On Wed, Apr 22, 2015 at 11:04 AM, J David <j.david.lists@xxxxxxxxx> wrote: > On Thu, Apr 16, 2015 at 8:02 PM, Gregory Farnum <greg@xxxxxxxxxxx> wrote: >> Since I now realize you did a bunch of reweighting to try and make >> data match up I don't think you'll find something like badly-sized >> LevelDB instances, though. > > It's certainly something I can check, just to be sure. Erm, what does > a LevelDB instance look like? After poking arond in the contents of > one of the OSD directories, I see what look like the PG directories > under current, but nothing labelled "LevelDB" jumps out. Uh, looks like it's the contents of the "omap" directory (inside of "current") are the levelDB store. :) > >> Final possibility which I guess hasn't been called out here is to make >> sure that your CRUSH map is good and actually expected to place things >> evenly. Can you share it? > > Here is the crush map: > > http://pastebin.com/yBtZFM6r > > It is pretty default; I don't think we (are smart enough to) have done > anything custom to it. Ah — I think you might be suffering from some of the issues that prompted the creation of the straw2 algorithm, since you have two close-but-different OSD sizes, a bunch of same-sized hosts, and one that's different. I could be wrong, but whenever you do upgrade to hammer you might want to pay the data movement price of making that change. (There are discussions in the release notes and elsewhere about this that you can look up.) -Greg _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |