[Pacemaker] Lighty doesn't come up always

[Torsten Bronger]

Hallöchen!

Dr. Michael Schwartzkopff writes:

> Am Dienstag, den 06.07.2010, 10:28 +0200 schrieb Torsten Bronger:
>
>> We have a two-node cluster with a virtual IP and Lighty running
>> on that node which has this IP currently.  Thus, our
>> configuration says:
>> 
>> [...]
>> 
>> The problem is that under some circumstances, Lighty is not
>> started.  Instead, crm_mon shows at the bottom:
>
> Remove the collocation constraint because it is implicitly given
> in the group. Then it shoud work.

Thank you, I removed the superfluous line.  However, the problem is
still there.

If I start Heartbeat on both nodes simultaneously, Lighty is not
started, and the log on one of the nodes says

Jul  6 11:27:08 olga pengine: [7362]: notice: group_print:  Resource Group: lighty_group
Jul  6 11:27:08 olga pengine: [7362]: notice: native_print:      Public-IP#011(ocf::heartbeat:IPaddr2):#011Stopped
Jul  6 11:27:08 olga pengine: [7362]: notice: native_print:      lighty#011(lsb:lighttpd):#011Stopped
Jul  6 11:27:08 olga pengine: [7362]: notice: clone_print:  Clone Set: pingclone
Jul  6 11:27:08 olga pengine: [7362]: notice: short_print:      Stopped: [ pingd:0 pingd:1 ]
Jul  6 11:27:08 olga attrd: [7356]: info: attrd_trigger_update: Sending flush op to all hosts for: terminate (<null>)
Jul  6 11:27:08 olga pengine: [7362]: info: native_merge_weights: Public-IP: Rolling back scores from lighty
Jul  6 11:27:08 olga pengine: [7362]: info: native_color: Resource Public-IP cannot run anywhere
Jul  6 11:27:08 olga pengine: [7362]: info: native_color: Resource lighty cannot run anywhere
Jul  6 11:27:08 olga pengine: [7362]: notice: RecurringOp:  Start recurring monitor (15s) for pingd:0 on mandy
Jul  6 11:27:08 olga pengine: [7362]: notice: RecurringOp:  Start recurring monitor (15s) for pingd:1 on olga
Jul  6 11:27:08 olga pengine: [7362]: notice: LogActions: Leave resource Public-IP#011(Stopped)
Jul  6 11:27:08 olga pengine: [7362]: notice: LogActions: Leave resource lighty#011(Stopped)

If I start both nodes one after the other with two minutes delay,
everything works fine.  Why is this?

Tschö,
Torsten.

-- 
Torsten Bronger, aquisgrana, europa vetus
                   Jabber ID: torsten.bronger at jabber.rwth-aachen.de
                                  or http://bronger-jmp.appspot.com