Ceph PVE dashboard Object store zonegroup error.

alexisshaw

New Member
Jan 21, 2025
5
2
3
Hi

I have been recently setting up a new cluster, and decided to use proxmox because it looked less daunting then spinning up a bare metal talos, rook-ceph kubernetes and kubevirt cluster.

One key requirement of this setup is it must support hyperconverged object storage. With minio deciding it wants to own everything hardware wise, the ceph object gateway sounded like just the ticket. However with pve not officially supporting it the setup process is obviously quite manual. Especially surrounding the keychains in the /etc/pve/private directory being immutable.

I managed to get almost everything working, however when setting up the dashboard I have been having an issue where the object gateway throws errors when listing the realms, zonesgroups and zones. Looking at the exception trace in the logs this seems to fault inside a function that calls the radosgw-admin binary. However the logging detail does not share enough details.

Debugging the function call i am pretty sure it is a permissions issue surrounding a keychain.

Specifically the crashes happen in
RgwMultisite.list_zonegroups(self) and
RgwMultisite.get_multisite_sync_status(self, daemon_name: str)

Both of these functions call MgrModule.send_rgwadmin_command which runs radosgw-admin.

Anyone who has an idea how to solve the issue would be appreciated.
 
That, and this was also useful

 
  • Like
Reactions: blackline
Also, you need to set you access key and secret key for the gateway in the dashboard using these:

ceph dashboard set-rgw-api-access-key -i access_key
ceph dashboard set-rgw-api-secret-key -i secret_key

There were a couple of other permissions issues that I solved by moving a few of the keys into the /etc/ceph folder, and ammending my ceph.conf file
 
  • Like
Reactions: blackline