xcp
xcp copied to clipboard
if VM is shut-down during migration, restart hang forever
Migration can fail for several reasons. If it fails, most of the time Linux kernel print in the console a message:
Freezing of tasks failed after 20.005 seconds (1 tasks refusing to freeze, wq_busy=0):
Xen seems to be unaware that hybernation failed so wait for migration forever. In this case could be an option to just shutdown the VM and restart it on another host instead of use live migration.
However if you shutdown a VM and than it's been restarted it on the host that was already embracing the migration just before the shutdown than the start task'll hang forever. So if you don't choose 'another host' than the pool could choose the wrong host.
I guess this is caused by a missed cleanup on the target-host of the live migration (but maybe i'm in wrong).
In order to fix this situation I runned:
xe vm-reset-powerstate uuid=<XXXX-XXXX-XXXX-XXXX> force=true
After the reset suddenly restarted without any input. It was likely that start was in queue and was finally just pushed when the powerstate reach the shutdown state.
It is indeed not good that one has to run xe vm-reset-powerstate in this kind of situation. If you have the opportunity to reproduce the issue on XS 7.6, please also report it to the XenServer project.
Also happened here on a 8.0.0 with all recent patches while evacuating the host for bios updates,
8.1 beta will be here very soon, please try as soon as you can :+1:
Actually the beta is already available: https://xcp-ng.org/forum/topic/2516/xcp-ng-8-1-0-beta-now-available
I don't know if anything was done regarding this issue.
We're running into this issue during a pool update from 8.1 -> 8.2