Re: [Xen-API] migration questions

[Mike McClurg <mike.mcclurg@xxxxxxxxxx>]

On 26/07/12 18:30, Thomas Gazagnaire wrote:
> Hi,
> 
> I've found that migrate_send *always* tries to use the new VDI
> storage motion feature, even when one of the VDI to move is on a
> shared storage and when the destination is on the same pool. I've get
> a tentative patch to fix that, I'll make a pull request after more
> testing, but just wanted to check that fixing this issue makes
> sense.

Yes, I think that this behaviour makes sense. There's no point in
copy/mirroring VDIs that are on a shared SR.

> Also when migrate_send is done from the master to the slave, still
> with a shared storage, the temporary remote VDI created on the slave
> has an associated tapdisk process on the slave, but the clean-up
> VDI.destroy call will be forwarded to the master which try to kill
> running tapdisk process -- but there is no tapdisk process there, and
> then vdi_destroy is forwarded to the slave again and finally that's
> failing because of the running tapdisk on the slave. A solution to
> fix this would be to speak directly to the remote storage layer to
> destroy the VDI, but it seems than no one does that in
> xapi_vm_migrate.ml. Any thoughts on this ?

That seems like a layering violation to me, and I don't know if it's
even possible. Jon (CC'd) will have to comment on this one...

Mike

_______________________________________________
Xen-api mailing list
Xen-api@xxxxxxxxxxxxx
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api