Nod3 HS - Cluster : Hostname lookup 'nod3' failled

[rospionne]

In my test simulations,
I intentionally crashed the nod3 on a 3-node cluster to simulate the definitive death of node 3. I completely reinstalled the nod3 so that it is like "new" and virgin. With the new name 'nod3b' and a new IP address. My VMs have their disk on CEPH.

The goal of the game is to set up a procedure for replacing a node that is the least dangerous. If I delete an OSD by mistake, the VMs become unavailable. A monitor, the CEPH cluster collapses. To limit this risk, before removing 'node3' from the cluster and adding the replacement 'nod3b' to have a margin of error per critical service.

However, when I go to the datacenter, "Cluster" section. The buttons are blocked by an error message: Hostname lookup 'nod3' failled - failed to get address info for: nod3: Name or service not known (500)

Do I have to remove the 'nod3' first with the risk of manipulation that this implies or is there a safer procedure? Otherwise, is there a way to keep the availability of these buttons so that I can add the 'nod3b' and have the possibility to remove 'nod3' later?

Thanks in advance for your answers.

 

[rospionne]

My solution alternative, to make this error disappear is to still use the command:

pvecm delnode nod3

After running the command, the error is no longer. I checked if CEPH still showed the 'nod3' and that I could still intervene on the OSD later. The answer is 'YES'.

Note: once this command is executed, remember to delete all traces of the 'nod3' node in the files, once you have finished clearing CEPH.