How can I force add two clusters with virtual machines running?

[hellomotow]

I want to know how I can force add two machines together that already have running virtual servers.

I have been careful and planned for this in advance each node uses different VM ids and users usernames are unique too so there will be no conflict there.

What are the risks and what exactly do I need to do? All I think I currently need to do is the following:

move/delete /etc/pve/nodes/XXX/qemu-server files to main node (second node will show as being empty)
copy the user.cfg file to main node

 

[wolfgang]

Hi,

this is not supported, and no guarantee is given that this will work.

It is correct that you must move the CT/VM config but not to the other node.

Let's say this operation is done on Node B:
I would recommend you to copy the entire /etc/pve to the user dir.
So If you forget something, you still have a copy of it.
Then you can use the "pvecm add <Node A> [options] --force"
The options are necessary for the link settings if you use different ones.
The force flag will delete(backup and remove) all settings (VM/CT config) on Node B.

If this is done, you two nodes are joint together.
now you have to move the VM/CT config back to the corresponding dir at /etc/pve/nodes/<NodeB>
/...
After this, you can merge the user.cfg, but this has to be done manually because users are cluster-wide the same file.
Don't forget the shadow.cfg in /etc/pve/priv/ because there are the hashed passwords included, and without you can't log in.
shadow.cfg has also be merged because it is also cluster-wide the same.

If you have a firewall configured, there are also extra files.

 

[hellomotow]

Hi,

this is not supported, and no guarantee is given that this will work.

It is correct that you must move the CT/VM config but not to the other node.

Let's say this operation is done on Node B:
I would recommend you to copy the entire /etc/pve to the user dir.
So If you forget something, you still have a copy of it.
Then you can use the "pvecm add <Node A> [options] --force"
The options are necessary for the link settings if you use different ones.
The force flag will delete(backup and remove) all settings (VM/CT config) on Node B.

If this is done, you two nodes are joint together.
now you have to move the VM/CT config back to the corresponding dir at /etc/pve/nodes/<NodeB>
/...
After this, you can merge the user.cfg, but this has to be done manually because users are cluster-wide the same file.
Don't forget the shadow.cfg in /etc/pve/priv/ because there are the hashed passwords included, and without you can't log in.
shadow.cfg has also be merged because it is also cluster-wide the same.

If you have a firewall configured, there are also extra files.

I understand this is not supported but what could really go wrong? If I do what you set out there should be no other risks?

 

[wolfgang]

I understand this is not supported but what could really go wrong?

I guess the worth thing is if you have backups of /etc/pve/ you have to reset the Proxmox Cluster FS and start from the beginning.

If I do what you set out there should be no other risks?

The most important thing is that you have a copy of the /etc/pve.
The VM data on the storage is safe. Also, your running VM/CT stays up.