wrote:
> On 2012-03-20T11:42:08, Andrew Beekhof <andrew at beekhof.net> wrote:
>
>> > I'm observing a little bit unintuitive behavior of migration logic when
>> > transition is aborted (due to CIB change) in the middle of the resource
>> > migration.
>> >
>> > That is:
>> > 1. nodea: migrate_to nodeb
>> > 2. transition abort
>> > 3. nodeb: stop
>> > 4. nodea: migrate_to nodec
>>
>> I'd like to see that a crm_report showing that behavior.
>> Because I'm looking at the same scenario and I see:
>>
>> 1. nodea: migrate_to nodeb
>> 2. transition abort
>> 3. nodea: stop
>> 4. nodeb: stop
>> 5. nodec: start
>
> Yes, that's what I see and expect too.
>
> What would obviously be very nice is if the PE could reconstruct that a
> migration is on-going and allow it to complete first, before again
> shuffling the resource to where it now believes it should be. (To avoid
> service downtime.) I wonder how hard that'd be? ;-)

You're welcome to find out :-)
   http://bugs.clusterlabs.org/show_bug.cgi?id=5045