Thanks Greg and Yehuda for the feedback. I'll upgrade to 0.48.3 next chance I get. Sounds like rebooting osd.2 may not be a bad idea either -- might fix the map disagreement. - Travis On Tue, Feb 12, 2013 at 1:35 PM, Gregory Farnum <greg@xxxxxxxxxxx> wrote: > On Tue, Feb 12, 2013 at 9:36 AM, Yehuda Sadeh <yehuda@xxxxxxxxxxx> wrote: >> On Tue, Feb 12, 2013 at 6:50 AM, Travis Rhoden <trhoden@xxxxxxxxx> wrote: >>> Sébastien, >>> >>> Thanks so much for trying! >>> >>> I really think the problem is probably related to the error/crash I >>> had last week. A couple of other tidbits: >> >> Note that 0.48.2 that you're using has a known data corruption issue >> related to osd kernel/machine crashes. If you've seen any of these it >> might explain your issue. 0.48.3 solves that issue. > > This might be part of it, but don't forget the client and OSD are > disagreeing on what the correct target for an object is — that's > probably either another indicator of these issues (corrupt map on the > OSD?) or a pointer to the real issue. > Anyway, this is a problem of some kind with either the OSD or with > RADOS, not with RBD. > -Greg _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |