On Sat, 8 Feb 2014, Christian Balzer wrote: > On Fri, 7 Feb 2014 19:22:54 -0800 (PST) Sage Weil wrote: > > > On Sat, 8 Feb 2014, Christian Balzer wrote: > > > On Fri, 07 Feb 2014 18:46:31 +0100 Christian Kauhaus wrote: > > > > > > > Am 07.02.2014 14:42, schrieb Mark Nelson: > > > > > Ok, so the reason I was wondering about the use case is if you were > > > > > doing RBD specifically. Fragmentation has been something we've > > > > > periodically kind of battled with but still see in some cases. > > > > > BTRFS especially can get pretty spectacularly fragmented due to > > > > > COW and overwrites. There's a thread from a couple of weeks ago > > > > > called "rados io hints" that you may want to look at/contribute to. > > > > > > > > Thank you for the hint. Sage's proposal on ceph-devel sounds good, so > > > > I'll wait for an implementation. > > > > > > > > > > Pardon me for stating the maybe painfully obvious, but wouldn't > > > setting the allocsize to 4MB (with XFS and the default Ceph object > > > size) do a world of good to prevent fragmentation? > > > > This is what we plan on doing, although I was thinking an allocation > > size of 1MB might be more appropriate as a default. In any case, > > though, the challenge is that not all objects are RBD objects, nor are > > all images using 4MB objects, so the OSD can't blindly do this; it needs > > to respond to a hint from the client. Ilya is working on this now. > > > Of course, for a generic implementation that would need to be done with > what was discussed in the "rados io hints" thread. > > In a use case where the sole usage would be RBD with the default object > size, mounting the XFS file systems with allocsize=4m might do the trick > for now though, right? If that option does what it sounds like it does, then yeah! sage _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |