[Pacemaker] help with configuration for Xen domU on two DRBD devices

[Florian Haas]

On 2010-09-13 01:56, Jai wrote:
> Hi
> 
> I'm new to pacemaker configurations and am trying to replace my old heartbeat two node cluster setup using haresources to pacemaker/corosync. I have mostly followed the instructions from http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf.
> 
> My configuration is for two drbd devices which are used as the filesystems for a xen DomU so they both need to be on the same node and in master role before the domU can be started. But I think I still have something wrong with my config (configs are below) as I had some trouble getting my colocation and order commands to work with master/slave resources.
> 
> Could you please have a look at my configuration and tell me what's wrong with it?
> 
> The problem:
> During testing most actions seemed to work as expected(shutdown active node, failover of resources in correct order succeed), however pulling the power from the active node did not produce the results expected, which where that node alpha would take over resources.
> 
> [...]
> 
> Failed actions:
>    testDomU_start_0 (node=alpha, call=17, rc=1, status=complete): unknown error
>    drbd0:1_monitor_10000 (node=alpha, call=31, rc=8, status=complete): master

There should error messages in the Pacemaker logs for both of these.

> [...]
>
> I expected drbd fencing to add both these rules but only one was added, the one for drbdTmp.
> 
> location drbd-fence-by-handler-drbdRoot drbdRoot \
> 	rule $id="drbd-fence-by-handler-rule-drbdRoot" $role="Master" -inf: #uname ne alpha
> location drbd-fence-by-handler-drbdTmp drbdTmp \
> 	rule $id="drbd-fence-by-handler-rule-drbdTmp" $role="Master" -inf: #uname ne alpha

Well I see none of those in your crm configure show output. Are you sure
that even one of them is being set?

> colocation testDomU-with-drbdRoot inf: testDomU drbdRoot:Master
> colocation testDomU-with-drbdTmp inf: testDomU drbdTmp:Master
> colocation drbdTmp-with-drbdRoot inf: drbdRoot:Master drbdTmp:Master

The third colocation is redundant, is it not?

> order order-drbdRoot-before-testDomU inf: drbdRoot:promote testDomU:start
> order order-drbdTmp-and-drbdRoot inf: drbdTmp:promote drbdRoot:promote

> # cat /etc/xen/test.sxp
> name = "testDomU"
> memory = 3000
> bootloader = "/usr/bin/pygrub"
> on_poweroff = "destroy"
> on_reboot = "restart"
> on_crash = "restart"
> vfb = [ "type=vnc,vncunused=1" ]
> disk = [ "phy:/dev/drbd0,xvda1,w", 
> 	"phy:/dev/drbd1,xvda2,w",
>        "phy:/dev/VGsys/LVswap1,xvda3,w", ]

Eeeeek. That's ugly. But I guess if you never want to be able to
migrate, it might actually work.

Take a closer look at the kernel logs (for DRBD) and /var/log/messages
(for Pacemaker). There should be something useful in there that will
help you troubleshoot this issue.

Cheers,
Florian

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 262 bytes
Desc: OpenPGP digital signature
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20100921/bab7006b/attachment-0003.sig>