[Linux-HA] Failed takeover of drbddisk/xen stack

"Robert Zöhrer | pronet.at" robert.zoehrer at pronet.at
Thu Jul 23 15:59:48 MDT 2009


I have a running LVM-DRBD-Xen-HVM/Win2k3 cluster-stack under debian/
lenny and want to manage it with heartbeat/crm.

I defined crm/cib with 2 resources (non ocf drbddisk and ocf Xen) within 
an ordered and collocated resource group.

To bind my stack to node1 on normal operation I've set one location 
constraint regarding the whole resource group.

My config in detail:

ha.cf: http://pastebin.com/m72309a54
/var/lib/heartbeat/crm/cib.xml: http://pastebin.com/m1c73356

Now when performing a manual takeover (standby command for resource 
holding node) I get errors while heartbeat wants  to stop the drbd 
resource on the loosing node. The drbd resource become unmanaged it the 
takeover process hangs.

Log in detail: http://pastebin.com/m3babd102

At this point when I switch back the resource holding node to active 
mode and perform a "clean up" to the group resource by hb_gui the 
resources restarts correctly on this node.

When I hardly shut down (power button) the resource holding node 
(instead of performing a manual takeover) the takeover seems to be 
working well.

On the other hand my location constraint doesn't seem to work .. maybe 
caused by the same reason?

I've defined CIB/CRM by hb_gui .. but believed to understand the CIB/XML 
concept. Also gui xml output seems consequentially ok (for me).

Any is appreciated.

Thx Robert

More information about the Linux-HA mailing list