[Xen-API] Fwd: Issue XCP 1.1: Shared LVM SR - unavailable host leaves VDI attached RW (xe vdi-introduce not available)

[Jakob Praher <jakob@xxxxxxxxxxx>]

Dear list,

We observed the following issue in conjunction w/ XCP 1.1 which is a
_severe_ problem.

Our environment:
- Two XCP 1.1 hosts in the same resource pool
- Shared LVM based storage repository based on DRBD
- VMs running on both systems stored on the shared SR

If suddenly one host dies unexpectedly (e.g. hardware defect, power
issue) the metadata still assumes the VMs that were running on the
unavailable host are in power-state running.

After ensuring the master to be the still available host and manually
resetting the power-state of the VMs running on the unavailable host we
faced the following problem:

 - The VDI state was still readwrite attached allthough the whole host
was unavailable. This also preserved when we made the system to forget
the unavailable host.

 - Even worse "xe vdi-introduce" does not work since it is not supported
by the LVM backend. All other tricks even snapshotting the disk does not
work, since XAPI tries to contact the dead host and this fails.

The problem is that as long the host is down for maintenance the VMs are
in a defective state!

What is the way out of this dilemma?

Best regards,
Jakob

P.S.: I have not only restarted xapi but also the available host.

_______________________________________________
xen-api mailing list
xen-api@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/mailman/listinfo/xen-api