The absurdity of migrating a VM vs. restoring a VM from backup

[AngryAdm]

I have a few nodes.

I have a Domain controller on one node using 15GB storage space in a 220G RAW file where more than 90% is empty.
I attempt to migrate this to a different node... 3 hours later, still not done. I stop the process as my patience ran out.

I decide to Migrate it the fast way:
-Backup DC with PBS. 10 minuttes later.. done
-Detach drives from the VM and delete the disks. 30 seconds.. done
-Migrate the VM without disks (as I just deleted them) 30 seconds.. done
-Restore the VM from PBS 10 minuttes later.. done.

Why does it take 20 minuttes to do it the fast way, but 3 hours+(an unknown amount of time) to migrate the 15GB used space using the UI?

This does not make sense!

 

[Denny]

I can't replicate your results. I have a 3 node cluster. Migration network is on a 10gig network with a small microtik switch. Moving a node with a single disk of 108G takes about 4.5 minutes.

 

[fabian]

you haven't given details about what kind of storage you are using on either end, but it's likely some combination that needs to fully initialize/transfer the disk including empty blocks.. PBS and VMA backups will optimize that away.

 

[AngryAdm]

raw disks in datasets created as storage (directory) on zfs

 

[fabian]

is an offline migration faster? can you post a task log?