Transfer WinSrv vms linked clone without setting a cluster

[dmvillar]

Hi, We are replacing a pve 7.x server with a new one with pve 8.x.

For this, we want to transfer all the windows vms to the new server without generating a cluster, since the currently operational pve 7.x server will go offline once everything is moved to the new pve 8.x

The vms in pve 7.x are windows "linked cloned" lvm.

A simple option is to backup the vms and transfer them, but they would no longer be "linked cloned" which would load the disk unnecessarily.

Is there any documented procedure to follow to pass the base template and the "linked clone" machines?

 

[bbgeek17]

Hi @dmvillar ,
There is no documented procedure for such operation, regrettably. You may be able to achieve your goal by manipulating the backend storage directly. However, it depends on the type of backend storage you are using and a number of other external factors.
The easiest way to do what you want is to add PVE8 hosts into your existing cluster and transition in orderly manner. I.e. removing PVE7 hosts as you migrate off of them, so that you are left with PVE8 only at the end of your migration.

While mixing major releases is not thoroughly tested, if at all, and frowned upon for long term use, short-term use is ok.

Good luck

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[dmvillar]

Hi @dmvillar ,
There is no documented procedure for such operation, regrettably. You may be able to achieve your goal by manipulating the backend storage directly. However, it depends on the type of backend storage you are using and a number of other external factors.
The easiest way to do what you want is to add PVE8 hosts into your existing cluster and transition in orderly manner. I.e. removing PVE7 hosts as you migrate off of them, so that you are left with PVE8 only at the end of your migration.

While mixing major releases is not thoroughly tested, if at all, and frowned upon for long term use, short-term use is ok.

Good luck

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

Thanks for responding.
Yes, I have the pve 8 server in "laboratory" mode until we have defined the process and I have installed another server in this lab with pve7 to test the best process, I have tried generating a cluster from pve7 adding the pve8 node and doing the migrations from the GUI, but the "linked clone" vms seem to be passed complete (they are no longer "linked cloned").

Since the main use is 200 exactly the same vms, the "Linked clone" mode is the most efficient in terms of storage, for this reason we want to maintain this format.

The other problem we find with this is that once the "cluster" is generated, it has difficulties in "disarming" it.
For this reason we are trying to do another type of data passing and configurations without using the "cluster" relationship.

At the moment the server with pve7 works very well, but we want to replace it soon with the new server (with much more capacity) and pve8 installed

 

[bbgeek17]

Yes, I have the pve 8 server in "laboratory" mode until we have defined the process and I have installed another server in this lab with pve7 to test the best process, I have tried generating a cluster from pve7 adding the pve8 node and doing the migrations from the GUI, but the "linked clone" vms seem to be passed complete (they are no longer "linked cloned").

You do not mention the type of storage you are using. PVE7 and PVE8 shared storage and linked clone functionality do not differ, as far as I know. A simple "qm migrate" of either the template, the linked clone, or both should transfer the VM over to PVE8 as is.

Perhaps you can expand more on your configuration, as well as demonstrate what you are doing and your results. Use CLI and paste everything in as TEXT using CODE tags (</>).

Best

Full disclosure: I did not build a lab of a mixed pve7/8 cluster to test your use case. My opinion is based on my experience and reasonable expectations of system functionality. If you can provide examples of your exact steps and results that would be helpful.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[bbgeek17]

The other problem we find with this is that once the "cluster" is generated, it has difficulties in "disarming" it.
For this reason we are trying to do another type of data passing and configurations without using the "cluster" relationship.

Depending on your storage (again) you can take an offline window and re-integrate your storage and VMs by : a) shutting down all VMs b) disable storage on PVE7 c) create equivalent storage entry on PVE8 d) copy/recreate VM configs on PVE8

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[dmvillar]

Depending on your storage (again) you can take an offline window and re-integrate your storage and VMs by : a) shutting down all VMs b) disable storage on PVE7 c) create equivalent storage entry on PVE8 d) copy/recreate VM configs on PVE8

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

The storage is not shared, each server manages its storage locally, the current storage type of the vms is LVM-Thin local-lvm (once transferred to the new one, the storage will be migrated to ZFS.

If I copied all the tests I think I would fill the forum with inaccurate information, I will be doing tests connecting them in a cluster again and commenting on the results.

I will be sharing the tests that I carry out and the results obtained so that they remain as information


If anyone has already gone through this type of process, I would be very grateful to read the experience.

Regards

 

[bbgeek17]

The storage is not shared, each server manages its storage locally, the current storage type of the vms is LVM-Thin local-lvm (once transferred to the new one, the storage will be migrated to ZFS.

The fact that you are using local storage is the key here. The only way that I know of to transfer your data in its current format is either by moving the disk physically to new servers, or making a duplicate of the disk.

You can look into things like https://github.com/mpalmer/lvmsync and let us know whether it worked for you.

Good luck

PS keep in mind when you transfer your data from LVM to ZFS you will loose the linked-clone relationship as well.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[dmvillar]

The fact that you are using local storage is the key here. The only way that I know of to transfer your data in its current format is either by moving the disk physically to new servers, or making a duplicate of the disk.

You can look into things like https://github.com/mpalmer/lvmsync and let us know whether it worked for you.

Good luck

PS keep in mind when you transfer your data from LVM to ZFS you will loose the linked-clone relationship as well.

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

I understand, the goal is not to lose the "linked clone" even when migrating the storage to ZFS, I know I don't have an easy path but I will try to achieve this successfully and document it.

 

[Falk R.]

I understand, the goal is not to lose the "linked clone" even when migrating the storage to ZFS, I know I don't have an easy path but I will try to achieve this successfully and document it.

If Linked Clone does not work, then migrate to ZFS and activate Dedup. This may even be more effective, as all patches are stored independently after the clone, although it is always the same data.