Hi Martin, > I too would happily welcome this potential in Ceph. It is evidently > quite attractive, obviously due to the plain number of Windows machines > just sitting on LAN:s with disks under-utilized, idleing. While it might > not be the perfect use case for Ceph, hopefully it would be doable? > > I'm seeing some possible issues with nodes going offline and online in a > higher-than-optimal frequency, but the idea is indeed attractive!
Ceph is actually relatively tolerant of this. The PG logs make resynchronizing replicas very fast. And you can adjust the time threshold before a "down" node is considered "out" and data is replicated to a new node (more or less analogous to a delay to stay in degraded mode before the hot spare is used in a RAID array). A larger issue is often that Ceph tends to treat the network as flat, which makes it relatively unsuitable for situations where a WAN is involved. There are some tricks to be played with CRUSH, but I think it only really makes sense for primary/secondary site situations. sage ------------------------------------------------------------------------- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://sourceforge.net/services/buy/index.php _______________________________________________ Ceph-devel mailing list Ceph-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ceph-devel
