[ClusterLabs] Resources within a Group

[Ken Gaillot]

On 08/28/2015 07:46 PM, Jorge Fábregas wrote:
> On 08/28/2015 10:06 AM, Ken Gaillot wrote:
>> I'm wondering if maybe you were using the term "group" generically? The
>> other responses were referring to Pacemaker's explicit group syntax.
> 
> Hi Ken,
> 
> Yes, I was referring to Pacemaker's group syntax.  I had this CRMish
> segment:
> 
> primitive p-vg01 LVM \
>         params volgrpname=vg01 exclusive=true \
>         op monitor interval=10 timeout=30 \
>         op stop interval=0 timeout=30 \
>         op start interval=0 timeout=30
> 
> primitive p-filesystem Filesystem \
>         params directory="/oracle" fstype=ext3 device="/dev/vg01/lvol1" \
>         op monitor interval=20 timeout=40 \
>         op stop interval=0 timeout=60 \
>         op start interval=0 timeout=60
> 
> group g-oraclefs p-vg01 p-filesystem \
>         meta target-role=Started
> 
> ...and I wasn't able to start the group "g-oraclefs".  Then I saw some
> errors that it coulnd't activate the VG followed by mounting attempts
> (where it mentioned that it coulnd't find /dev/vg01/lvol1").  That's
> when I thought: "Why is it trying to start the 2nd resource when the
> first one failed?"

"Couldn't find" isn't the same as "couldn't mount" ... perhaps the
message is coming from a probe (the cluster making sure the resource
isn't already running) rather than a start?

The probe (a one-time monitor operation) will check whether the
directory is a mountpoint or not. I'm guessing if the directory doesn't
exist, it'll print that message but proceed normally. If it's not
causing any cluster errors, I think you can ignore it.

If the cluster really is trying to start p-filesystem even though p-vg01
failed, then something's really screwy. Pastebin the logs around the
problem if that's the case.

> I don't have the problem now but one of the errors was that I had set a
> Target-Role of "Started" for the group itself but forgot that I had a
> Target-Role of "Stopped" for both primitives (set at creation time).
> This got me for a while!  I finally figured it out when
> starting/stopping via the crm shell that warned me.  Thru the Hawk web
> interface it wasn't clear.
> 
> 
> Thanks,
> Jorge
> 
> _______________________________________________
> Users mailing list: Users at clusterlabs.org
> http://clusterlabs.org/mailman/listinfo/users
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org
>