replace node in cluster

ildar94

New Member
Feb 24, 2025
5
1
3
Hi all!
I have a pve cluster and a cef storage of 3 nodes. On one node, both OSDs are in the down and out status.
I need to replace this node with another node.
I added a new node to the proxmox cluster, installed cef.
Tell me what are my next steps?
Do I need to delete the OSD on the node I want to remove from the cluster?
Or do I first need to create an OSD on a new node and add them to the pool?
 
Hi @ildar94

- First, add the new OSDs on your new node to the cluster
- Wait for the cluster to rebalance
- Only after rebalancing is complete, remove the failed OSD
- Remove Monitor / Managers from failed node
- Clean up the old node:
-- If you want to completely remove the node from the Ceph cluster, remove it from the CRUSH map: ceph osd crush rm {node-name}

Additional docs:

https://dzone.com/articles/how-to-remove-node-from-proxmox-cluster#:~:text=To remove a node from,and clean up Ceph components.&text=If you are learning Proxmox,is build a Proxmox cluster.

https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_remove_a_cluster_node
 
Last edited:
  • Like
Reactions: ildar94
Hi @ildar94

- First, add the new OSDs on your new node to the cluster
- Wait for the cluster to rebalance
- Only after rebalancing is complete, remove the failed OSD
- Remove Monitor / Managers from failed node
- Clean up the old node:
-- If you want to completely remove the node from the Ceph cluster, remove it from the CRUSH map: ceph osd crush rm {node-name}

Additional docs:

https://dzone.com/articles/how-to-remove-node-from-proxmox-cluster#:~:text=To remove a node from,and clean up Ceph components.&text=If you are learning Proxmox,is build a Proxmox cluster.

https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_remove_a_cluster_node
Hi Lukas,
Thank you a lot
 
I just created two OSDs on a new node. The new OSDs show up as ghost OSDs. Maybe I also need to add the monitor, manager and meta services to the new node to start the cluster rebalance process?

1740431092947.png
 
Hello @ildar94

for my curiosity simulated your issue with to reproduce steps.

Original cluster: proxmox1A, proxmox1B, proxmox1C
1) Simulated failure on proxmox1A - detached disks with CEPH OSDs
2) Created new node (proxmox1d) and add in to the cluster + install CECP
3) Add new OSDs on (proxmox1d)

Mo issues or data los observed

1740563309395.png

1740563409407.png

How is now your situation ?

L.
 
  • Like
Reactions: ildar94
Hello @ildar94

for my curiosity simulated your issue with to reproduce steps.

Original cluster: proxmox1A, proxmox1B, proxmox1C
1) Simulated failure on proxmox1A - detached disks with CEPH OSDs
2) Created new node (proxmox1d) and add in to the cluster + install CECP
3) Add new OSDs on (proxmox1d)

Mo issues or data los observed

View attachment 82913

View attachment 82915

How is now your situation ?

L.
Hi Lukas,
Thank you for you help.
I just had to reboot the new node. After that I saw new OSDs and the rebalancing process started.
 
  • Like
Reactions: Johannes S