[virt-tools-list] KVM Live Migration

[Justin Clift]

On 10/06/2010 01:40 PM, Mike Hall wrote:
>> As a thought, do you have SELinux enabled on the servers host
>> servers, and are the host servers patched to the latest available
>> patches?
>
> SELinux is/was enabled virt_use_nfs == on No patches, bog standard
> install with available RPM updates
>
> I tried disabling SELinux (setenforce 0), but same result. The guest
> is pingable from the VM host, but not the network. Becomes pingable
> again if re-migrated back to original host. Possibly some sort of
> switch/ARP update required?
>
> I'm pretty sure the shared NFS stuff is working OK, the guests do
> actually migrate and restart, its just that rebooting and losing
> network connectivity isn't the desired outcome :-)
>
> I'm looking into libvirt hook scripts, not sure if they're relevant
> yet or not. Also not sure if I'm dealing with one problem or two
> separate problems (rebooting + network).

Yeah, it sounds weird.

If you try kicking off the migrations from the command line, using virsh
instead of virt-manager, what happens?

ie.  from a command line on the source host:

   $ sudo virsh migrate myguestname qemu+ssh://destserver/system --live

Suggesting this purely because using the command line options gives a
more obvious mental handle on what's being attempted.

The only other thing I can think of, and this is a long shot, is that
maybe there's a firewall enabled on one or both (or the NFS server),
and maybe that's "somehow" blocking communication flow at some point?

Not sure how that could happen, but if it's not a production host (?),
then it's pretty easy to test by turning off the firewall. :)