mergin two nodes to cluster

[quax]

searched through the forum, but coudln't find the answer.

situation: I have two pve nodes with _non-overlapping_ CTIDs. I wish to join them to one cluster (they are in the same subnet).

If I promote the one node to master and join with the another node to the new cluster, what happens to the existing vservers (openvz) on the slave node? My big concern is, that the slave-vservers get damaged due to sync-process.

How schall I proceed?

ps: sorry, I don't have a lab-setup to test it

update: pve versions on both nodes is in sync

 

[dietmar]

If I promote the one node to master and join with the another node to the new cluster, what happens to the existing vservers (openvz) on the slave node?

You need to backup all config files in /etc/qemu-sever and /etc/vz/conf on the slave node, and move them out of the way. Then join the cluster and move them back. But I would backup everything before doing such operations.

 

[quax]

Thanks for advise. Backups did I have, but after some simple tests and thoughts, I decided to try it without hiding existing configs. And it worked!

Thus for the log, since I couldn't find it wiki nor in the formum (please approve, that this method is a right way to do it):

Two nodes with two sets of non overlapping live-CTIDs on same subnet:
Promote the first hardware-node to master with pveca -c
Add the second hardware-node to the new created cluster with pveca -a -h <master-IP>
After few seconds of syncing everything is fine. all openvz nodes (didn't test ist with kvm) stay intact and operational and appear in the web-frontend on right place.

To get a clean result, I removed /etc/pve/master/ & /etc/pve/cluster.cnf on both nodes prior merging, which were left from previous testings.

version of both nodes: pve-manager/1.9/6567

 

[dietmar]

please approve, that this method is a right way to do it

This should normally complain about existing VMs - no idea why it worked for you.

 

[udo]

Thanks for advise. Backups did I have, but after some simple tests and thoughts, I decided to try it without hiding existing configs. And it worked!

Thus for the log, since I couldn't find it wiki nor in the formum (please approve, that this method is a right way to do it):

Two nodes with two sets of non overlapping live-CTIDs on same subnet:
Promote the first hardware-node to master with pveca -c
Add the second hardware-node to the new created cluster with pveca -a -h <master-IP>
After few seconds of syncing everything is fine. all openvz nodes (didn't test ist with kvm) stay intact and operational and appear in the web-frontend on right place.

To get a clean result, I removed /etc/pve/master/ & /etc/pve/cluster.cnf on both nodes prior merging, which were left from previous testings.

version of both nodes: pve-manager/1.9/6567

Hi,
and be sure that all required iso-images from the new slave are copied before to the new master. Otherwise they will be deleted (not the best conditions for an cd-bootable guest).

Udo