On Wed, Jan 22, 2014 at 10:43 PM, Schlacta, Christ <aarc...@aarcane.org> wrote: > can ceph handle a configuration where a custer node is not "always on", but > rather gets booted periodically to sync to the cluster, and is also > sometimes up full time as demand requires? I ask because I want to put an > OSD on each of my cluster nodes, but some cluster nodes only come up as > demand dictates, but are offline most of the time. It would be a trivial > matter to have them cycle on once or twice a day for the purposes of syncing > ceph, then shut down again if unneeded, but running them full time when > there's simply no demand would be wasteful. > > dmick on IRC says he can't imagine ceph handling it cleanly, but I was > hoping someone here might have some more knowledge or experience to help > guide me.
It's not likely to work out well, although with a sufficiently custom configuration and use case you might succeed. The issue is that Ceph is built for online always-on replication; it's not using an AFS or XtreemFS model of copying data around. When you turn those extra nodes off, it's going to notice, and (by default), eventually mark them "out". Even if you disable that, the logs aren't big enough for log-based recovery to work unless you don't have any write activity happening, which means it's going to have to backfill the data to the "on-demand" OSDs every single time they turn on, and that's just not going to work out well when the whole reason you're turning them on is because demand has just increased. -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
