Moving VM's between Proxmox VE 3.x and 4.x cluster

[wosp]

Currently we have a Proxmox VE 3.x cluster with 3 nodes (node1, node2 and node3). We want to upgrade this cluster to Proxmox VE 4.x but with the lowest possible downtime and, we don't want to upgrade but do a clean install. Since we also needed some more physical resources, we decided to buy 2 additional nodes and install them with Proxmox VE 4.x and also installed a VM with Proxmox VE 4.x on the current Proxmox VE 3.x cluster, to create the new cluster (node4, node5 and node6). The storage we use is an external Ceph cluster (not on the Proxmox VE nodes, but dedicated storage nodes). So we have added the Ceph storage also to our Proxmox VE 4.x cluster now and some test VM's are working fine. Now we want to move the VM's on the Proxmox VE 3.x cluster to the new Proxmox VE 4.x cluster, after which we can re-install the Proxmox VE 3.x nodes with Proxmox VE 4.x and let them join the new cluster (and remove the virtual node6 from the new Proxmox VE 4.x cluster).

We did some testing and came to the following method. In this example the VM that needs to be moved have ID '100' and is running currently on node1 (Proxmox VE 3.x) and needs to be migrated to node4 (Proxmox VE 4.x).

1. Stop HA of the VM on Proxmox VE 3.x cluster
2. Stop the VM itself on Proxmox VE 3.x cluster
3. On one of the nodes in the Proxmox VE 3.x cluster execute:

# scp /etc/pve/nodes/node1/qemu-server/100.conf root@node4:/etc/pve/nodes/node4/qemu-server/100.conf

4. Enable HA for the moved VM on the Proxmox VE 4.x cluster
5. VM will be booted automatically and is online again.
6. On one of the nodes in the Proxmox VE 3.x cluster execute:

# rm /etc/pve/nodes/node1/qemu-server/100.conf

Do I forget something or is the method above OK? Thank you!

 

[udo]

Hi,
this will work if the storage and network config is the same on both cluster (but I don't have experiences with HA).
One remark for a shorter downtime:
scp the config before you shutdown the VM on node1, then you can start the VM on node 4 just after the shutdown on node1 is finished.

But be shure that you don't create an VM with an VMID which is used on the old cluster - if you delete them you have an high risk to delete the VM-files from the old cluster (because two cluster use the same storage!).

Udo

 

[ghusson]

Advised method is backup on old PVE / restore on new PVE.
It implies more downtime but you are sure to get a good migration.
You don't ever know if the VM config files have new defaults settings or format change...

 

[wosp]

Backup and restore is taking to much downtime. That's simply no option. In case this is really needed we better do a upgrade to Proxmox VE 4.x of the current Proxmox VE 3.x cluster and then join the new nodes in the current cluster, migrate the VM's and re-install the "old" nodes.

But I would like to get an answer from a Proxmox VE staff member. Is there anything changed in VM .conf files between Proxmox VE 3.x and 4.x?

Edit:

Just did a new test and found out only 1 change:

backup=no

Is changed to:

backup=0

However, even with "backup=no" Proxmox VE 4.x still shows backup is disabled for this harddisk. When enabling backup and then disable again, the configfile is changed to "backup=0". So seems not to be a big issue. Anyone from Proxmox VE team with more info?

 

[udo]

Backup and restore is taking to much downtime. That's simply no option. In case this is really needed we better do a upgrade to Proxmox VE 4.x of the current Proxmox VE 3.x cluster and then join the new nodes in the current cluster, migrate the VM's and re-install the "old" nodes.

But I would like to get an answer from a Proxmox VE staff member. Is there anything changed in VM .conf files between Proxmox VE 3.x and 4.x?

Edit:

Just did a new test and found out only 1 change:

backup=no

Is changed to:

backup=0

However, even with "backup=no" Proxmox VE 4.x still shows backup is disabled for this harddisk. When enabling backup and then disable again, the configfile is changed to "backup=0". So seems not to be a big issue. Anyone from Proxmox VE team with more info?

Hi,
I'm not from the proxmox team, but have migrated till now some (22) VMs without trouble with thus method.

Udo

 

[wosp]

Thanks Udo,

Last weekend we did migration of +/- 50 VMs without any issue. In a couple of weeks we will migrate the remaining VM's also with less then 1 minute downtime