[rfe] implement fsfreeze retry when creating snapshot

[Soroush Kaviani]

Hello
When i create snapshot from some of my instances with Windows OS the snapshot is deleted from list and when I check the nova compute log i see this error:

Nov 29 09:37:32 computeXX libvirtd[2880]: internal error: unable to execute QEMU agent command 'guest-fsfreeze-thaw': couldn't hold writes: fsfreeze is limited up to 10 seconds:

Qemu Guest Agent version in windows instances are updated but problem not solved.
now the only way to create snapshot from these instances is pausing and then create snapshot.

I have send guest-fs-freeze command manually to instance and i see sometime freezing maybe timed_out but if the fs-freeze request be repeated fs-freeze is done with success.

I think if nova repeats the fs_freeze request to qemu-guest-agent of instances when get timeout, the problem solve and live snapshot for all windows instances is without any problem.

thanks for your attention

[Balazs Gibizer (balazs-gibizer)]

This feels like a small feature request. The best would be to be able to pass a timeout parameter from nova to libvirt then nova can make the timeout configurable. As far as I see libvirt does not provide such parameter[1]. So yes the other option is to implement a retry loop on the nova-compute side. If you have the motivation to push this small feature forward then I suggest to show up on #openstack-nova IRC channel (on OFTC) to discuss this forward.

Marking this as valid as an rfe

[1] https://libvirt.org/html/libvirt-libvirt-domain.html#virDomainFSFreeze
[2] https://wiki.openstack.org/wiki/Meetings/Nova