High: Xen: retry domain lookup in repeating monitor and stop
If the guest is rebooting, it may completely disappear from the
list of defined guests, thus xm/xen-list would return with not
running. Apparently, this period lasts only for a second or two.
It is easily reproducible by rebooting the guest.
The effect is that a guest can be started on two nodes in
parallel or even twice on the same node thus resulting in guest's
filesystems corrupted.
In both stop and repeating monitor actions, if we encounter no
domain (OCF_NOT_RUNNING), we keep trying for the next 5 seconds.
The impact of this patch is that when stopping a non-existent
domain or unexpected disappearance of a running domain there will
be an extra 5 second delay.
This should be repaired within the xen domain.