CEPH cluster_network defaults to a single node IP rather than to a subnet

[kiwistagnz]

I've noticed that in the CEPH Configuration on /etc/pve/ceph.conf in [global], cluster_network it sets the IP as one specific node's IP rather than the subnet. This doesn't matter which node we log into - it's still the same.

For example:

[global]
        auth_client_required = cephx
        auth_cluster_required = cephx
        auth_service_required = cephx
        cluster_network = 10.1.2.3/24
....

Reading up on documentation, cluster_network should be 10.1.2.0/24

Has anyone else struck this or is this some sort of bug where the last CEPH node writes itself as the IP?

The same issues exists on a few clusters I have checked.

 

[SteveITS]

AFAIK that’s normal in PVE. And equal?

 

[Alwin Antreich]

Reading up on documentation, cluster_network should be 10.1.2.0/24

When you run the ceph install wizard, the configuration is done only once. And yes, the PVE node adds it's own IP in CIDR notation. Ceph is clever enough to get the network range from it.

Has anyone else struck this or is this some sort of bug where the last CEPH node writes itself as the IP?

The `pveceph init` step should only be run once and not in parallel, as yes, the last write of the file will win.