Ceph

git: v0.48.2 3e02b2fad88c2a95d9c0c86878f10d1beb780bfe

In the process of scrubbing, the primary encounters an object that has clones but is missing a snapset. This causes an assert failure in src/osd/ReplicatedPG.cc _scrub.

These are the last lines of the log before the assert fail (truncated):

2012-10-24 08:20:07.314332 7f8f2350f700 20 osd.10 12873 pg[4.2( v 12611'10870388 (12597'10865322,12611'10870388] n=51440 ec=12 les/c 12869/12873 12860/12860/12841) [10,12,13] r=0 lpr=12860 lcod 0'0 mlcod 0'0 active+clean+scrubbing] scrub  268abeca/rb.0.c.000000004ea5/head//4 268abeca/rb.0.c.000000004ea5/head//4(3348'147708 client.6315.0:46760 wrlock_by=unknown.0.0:0)
2012-10-24 08:20:07.314340 7f8f2350f700 20 osd.10 12873 pg[4.2( v 12611'10870388 (12597'10865322,12611'10870388] n=51440 ec=12 les/c 12869/12873 12860/12860/12841) [10,12,13] r=0 lpr=12860 lcod 0'0 mlcod 0'0 active+clean+scrubbing] scrub  3983aeca/rb.0.9.000000000044/head//4 3983aeca/rb.0.9.000000000044/head//4(9455'10494008 client.8049.0:13400 wrlock_by=unknown.0.0:0)
2012-10-24 08:20:07.314352 7f8f2350f700  0 log [ERR] : scrub 4.2 3983aeca/rb.0.9.000000000044/d6//4 on disk size (0) does not match object info size (4194304)
2012-10-24 08:20:07.314354 7f8f2350f700 20 osd.10 12873 pg[4.2( v 12611'10870388 (12597'10865322,12611'10870388] n=51440 ec=12 les/c 12869/12873 12860/12860/12841) [10,12,13] r=0 lpr=12860 lcod 0'0 mlcod 0'0 active+clean+scrubbing] scrub  3983aeca/rb.0.9.000000000044/d6//4 3983aeca/rb.0.9.000000000044/d6//4(8231'10144560 client.7605.0:74 [d6,d1,cc,c7,c2,bd,b8,b3,ae])
2012-10-24 08:20:07.315112 7f8f2350f700 -1 osd/ReplicatedPG.cc: In function 'virtual int ReplicatedPG::_scrub(ScrubMap&, int&, int&)' thread 7f8f2350f700 time 2012-10-24 08:20:07.314362
osd/ReplicatedPG.cc: 6540: FAILED assert(head != hobject_t())

 ceph version 0.48.2argonaut (commit:3e02b2fad88c2a95d9c0c86878f10d1beb780bfe)
 1: (ReplicatedPG::_scrub(ScrubMap&, int&, int&)+0x1924) [0x53a6d4]
 2: (PG::scrub_finalize()+0x1241) [0x628851]
 3: (OSD::ScrubFinalizeWQ::_process(PG*)+0xc) [0x5d8e3c]
 4: (ThreadPool::worker()+0x4d5) [0x79f835]
 5: (ThreadPool::WorkThread::entry()+0xd) [0x5d87cd]
 6: (()+0x7e9a) [0x7f8f34a07e9a]
 7: (clone()+0x6d) [0x7f8f336a74bd]
 NOTE: a copy of the executable, or `objdump -rdS <executable>` is needed to interpret this.

The PG in question is pg[4.2(

The log can be found on cephdrop: rlr219-ceph-osd.10.log.1.gz

This bug affects a user who came to #ceph with a crashing OSD.