PVE to PVE migration

T

tralce

Member
Jan 16, 2022
2
0
6
36
Hi all. I'm looking for recommendations.

I am preparing to migrate the VMs on a four-node cluster onto one new Proxmox server. The existing four nodes are running Proxmox VE 7.2-4. All of the VMs are stored in Ceph, which is running on two of the nodes. The new server - which has not been configured yet - will be running Proxmox VE 8.1, or whatever the current ISO version is when we provision the server.

My question is, what would be the best strategy to migrate the VMs to the new machine? Concerns include downtime, speed of migration, and data integrity, as if to imply these are ever not a concern.

Thanks!
 
One way to migrate would be to export a part of your new storage as NFS to the old cluster, live migrate the disk to the NFS storage, then shutdown the VM, copy the VM configuration to the new node, start it on the new cluster (downtime as long as the OS shuts down / boots up) and then live migrate the disk to the target storage (e.g. also ceph). The storage name in PVE should be the same for the mounted NFS on the old cluster and the configure directory storage on the new cluster, so that you can just copy the VM config without changing the storage directory.
 
  • Like
Reactions: alexskysilk, Dunuin and Kingneutron
LnxBil said:
start it on the new cluster (downtime as long as the OS shuts down / boots up) and then migrate the disk to the target storage (e.g. also ceph). The storage name in PVE should be the same for the mounted NFS on the old cluster and the configure directory storage on the new cluster, so that you can just copy the VM config without changing the storage directory.
Click to expand...
Op said they only have 1 new server. So not a cluster or Ceph candidate. Which is why it would be important to understand whether the old nodes (which are also running Ceph) are staying and the new node gets added to existing cluster, or whether its a inter-cluster/node migration to a completely different storage.

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
Last edited:
bbgeek17 said:
Op said they only have 1 new server. So not a cluster or Ceph candidate. Which is why it would be important to understand whether the old nodes (which are also running Ceph) are staying and the new node gets added to existing cluster, or whether its a inter-cluster/node migration to a completely different storage.
Click to expand...
Yes, of course it's important and you're right. We still wait for the answer from the OP. My answer was more a general cluster2cluster migration despite their used storage technology and with a dedicated shared storage each.
 
tralce said:
My question is, what would be the best strategy to migrate the VMs to the new machine? Concerns include downtime, speed of migration, and data integrity, as if to imply these are ever not a concern.
Click to expand...
Step 1. make backups.
step 2. restore on new infrastructure.

downtime is a function of incurred deltas. Static vms can just be flipped in place (turn old one off, turn new one on) and more dynamic can have the outage time minimized by either syncing the difference from old to new, or by using differential backups.
 
bbgeek17 said:
The following aspects need clarification:
- What are your plans for the old 4 nodes when/if you migrate everything to a single new one?
- What are your plans for backend storage on the new node?


Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
Click to expand...
I am hoping that one of the old nodes can become host to a Proxmox Backup Server, but essentially all the existing nodes will be decommissioned post-migration.

The new node has a large JBOD of SSDs, so we're planning to use LVM or ZFS - haven't fully committed to one yet, as we're at the end of the planning stage.

Sincerest apologies for my delay. I had mistakenly believed that this platform would email me reply notifications, as I ticked the corresponding box.
 
tralce said:
The new node has a large JBOD of SSDs, so we're planning to use LVM or ZFS - haven't fully committed to one yet, as we're at the end of the planning stage.
Click to expand...
So sounds like you are not planning a Proxmox HA cluster for your data (where downtime and data integrity were a concern). Or perhaps you plan to add other nodes later, but you are _not_ planning to join this node into existing cluster.

In that case, as mentioned by various contributors to this thread, you have a few options:
- remote migration
- nfs reach-around
- backup and restore

the last two allow you to pre-migrate and measure aspects of the process. The first one is experimental and does not allow for non-disruptive pre-migration. However you can clone a production VM to then gain experience with remote migration.

Good luck

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
  • Like
Reactions: alexskysilk
LnxBil said:
It does. I got a mail that you replied. Maybe something wrong with your settings?
Click to expand...
You have to "watch" a thread and it will ask you if you want additional email notifications or not. There should also be an option to auto-watch threads you contributed to.
 
You must log in or register to reply here.
Top Bottom