[Users] Migration from NFS backed VM to iSCSI LUN without exporting them on the Ovirt Export Share

Itamar Heim iheim at redhat.com
Fri Jan 17 23:05:23 UTC 2014


On 01/18/2014 12:57 AM, squadra wrote:
> On Fri, Jan 17, 2014 at 11:48 PM, Itamar Heim <iheim at redhat.com
> <mailto:iheim at redhat.com>> wrote:
>
>     On 01/09/2014 10:16 AM, squadra wrote:
>
>         Hello folks,
>
>         since the nfs discussion remembered me that i got some vm left to
>         migrate... here some kind of special question. i am open for
>         non-bestpractice hacky solutions, too
>
>         Situation is:
>
>         - 2 Ovirt Cluster - Same DC - one NFS backed, one iSCSI backed
>         - NFS Share and iSCSI Share are exported from the same phys. Machine
>         - Bot DC use the same filer, just different Luns / Protocolls
>
>
>         so, i thought about something simple like just move vm folder
>         from a to
>         b and do a little bit database voodoo?
>
>         anyone did something like this yet? or is storage live migration
>         already
>         working for this? the docs didnt tell me very much about this.
>
>
>     are the VMs thin or pre-allocated? with or without snapshots?
>
>
> mostly thin, but i wouldnt care about loosing the thin feature since the
> underlaying filer runs zfs with compression. and nop, no snaps through
> ovirt.
>
>
> you got some ugly hack suggestion for me? :D
>

well, no voodo for nfs to iscsi.

you can try and create the same VMs/disks (manually to check, then 
script with a different vm name (due to uniqueness) on the new DC.
then try to dd the source disk to the target one.

unless your storage can somehow do such mapping of LVs.

3.4 will hopefully allow to attach both nfs and iscsi domains to same 
DC, then you could try to move the VMs offline (or online) between the two.



More information about the Users mailing list