On Wed, Aug 3, 2011 at 12:46 PM, H. Andres Lagar-Cavilla <[email protected]> wrote: > Hi all, > I''ve been looking at ceph for a little while and following the mailing list. > We are thinking of using just rados (i.e. no mds) for a data warehouse that > does not need posix compliance. Three questions come to mind > > 1. what is a good kernel+backing filesystem combo to use on each individual > rados node. Recently people have been reporting xattrs issues with 3.0 and > ext4, and performance leaks with btrfs. What is recommended by core ceph > developers as stable?
We generally recommend the use of btrfs with ceph, as there are various optimizations that ceph employs with it. We use btrfs on 2.6.38+ and it's reasonably stable. > > 2. Layout: we have 10 beefy nodes (maybe more, stick with 10 for now). I'd > like to use all ten for osds, but we also need monitors. So: > 2.1. Do monitors need beefy machines (I guess not) Usually no. > 2.2. two/three monitors is ok for O(10) osds? Three are needed to form a quorum. > > 3. API examples. The rgw looks like a prime example of a chunk of code that > interfaces directly with rados. Any other examples? librbd? > the 'rados' tool, librbd, testrados and testradospp (compile with --with-debug). Yehuda -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
