I haven't tested it but even without that, I don't think it is a good idea to have another Ceph deployment tool besides PVE as there are likely some conflicts.
As I mentioned, on a hyperconverged PVE Ceph cluster, PVE is handling the Ceph deployment. Introducing another tool will most likely break stuff. One thing that I expect to break is the delivery of up to date ceph.conf files. PVE is using the pmxcfs to distribute it across the cluster and /etc/ceph/ceph.conf is only a symlink to /etc/pve/ceph.conf.
Using ceph-adm which deploys these via SSH (AFAICT) will most likely overwrite the /etc/ceph/ceph.conf file and thus breaking the symlink and PVEs assumption that all nodes have the same ceph.conf file.
I had to create osds also, which have the same issue, so I found this ceph-adm stuff.
It also seems to be easier to reload the ceph services after upgrade.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.