Ceph-fs deployment

Hi,

I’m deploying a ceph cluster for hw nodes and having some trouble to deploy ceph-fs. I’m running in to problem where the ceph-mds-service doesn’t seem to start.
The cluster is up otherwise and health is OK. This is a deployment for only CEPH, no openstack or anything else related.

I’m deploying MDS as LXD for one node at the moment just to test it out. The LXD container is created with proper ip-addresses and I can access the container. The MDS-service fails to start because of missing keyring which it should create automatically from ceph auth. The user mds.juju-xxx is created and it has the correct CAPS-privileges. I have tried to add the keyring myself but I only end up on an error of monclient(hunting): handle_auth_bad_method server allowed_methods [2] but i only support [2]

My config for the ceph-fs-application looks like:

ceph-fs:
pool-type: erasure
ceph-osd-replication-count: 3
metadata-pool: cephfs_metadata
source: focal
use-syslog: true
ceph-public-network: XXX
loglevel: 5

I have the relations added after deploying the application, but the service just doesn’t seem to start. Any help for this?

Thanks,
-O

Hi, to get an overall picture of your deployment can you post the output to juju status ceph* --relations? Please border the output with three backticks to make the output render in a clear way (i.e. monospaced).

Also, I see that for option source you have ‘focal’, but it should either be ‘distro’ or a UCA OpenStack release such as ‘focal-wallaby’. Please see Ceph and the UCA.