On Fri, 1 Jul 2011, Gregory Farnum wrote: > On Thu, Jun 30, 2011 at 5:54 PM, Mark Nigh <mnigh@xxxxxxxxxxxxxxx> wrote: > > Yes, I did increase the max_mds prior to starting the cmds on the second node. Should I have started the daemon and then increase the mds? > Well if you increase the max_mds that tells the system to make more > active MDSes. "Active" means that the MDS is authoritative for part of > the namespace hierarchy, will be serving clients, etc. > If you just want standbys then you simply need to start up extra cmds processes. > > > I decrease it to one (1) and rebooted both cmds and they are still in replay. Is there a way to get them into active? > Unfortunately we don't have a way right now to reduce the number of > active MDSes. Most of the machinery is there but it's not complete or > well-tested. You've probably confused the system by telling it to have > fewer MDSes than it's already got, so you're going to have to put > max_mds back to 2 to get this cluster back up. There are two parts here: - 'ceph mds stop <num>' will tell the given mds rank to export its subtrees and leave the active set. The daemon will either shut down or go back to standby (I forget which :). - Setting max_mds to a lower value will prevent any new or standby MDSs from (re)joining the active set. The first part isn't yet part of our testing matrix but should work! sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
