On Tue, Jun 3, 2008 at 8:17 AM, Junko IKEDA <[EMAIL PROTECTED]> wrote:
>> Then (because of the probe) we find out it _is_ running afterall and
>> we end up in the situation contained in pe-input-6.bz2
>>
>> We only guarantee that the probe for rscX completes before we start the
> rscX.
>
> Start failures which are set as on_fail=block also induce the unmanaged
> status as the same as stop failures.

Right.  Thats what specifying on_fail=block does - it makes the
resource unmanaged.

> When I run crm_resource -C to recover the start failure,
> It seems that probe for node (probe_complete) is also called not only probe
> for rscX.

Right - its just a "we've done all the probing for this node" action.
Nothing to worry about.

> So, all the resources in group would start again in order if I use Heartbeat
> 2.1.3.
> (Pacemaker would start only the resource which are not started)

right

> Does this mean, if group contains some primitive resources,
> start operations would be called in order,

yes

> but stop operations might not be depending on conditions?

If by "conditions" you mean "removing all knowledge of the resource's
state", then yes :-)

It probably wouldn't be too hard to implement this though (ordering
the probe before the stop action).
If you file an enhancement request I'll try and get it done for 1.0.
_______________________________________________
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