On Mon, Dec 2, 2013 at 9:03 PM, Miguel Afonso Oliveira <m.a.oliveira@xxxxxxxxxxx> wrote: > Hi, > > Sorry for the very late reply. I have been trying a lot of things... > > > On 25/10/13 22:40, Yan, Zheng wrote: >> >> On Sat, Oct 26, 2013 at 2:05 AM, Gregory Farnum <greg@xxxxxxxxxxx> wrote: >>> >>> Are you sure you're using only CephFS? Do you have any snapshots? >>> -Greg >>> Software Engineer #42 @ http://inktank.com | http://ceph.com >>> >>> >>> On Fri, Oct 25, 2013 at 2:59 AM, Miguel Afonso Oliveira >>> <m.a.oliveira@xxxxxxxxxxx> wrote: >>>> >>>> Hi, >>>> >>>> I have a recent ceph deployment with version: >>>> >>>> ceph version 0.67.4 (ad85b8bfafea6232d64cb7ba76a8b6e8252fa0c7) >>>> >>>> on 4 12TB OSDs: >>>> >>>> GLOBAL: >>>> SIZE AVAIL RAW USED %RAW USED >>>> 49143G 8285G 40858G 83.14 >>>> >>>> POOLS: >>>> NAME ID USED %USED OBJECTS >>>> data 0 20396G 41.50 7342052 >>>> metadata 1 276M 0 81826 >>>> rbd 2 0 0 0 >>>> >>>> and this morning I started to get a warning about a full OSD: >>>> >>>> cluster 14320bfb-8b8c-4280-afee-df63172b1d0c >>>> health HEALTH_WARN 1 near full osd(s) >>>> monmap e3: 3 mons at >>>> >>>> {gridio1=10.112.0.148:6789/0,gridio2=10.112.0.149:6789/0,gridio3=10.112.0.150:6789/0}, >>>> election epoch 44, quorum 0,1,2 gridio1,gridio2,gridio3 >>>> osdmap e498: 4 osds: 4 up, 4 in >>>> pgmap v485463: 6144 pgs: 6142 active+clean, 2 >>>> active+clean+scrubbing+deep; 20396 GB data, 40858 GB used, 8285 GB / >>>> 49143 >>>> GB avail; 2252B/s wr, 0op/s >>>> mdsmap e54: 1/1/1 up {0=gridio4=up:active} >>>> >>>> However when I use a du on the mount point I get: >>>> >>>> [root@ce01 /]# du -bsh grid/ >>>> 31G grid/ >>>>>> >>>>>> [root@ce01 /]# du -bsh grid/ >>>>>> 31G grid/ >>>> >>>> what is the output of 'getfattr -d -m - grid/' ? > > > [root@ce01 ~]# getfattr -d -m - /grid > getfattr: Removing leading '/' from absolute path names > # file: grid > ceph.dir.layout="stripe_unit=4194304 stripe_count=1 object_size=4194304 > pool=data" > > > >> sounds like the 'purge strays' bug. try umounting all clients and >> restarting the mds. > > > I think you have nailed it on the head! I have now updated to: > > ceph version 0.72.1 (4d923861868f6a15dcb33fef7f50f674997322de) > > but I still see the same behavior. Is there anything else I can do other > than > keep having to do this every time until the bug is solved? Any idea when > will > that be? Next release? If your issue is caused by the bug I presume, you need to use the newest client (0.72 ceph-fuse or 3.12 kernel) Regards Yan, Zheng > > Cheers, > > MAO > > > _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |