[Pacemaker] RFC: How do you upgrade?

[Junko IKEDA]

> >>> I'd like to get a sense of how people upgrade their clusters...
> >>> * rolling upgrade (node-by-node)
> >>> * big-bang (all nodes at once with all resources stopped)
> >>> * detach and reattach (all nodes at once with resources still running)
> >>> * other?
> >>> I'm interested to know how popular the various methods are.
> >>> Andrew
> >> I prefer no. 3 (detach and reattach).
> >>
> >> Works quite nicely.
> >
> > Do you do that node-by-node or cluster-wide?
> 
> Cluster wide.
> 
> Backup cib.xml
> Stop all monitor operations
> Set all resources unmanaged
> Stop heartbeat on all nodes
> Upgrade heartbeat / install pacemaker on all nodes
> Start heartbeat on all nodes
> 
> See what the cluster would do when re-activating it
> <beekhof> before you turn back on is-managed...
> <beekhof> cibadmin -Ql > tmp.xml
> <beekhof> vi tmp.xml (change is-managed)
> <beekhof> ptest -VVVVV -X tmp.xml -D tmp.dot
> <beekhof> look in tmp.dot and see what we'd do
> 
> If that's to my linking:
> 
> Set all resources managed
> Start all monitor operations

well...  I'll try this way.

order of preference;
3 (detach and reattach) > 1 (rolling upgrade) > 2 (big-bang)

So far, we have chosen 1 or 2 (mainly 2 with hesitancy),
if 3 is possible (and stable), it's the best one.
We have no experience of updating on VM .

Thanks,
Junko