On Thu, Jan 23, 2014 at 3:35 AM, bf <bf31415@xxxxxxxxx> wrote: > Gregory Farnum <greg@...> writes: > > >> Yes, Ceph does all the heavy lifting. Multiple PGs with the same OSDs >> can happen (eg, if you only have two OSDs, all PGs will be on both), >> but it behaves about as well as is possible within the configuration >> you give it. >> -Greg >> Software Engineer #42 <at> http://inktank.com | http://ceph.com >> > > Thanks Greg. A few more questions-- again maybe obvious. > > If a pool defines a num of copies as 6, does that mean each PG will have > 6 and only 6 OSDs? Yes. > How is the primary OSD determined per PG? CRUSH determines that — at the moment it's simply the first OSD in the generated list, but that's subject to change. > When an OSD within a PG fails, presumably the PG would be operating outside > of the target diversity/redundancy of the pool. Does that PG start to search > for another OSD that can be brought into the PG to satisfy the pool diversity/ > redundancy requirements? Is there a way to tell when a PG is in this hazard > state? If an OSD gets marked out, then the map has changed and the CRUSH algorithm outputs a different OSD in its place. There's no searching involved, just a computation. Whenever a PG is not fully replicated, it's marked "degraded" in the pg map. For further details, you should check out the documentation (ceph.com/docs) or the research papers (http://ceph.com/resources/publications/). -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |