On Mon, Nov 25, 2013 at 8:10 AM, Laurent Barbe <laur...@ksperis.com> wrote: > Hello, > > Since yesterday, scrub has detected an inconsistent pg :( : > > # ceph health detail (ceph version 0.61.9) > HEALTH_ERR 1 pgs inconsistent; 9 scrub errors > pg 3.136 is active+clean+inconsistent, acting [9,1] > 9 scrub errors > > # ceph pg map 3.136 > osdmap e4363 pg 3.136 (3.136) -> up [9,1] acting [9,1] > > But when I try to repair, osd.9 daemon failed : > > # ceph pg repair 3.136 > instructing pg 3.136 on osd.9 to repair > > 2013-11-25 10:04:09.758845 7fc2f0706700 0 log [ERR] : 3.136 osd.9 missing > 96ad1336/rb.0.32a6.238e1f29.000000034d6a/5ab//3 > 2013-11-25 10:04:09.759862 7fc2f0706700 0 log [ERR] : repair 3.136 > 96ad1336/rb.0.32a6.238e1f29.000000034d6a/5ab//3 found clone without head > 2013-11-25 10:04:12.872908 7fc2f0706700 0 log [ERR] : 3.136 osd.9 missing > e5822336/rb.0.32a6.238e1f29.000000036552/5b3//3 > 2013-11-25 10:04:12.873064 7fc2f0706700 0 log [ERR] : repair 3.136 > e5822336/rb.0.32a6.238e1f29.000000036552/5b3//3 found clone without head > 2013-11-25 10:04:14.497750 7fc2f0706700 0 log [ERR] : 3.136 osd.9 missing > 38372336/rb.0.32a6.238e1f29.000000011379/5bb//3 > 2013-11-25 10:04:14.497796 7fc2f0706700 0 log [ERR] : repair 3.136 > 38372336/rb.0.32a6.238e1f29.000000011379/5bb//3 found clone without head > 2013-11-25 10:04:57.557894 7fc2f0706700 0 log [ERR] : 3.136 osd.9 missing > 109b8336/rb.0.32a6.238e1f29.00000003ad6b/5ab//3 > 2013-11-25 10:04:57.558052 7fc2f0706700 0 log [ERR] : repair 3.136 > 109b8336/rb.0.32a6.238e1f29.00000003ad6b/5ab//3 found clone without head > 2013-11-25 10:17:45.835145 7fc2f0706700 0 log [ERR] : 3.136 repair stat > mismatch, got 8289/8292 objects, 1981/1984 clones, 26293444608/26294251520 > bytes. > 2013-11-25 10:17:45.835248 7fc2f0706700 0 log [ERR] : 3.136 repair 4 > missing, 0 inconsistent objects > 2013-11-25 10:17:45.835320 7fc2f0706700 0 log [ERR] : 3.136 repair 9 > errors, 5 fixed > 2013-11-25 10:17:45.839963 7fc2f0f07700 -1 osd/ReplicatedPG.cc: In function > 'int ReplicatedPG::recover_primary(int)' thread 7fc2f0f07700 time 2013-11-25 > 10:17:45.836790 > osd/ReplicatedPG.cc: 6643: FAILED assert(latest->is_update()) > > > The object (found clone without head) concern the rbd images below (which is > in use) : > > # rbd info datashare/share3 > rbd image 'share3': > size 1024 GB in 262144 objects > order 22 (4096 KB objects) > block_name_prefix: rb.0.32a6.238e1f29 > format: 1 > > > Directory contents : > In OSD.9 (Primary) : > /var/lib/ceph/osd/ceph-9/current/3.136_head/DIR_6/DIR_3/DIR_3/DIR_1# ls -l > rb.0.32a6.238e1f29.000000034d6a* > -rw-r--r-- 1 root root 4194304 nov. 6 02:25 > rb.0.32a6.238e1f29.000000034d6a__7ed_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 8 02:40 > rb.0.32a6.238e1f29.000000034d6a__7f5_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 9 02:44 > rb.0.32a6.238e1f29.000000034d6a__7fd_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 12 02:52 > rb.0.32a6.238e1f29.000000034d6a__815_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 14 02:39 > rb.0.32a6.238e1f29.000000034d6a__825_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 16 02:45 > rb.0.32a6.238e1f29.000000034d6a__835_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 19 01:59 > rb.0.32a6.238e1f29.000000034d6a__84d_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 20 02:25 > rb.0.32a6.238e1f29.000000034d6a__855_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 22 02:18 > rb.0.32a6.238e1f29.000000034d6a__865_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 23 02:24 > rb.0.32a6.238e1f29.000000034d6a__86d_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 23 02:24 > rb.0.32a6.238e1f29.000000034d6a__head_96AD1336__3 > > In OSD.1 (Replica) : > /var/lib/ceph/osd/ceph-1/current/3.136_head/DIR_6/DIR_3/DIR_3/DIR_1# ls -l > rb.0.32a6.238e1f29.000000034d6a* > -rw-r--r-- 1 root root 4194304 oct. 11 17:13 > rb.0.32a6.238e1f29.000000034d6a__5ab_96AD1336__3 <--- ???? > -rw-r--r-- 1 root root 4194304 nov. 6 02:25 > rb.0.32a6.238e1f29.000000034d6a__7ed_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 8 02:40 > rb.0.32a6.238e1f29.000000034d6a__7f5_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 9 02:44 > rb.0.32a6.238e1f29.000000034d6a__7fd_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 12 02:52 > rb.0.32a6.238e1f29.000000034d6a__815_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 14 02:39 > rb.0.32a6.238e1f29.000000034d6a__825_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 16 02:45 > rb.0.32a6.238e1f29.000000034d6a__835_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 19 01:59 > rb.0.32a6.238e1f29.000000034d6a__84d_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 20 02:25 > rb.0.32a6.238e1f29.000000034d6a__855_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 22 02:18 > rb.0.32a6.238e1f29.000000034d6a__865_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 23 02:24 > rb.0.32a6.238e1f29.000000034d6a__86d_96AD1336__3 > -rw-r--r-- 1 root root 4194304 nov. 23 02:24 > rb.0.32a6.238e1f29.000000034d6a__head_96AD1336__3 > > > The file rb.0.32a6.238e1f29.000000034d6a__5ab_96AD1336__3 is only present on > replica on osd.1. It seems that this snapshot (5ab) no longer exists. > > # ceph osd dump | grep snap > removed_snaps [1~c,e~23] > removed_snaps > [1~7,9~1,d~2,14~789,7a0~1,7a2~3,7a8~1,7aa~43,7f1~1,7f3~2,7f9~1,7fb~2,801~1,803~2,809~1,80b~2,811~1,813~2,819~1,81b~2,821~1,823~2,829~1,82b~2,831~1,833~2,839~1,83b~2,841~1,843~2,849~1,84b~2,851~1,853~2,859~1,85b~2,861~1,863~2,869~1,86b~2,871~1,873~2,879~1,87b~39,8ba~49] > > # for i in `rbd snap ls datashare/share3 | cut -f3 -d ' '`; do printf '%x, ' > $i; done > 7ed, 7f5, 7fd, 805, 80d, 815, 81d, 825, 82d, 835, 83d, 845, 84d, 855, 85d, > 865, 86d, 875, 8b4, 905 > > > How can I be sure that this file is no more useful? > > If these files are no longer used, do you think I can remove them manually > in osd.1 ? > > Like this ? : > > $ ceph osd set noout > $ service ceph stop osd.1 > > $ cd /var/lib/ceph/osd/ceph-1/current/3.136_head > $ mv > ./DIR_6/DIR_3/DIR_3/DIR_1/rb.0.32a6.238e1f29.000000034d6a__5ab_96AD1336__3 > ./DIR_6/DIR_3/DIR_3/DIR_2/rb.0.32a6.238e1f29.000000036552__5b3_E5822336__3 > ./DIR_6/DIR_3/DIR_3/DIR_2/rb.0.32a6.238e1f29.000000011379__5bb_38372336__3 > ./DIR_6/DIR_3/DIR_3/DIR_8/rb.0.32a6.238e1f29.00000003ad6b__5ab_109B8336__3 > /root/temp_obj_backup > > $ service ceph start osd.1 > $ ceph osd unset noout > > $ ceph pg repair 3.136
I ran this by Sam to get more ideas of the cause, and he suggests checking dmesg on the offending node (guessing that your local filesystem is having some issues). Manually removing the offending object from the replica and then running repair (to fix any lingering stat mismatches) should deal with it, yes. -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com