Hi Alan,
This example helps, but there's one thing I can't seem to grasp. I
understand how these rules make sure that the file system stays on the
same node as the primary drbd, but what determines which node the
primary drbd is on? With a mirrored partition from drbd, I really don't
care which node takes over as master, but I need one of them to do it.
Right now it comes up as two slaves, waiting for one or the other to ask
to be promoted. I was hoping someone out there had a complete cib.xml
file configuring drbd with something else (i.e. database, apache, IP
address, etc), but I've gotten no response on that (other than your
example :)

Thanks,
Doug
On Fri, 2007-04-13 at 11:51 -0600, Alan Robertson wrote:

> Knight, Doug wrote:
> > I am trying to configure HB to monitor a drbd/postgreSQL configuration,
> > using the HB GUI. I have manually verified that drbd can be failed back
> > and forth between the two nodes in the HA cluster. I have also verified
> > that postgreSQL can run on either node, dependent upon the drbd device
> > being colocated and mounted with it. However, when I try to create a
> > group containing the drbd resource, I get an "activation failed"
> > response from the GUI and it won't create it. I started by executing a
> > cibadmin --cib_erase, so I started with a completely clean cib.xml. Then
> > I used the GUI to add a group, call it grp_pgsql_mirror. The first
> > resource I define is the drbd, call it rsc_drbd_7788, using the drbd
> > ocf/heartbeat-supplied script. Master/Slave ID is ms_drbd_7788. Once
> > I've entered all the info, clicking Add puts a small window up on the
> > screen with "Activation Failed" in it and an OK button. I'm trying to
> > use a group so I can fix the ordering of the resources I need to handle
> > the drbd/postgreSQL resource (i.e. drbd, filesystem, IP address, and
> > postgreSQL resources). I'm about ready to manually write the cib.xml,
> > but my XML writing skills aren't the best, and I would really prefer to
> > use a tool to create or make changes to the cib.xml. Any ideas or help
> > would be great. If anyone has a cib.xml with a similar configuration
> > they'd be willing to share, that would be great too.
> 
> 
> Master/slave (clone) resources cannot go into groups, IIRC.
> 
> You need to have the group depend on the clone being in master state.
> 
> Here's a CIB idiom for doing this:
>       http://linux-ha.org/CIB/Idioms/ColoWithMaster
> 
> I haven't tested this, and it's more than possible you can't do all this
> with the GUI.  But, you should be able to create the master/slave
> resource and the group just fine.
> 
_______________________________________________
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