[SOLVED] How to reinstall a cluster node with CT

[Vasu Sreekumar]

1. I have a cluster with 2 nodes. Each node has 5 CTs each. Proxmox is version 5.1.46.

2. Each node has two drives with ZFS, first one holds Proxmox (local and local-zfs). Other drive is attached as data to hold the CTs.

3. I take backup of /etc/pve/lxc/ folder from both nodes to another server.

4. Suppose if one node's first drive fails, what is best way to reinstall that node?

If it is stand alone node without cluster, all I have to do is .... reinstall promxox on first drive, attach second drive again as data, copy files back to /etc/pve/lxc/ folder and start each CT.

 

[Alwin]

Roughly and in generell yes. In a cluster, the configs under /etc/pve will be synced when the node will be added again to the cluster. But the failed node has to be removed from the rest of the cluster before you can join it again.

 

[Vasu Sreekumar]

1. I removed node from cluster
2. Reinstalled node, copied /etc/pve/lxc/ folder files to make all LXC guests come online.
3. Now I can not add node back to cluster, we can not add a node with existing guests.

So what is the work around to add node back to cluster?

 

[Alwin]

2. Reinstalled node, copied /etc/pve/lxc/ folder files to make all LXC guests come online.
3. Now I can not add node back to cluster, we can not add a node with existing guests.

If the new installed node has the same name, it has to be removed from the cluster first. Then you add the node, but it has to be empty (no vmid.conf files).
https://pve.proxmox.com/pve-docs/

 

[Vasu Sreekumar]

"Roughly and in generell yes."

So you answer was wrong? you said it is possible.

My original post clearly says node is not empty. Please read below.

1. I removed node from cluster
2. Reinstalled node, copied /etc/pve/lxc/ folder files to make all LXC guests come online. Node is not empty now, it has guests.
3. Now I can not add node back to cluster, we can not add a node with existing guests.

So what is the work around to add node back to cluster?

 

[Alwin]

When a node joins the cluster, the database of the joining node will be overwritten, with the state of the cluster. That's why the node needs to be empty (no configs). As you made a copy of your vmid.conf(s), you add them after joining the node to the cluster. Or restore your VM/CT from backup after joining.

 

[Vasu Sreekumar]

1. I am reinstalling a failed node, so I have no backup. Second drive has all guest images.
2. I have /etc/pve/lxc/ folder files backup.
3. So from fresh Proxmox installation on first drive, I will add second drive then I will join the cluster again, then I will copy /etc/pve/lxc/ folder files.

I hope this works, I will test it tomorrow.

 

[Alwin]

1. I am reinstalling a failed node, so I have no backup. Second drive has all guest images.

I added the restore your VM/CT, as the vmid.conf is added to the backup, to be able to restore the machine on a different PVE host. You can also extract the config from the backup (with restoring the whole VM/CT).

 

[Vasu Sreekumar]

It worked perfect.

I added a node with guests to the cluster successfully.

 

[Vasu Sreekumar]

So it turned out to be....

We can add a node with guests to a cluster, if we want.

 

[Vasu Sreekumar]

I have the issue solved.

Is there any a way to mark this post as SOLVED?

 

[Vasu Sreekumar]

OK I found it and marked solved.

 

[Vasu Sreekumar]

We tested same procedure on 4 more clusters to add new nodes with live guests.

All worked fine.