On Mon, Jan 5, 2015 at 10:56 AM, debian Only <onlydebian@xxxxxxxxx> wrote: > I try myself, use these command and recover to Health Ok now. But i do not > know why were these command work, in my opinion , fail mds node first and rm > failed mds node > > root@ceph01-vm:~# ceph mds fail 0 Okay, good -- this is telling ceph that the daemon that was serving rank 0 will not be heard from again (use this after actually stopping the daemon). > root@ceph01-vm:~# ceph mds rm 0 mds.ceph06-vm > mds gid 0 dne This command is a no-op, as you're asking it to remove a daemon with id '0' from the map, and no such daemon is in the map. You do not need to do this. The second argument here is also being ignored, as "mds rm" takes a single argument (MDS daemon GID) > root@ceph01-vm:~# ceph mds newfs 1 0 --yes-i-really-mean-it > filesystem 'cephfs' already exists Again, `newfs` has nothing to do with managing individual MDS daemons or ranks. It is a global operation that recreates a whole filesystem. > root@ceph01-vm:~# ceph mds rmfailed 0 You should not need to do this if you stopped the daemon before doing the "fail 0" earlier. I have just tried this myself to check the procedure, and it is really simple: * Stop the old MDS * Run "ceph mds fail 0" * Start the new MDS * Clean up the local daemon files for the old MDS You do not need "mds rm", "mds rmfailed", "newfs" or any of that! Regards, John _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |