On Tue, Jan 11, 2011 at 3:31 PM, RaSca <[email protected]> wrote:
> Il giorno Mar 11 Gen 2011 13:36:58 CET, RaSca ha scritto:
>> Hi all,
>> I've got two group of resources, say A and B. A depends on B, so if B
>> isn't up A must NOT be started. This is the situation:
>> group B Bres1 Bres2 Bres3
>> colocation B_ON_B_ms-r1 inf: B B_ms-r1:Master
>> order B_AFTER_B_ms-r1 inf: B_ms-r1:promote B:start
>> group A  Ares1 Ares2 Ares3
>> colocation A_ON_A_ms-r0 inf: A A_ms-r0:Master
>> order A_AFTER_A_ms-r0 inf: A_ms-r0:promote A:start
>> order A_AFTER_B inf: B:start A:start
>> In a running configuration everything works fine, all the services
>> switch in case of failures or if I force a manual move.
>> The problem is on startup, because for some reason Pacemaker (1.0.10)
>> start first of all Ares1, that of course fails because B is not started.
>> I think that the last order constraint must obligate A to not start
>> before B is done, but things are not going in this way.
>> What am I missing?
>> Thanks a lot!
>
> To be more specifically, this is the log of what happens when I startup
> just one node: http://pastebin.com/YsP4B94r and this is my
> configuration: http://pastebin.com/77gwm1Gr
>

Not sure where you are with this, but these logs indicate that some
things are already running (rc=0) when the cluster starts:

Jan 11 15:11:52 SE4 crmd: [32095]: info: match_graph_event: Action
db_iscsi-lun_monitor_0 (7) confirmed on se4 (rc=0)
Jan 11 15:11:52 SE4 crmd: [32095]: info: match_graph_event: Action
db_iscsi-target_monitor_0 (6) confirmed on se4 (rc=0)

This may create the impression that pacemaker started things in the
wrong order, even though it didn't.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to