[PVE-User] Online migration problems with pve 4.2

[Albert Dengg]

hi,
On Thu, May 12, 2016 at 08:45:30AM +0200, Eneko Lacunza wrote:
> Hi Albert,
> 
> I just updated yesterday afternoon our cluster to PVE 4.2 non-subscrition
> and migration works OK. Did so using your procedure.
> 
> I found migration problems about a month ago, you can search in the mailing
> list the good advice I got from the developers.
i think i know wich thread you mean, and they advised you to upgrade
to
qemu-server 4.0-59
pve-qemu-kvm 2.5-7

however, i'm already running
qemu-server: 4.0-72
pve-qemu-kvm: 2.5-14

> 
> Can you attach the full migration task log? What storage are you using?
here is the log from the latest try:
task started by HA resource agent
May 12 12:10:26 starting migration of VM 105 to node 'pve1'
(172.23.22.1)
May 12 12:10:26 copying disk images
May 12 12:10:26 starting VM 105 on remote node 'pve1'
May 12 12:10:28 starting ssh migration tunnel
May 12 12:10:29 starting online/live migration on localhost:60000
May 12 12:10:29 migrate_set_speed: 8589934592
May 12 12:10:29 migrate_set_downtime: 0.1
May 12 12:10:31 ERROR: online migrate failure - aborting
May 12 12:10:31 aborting phase 2 - cleanup resources
May 12 12:10:31 migrate_cancel
May 12 12:10:33 ERROR: migration finished with problems (duration
00:00:07)
TASK ERROR: migration problems

i also saw the task starting the vm on the new node and starting the
migration tunnel, which succeeded.

as for storage:
we are using ceph storage for the VMs

just for the record:
i reran some test just now:
* offline migration works (i have a test vm that i could use for
  that
* even though i got an error yesterday, it worked just now for a
  freshly booted VM

i still would like to find a way that does not involve shutting down
all VMs to get migration working again...the next possible time for
that would be at some point end of next week most likly...

> 
> Not really related to this, but you should remove 2.6.32 and 3.10 kernels,
> they're not able to boot PVE 4/Debian jessie as far as I've tested.
yes i know, i just saw it...however i really don't think non-running
kernels are the problem here...

> 
> Versions on our cluster:
> # pveversion -v
> proxmox-ve: 4.2-49 (running kernel: 4.4.8-1-pve)
> pve-manager: 4.2-4 (running version: 4.2-4/2660193c)
hm...slightly newer, however i think that's becaus you are running
non-subscription..

thanks for all the pointers sofar...

regards,
albert
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.proxmox.com/pipermail/pve-user/attachments/20160512/2a4a18c9/attachment.sig>