Almost live migration between two differents nodes (version 3 --> 4 example) with NFS

[whitewater]

Hello,
I migrate VMs from a cluster PVE 3 to a PVE 4.
Ideally, it should use a backup / restore of each VM from the old node proxmox 3 (A), to the new node proxmox 4 (B).

Here's an idea to have very little downtime :
Migrate hard drives between each node with NFS storage, then, stop VM on old node and restart it on the new.

Here a little tuto :
Node B : install and config NFS server.

apt-get install nfs-kernel-server

think nfs storage will be on /var/lib/vz/nfs :

mkdir /var/lib/vz/nfs

configuration for nfs, add this line to /etc/exports :

/var/lib/vz/nfs    <ip of client, node A>/24(rw,no_root_squash,anonuid=1000,anongid=1000,sync,no_subtree_check)

then restart the service :

service nfs-kernel-server restart

now add this nfs storage on both node with disk image attribute.

Then, for each VM :
- migrate hard drives node A to B by this NFS storage.
- In node B : Copy the VM conf file (or copy / paste to a new file) to /etc/pve/qemu-server, adjusting hard drives line in this file, for pointing to nfs storage, with VM number as folder, like "local storage".

For exemple, if you have :

virtio0: drbdr0:vm-102-disk-1,cache=writethrough,size=165964M

change it by :

virtio0: nfs:102/vm-102-disk-1,cache=writethrough,size=165964M

- When migration is finished, shutdown the VM on the node A. When it is stopped, start it on node B.
- Correct "start on boot" option if needed.
- Last migration of hard drive to another storage.

Et voila
certainly useful for other uses.