Ceph failure - missing /etc/pve/ceph directory

bagels1b

Member
Nov 7, 2022
19
2
8
I have an 8.1.4 ProxMox Ceph cluster. I'm trying to add a new host that is running the latest 8.2.2 to the cluster but when I try and add the OSD I get the failure "pveceph configuration not initialized - missing '/etc/pve/ceph'". There's a commit that went in on Feb 16, 2024 that looks like it adds this directory. From the description it looks like it will create this directory as part of ceph init on new installs (after 8.1.5) and create it if upgrading from a previous release. But my scenario isn't either of those since I'm not running ceph init or upgrading a host already in the cluster. Is this a valid scenario that should be handled?

As a workaround, I manually added the directory /etc/pve/ceph to an existing 8.1.4 node in the cluster and then running 'pveceph osd create /dev/sdb' on the new 8.2.2 host is successful. Does anyone know if this is a valid workaround or will I potentially run into issues down the road? The /etc/pve/ceph directory is currently empty after adding the OSD but it appears to have been added properly.

Thanks
 
  • Like
Reactions: jdw
Thanks, I just ran into this as well!

Creating the (empty) missing directory by hand seems to have worked here without any ill effects (so far).