After some more searching, I've found that the source of the problem
is with the mds and not the mon.. The mds crashes, generates a core
dump that eats the local space, and in turn the monitor (because of
leveldb) crashes.
The error in the mds log of one host:
2014-04-01 15:46:34.414615 7f870e319700 0 -- 10.141.8.180:6836/13152
>> 10.141.8.180:6789/0 pipe(0x517371180 sd=54 :42439 s=4 pgs=0 cs=0
l=1 c=0x147ac780).connect got RESETSESSION but no longer connecting
2014-04-01 15:46:34.438792 7f871194f700 0 -- 10.141.8.180:6836/13152
>> 10.141.8.180:6789/0 pipe(0x1b099f580 sd=8 :43150 s=4 pgs=0 cs=0
l=1 c=0x1fd44360).connect got RESETSESSION but no longer connecting
2014-04-01 15:46:34.439028 7f870e319700 0 -- 10.141.8.180:6836/13152
>> 10.141.8.182:6789/0 pipe(0x13aa64880 sd=54 :37085 s=4 pgs=0 cs=0
l=1 c=0x1fd43de0).connect got RESETSESSION but no longer connecting
2014-04-01 15:46:34.468257 7f871b7ae700 -1 mds/CDir.cc: In function
'void CDir::_omap_fetched(ceph::bufferlist&,
std::map<std::basic_string<char, std::char_traits<char>,
std::allocator<char> >, ceph::buffer::list,
std::less<std::basic_string<char, std::char_traits<char>,
std::allocator<char> > >, std::allocator<std::pair<const
std::basic_string<char, std::char_traits<char>, std::allocator<char>
>, ceph::buffer::list> > >&, const std::string&, int)' thread
7f871b7ae700 time 2014-04-01 15:46:34.448320
mds/CDir.cc: 1474: FAILED assert(r == 0 || r == -2 || r == -61)
ceph version 0.78 (f6c746c314d7b87b8419b6e584c94bfe4511dbd4)
1: (CDir::_omap_fetched(ceph::buffer::list&, std::map<std::string,
ceph::buffer::list, std::less<std::string>,
std::allocator<std::pair<std::string const, ceph::buffer::list> > >&,
std::string const&, int)+0x4d71) [0x77c3c1]
2: (Context::complete(int)+0x9) [0x56bb79]
3: (Objecter::handle_osd_op_reply(MOSDOpReply*)+0x11a6) [0x806dd6]
4: (MDS::handle_core_message(Message*)+0x9c7) [0x5901d7]
5: (MDS::_dispatch(Message*)+0x2f) [0x59028f]
6: (MDS::ms_dispatch(Message*)+0x1ab) [0x591d4b]
7: (DispatchQueue::entry()+0x582) [0x902072]
8: (DispatchQueue::DispatchThread::entry()+0xd) [0x85ef4d]
9: /lib64/libpthread.so.0() [0x34c36079d1]
10: (clone()+0x6d) [0x34c32e8b6d]
NOTE: a copy of the executable, or `objdump -rdS <executable>` is
needed to interpret this.
--- begin dump of recent events ---
-10000> 2014-04-01 15:46:34.313794 7f871b7ae700 1 --
10.141.8.180:6836/13152 --> 10.141.8.182:6789/0 -- auth(proto 0 42
bytes epoch 1) v1 -- ?+0 0x5b301c800 con 0x377e5c0
-9999> 2014-04-01 15:46:34.313811 7f871b7ae700 10
monclient(hunting): renew_subs
-9998> 2014-04-01 15:46:34.313829 7f871b7ae700 1 --
10.141.8.180:6836/13152 <== osd.3 10.141.8.180:6827/4366 33052 ====
osd_op_reply(4176829 10000c23f15.00000000 [getxattr] v0'0 uv0 ack =
-108 (Cannot send after transport endpoint shutdown)) v6 ==== 187+0+0
(3499028751 0 0) 0x51a1fb200 con 0x37ea3c0
-9997> 2014-04-01 15:46:34.313858 7f871b7ae700 5 mds.0.16 is_laggy
600.641332 > 15 since last acked beacon
-9996> 2014-04-01 15:46:34.313865 7f871b7ae700 5 mds.0.16
initiating monitor reconnect; maybe we're not the slow one
-9995> 2014-04-01 15:46:34.313867 7f871b7ae700 10
monclient(hunting): _reopen_session rank -1 name
-9994> 2014-04-01 15:46:34.313872 7f871b7ae700 1 --
10.141.8.180:6836/13152 mark_down 0x377e5c0 -- 0xae9cb480
-9993> 2014-04-01 15:46:34.313888 7f870e319700 2 --
10.141.8.180:6836/13152 >> 10.141.8.182:6789/0 pipe(0xae9cb480 sd=54
:0 s=4 pgs=0 cs=0 l=1 c=0x377e5c0).connect couldn't read banner, Success
-9992> 2014-04-01 15:46:34.313902 7f870e319700 3 --
10.141.8.180:6836/13152 >> 10.141.8.182:6789/0 pipe(0xae9cb480 sd=54
:0 s=4 pgs=0 cs=0 l=1 c=0x377e5c0).connect fault, but state = closed
!= connecting, stopping
-9991> 2014-04-01 15:46:34.313912 7f871b7ae700 10
monclient(hunting): picked mon.ceph001 con 0x377aec0 addr
10.141.8.180:6789/0
-9990> 2014-04-01 15:46:34.313923 7f871b7ae700 10
monclient(hunting): _send_mon_message to mon.ceph001 at
10.141.8.180:6789/0
-9989> 2014-04-01 15:46:34.313927 7f871b7ae700 1 --
10.141.8.180:6836/13152 --> 10.141.8.180:6789/0 -- auth(proto 0 42
bytes epoch 1) v1 -- ?+0 0x5b301f500 con 0x377aec0
-9988> 2014-04-01 15:46:34.313935 7f871b7ae700 10
monclient(hunting): renew_subs
-9987> 2014-04-01 15:46:34.313944 7f871b7ae700 1 --
10.141.8.180:6836/13152 <== osd.2 10.141.8.180:6800/2233 33052 ====
osd_op_reply(2116129 100003f7763.00000000 [getxattr] v0'0 uv0 ack =
-108 (Cannot send after transport endpoint shutdown)) v6 ==== 187+0+0
(567486810 0 0) 0x4ea1b9e00 con 0x37ebb20
.....
.....
-19> 2014-04-01 15:46:34.440672 7f871b7ae700 1 --
10.141.8.180:6836/13152 mark_down 0x4eff853e0 -- 0xa8693980
-18> 2014-04-01 15:46:34.440684 7f871194f700 2 --
10.141.8.180:6836/13152 >> 10.141.8.182:6789/0 pipe(0xa8693980 sd=54
:0 s=4 pgs=0 cs=0 l=1 c=0x4eff853e0).connect couldn't read banner,
Success
-17> 2014-04-01 15:46:34.440697 7f871b7ae700 10
monclient(hunting): picked mon.ceph001 con 0x4eff82260 addr
10.141.8.180:6789/0
-16> 2014-04-01 15:46:34.440703 7f871b7ae700 10
monclient(hunting): _send_mon_message to mon.ceph001 at
10.141.8.180:6789/0
-15> 2014-04-01 15:46:34.440701 7f871194f700 3 --
10.141.8.180:6836/13152 >> 10.141.8.182:6789/0 pipe(0xa8693980 sd=54
:0 s=4 pgs=0 cs=0 l=1 c=0x4eff853e0).connect fault, but state = closed
!= connecting, stopping
-14> 2014-04-01 15:46:34.440708 7f871b7ae700 1 --
10.141.8.180:6836/13152 --> 10.141.8.180:6789/0 -- auth(proto 0 42
bytes epoch 1) v1 -- ?+0 0x5e77ec140 con 0x4eff82260
-13> 2014-04-01 15:46:34.440716 7f871b7ae700 10
monclient(hunting): renew_subs
-12> 2014-04-01 15:46:34.440746 7f871b7ae700 1 --
10.141.8.180:6836/13152 <== osd.10 10.141.8.180:6806/2663 33077 ====
osd_op_reply(1698429 100003917b8.00000000 [getxattr] v0'0 uv0 ack =
-108 (Cannot send after transport endpoint shutd
own)) v6 ==== 187+0+0 (1851692513 0 0) 0x369dc7080 con 0x37ea680
-11> 2014-04-01 15:46:34.440770 7f871b7ae700 5 mds.0.16 is_laggy
600.768243 > 15 since last acked beacon
-10> 2014-04-01 15:46:34.440777 7f871b7ae700 5 mds.0.16
initiating monitor reconnect; maybe we're not the slow one
-9> 2014-04-01 15:46:34.440779 7f871b7ae700 10
monclient(hunting): _reopen_session rank -1 name
-8> 2014-04-01 15:46:34.440781 7f871b7ae700 1 --
10.141.8.180:6836/13152 mark_down 0x4eff82260 -- 0xc3287800
-7> 2014-04-01 15:46:34.440796 7f871063c700 2 --
10.141.8.180:6836/13152 >> 10.141.8.180:6789/0 pipe(0xc3287800 sd=8
:43167 s=4 pgs=0 cs=0 l=1 c=0x4eff82260).connect couldn't write my
addr, Interrupted system call
-6> 2014-04-01 15:46:34.440810 7f871063c700 3 --
10.141.8.180:6836/13152 >> 10.141.8.180:6789/0 pipe(0xc3287800 sd=8
:43167 s=4 pgs=0 cs=0 l=1 c=0x4eff82260).connect fault, but state =
closed != connecting, stopping
-5> 2014-04-01 15:46:34.440822 7f871b7ae700 10
monclient(hunting): picked mon.ceph002 con 0x4eff865c0 addr
10.141.8.181:6789/0
-4> 2014-04-01 15:46:34.440832 7f871b7ae700 10
monclient(hunting): _send_mon_message to mon.ceph002 at
10.141.8.181:6789/0
-3> 2014-04-01 15:46:34.440837 7f871b7ae700 1 --
10.141.8.180:6836/13152 --> 10.141.8.181:6789/0 -- auth(proto 0 42
bytes epoch 1) v1 -- ?+0 0x5e77e8900 con 0x4eff865c0
-2> 2014-04-01 15:46:34.440848 7f871b7ae700 10
monclient(hunting): renew_subs
-1> 2014-04-01 15:46:34.440860 7f871b7ae700 1 --
10.141.8.180:6836/13152 <== osd.3 10.141.8.180:6827/4366 33077 ====
osd_op_reply(4179177 100001f2ef1.00000000 [omap-get-header
0~0,omap-get-vals 0~16] v0'0 uv0 ack = -108 (Cannot send a
fter transport endpoint shutdown)) v6 ==== 229+0+0 (958358678 0 0)
0x2cff7aa80 con 0x37ea3c0
0> 2014-04-01 15:46:34.468257 7f871b7ae700 -1 mds/CDir.cc: In
function 'void CDir::_omap_fetched(ceph::bufferlist&,
std::map<std::basic_string<char, std::char_traits<char>,
std::allocator<char> >, ceph::buffer::list, std::less<std::ba
sic_string<char, std::char_traits<char>, std::allocator<char> > >,
std::allocator<std::pair<const std::basic_string<char,
std::char_traits<char>, std::allocator<char> >, ceph::buffer::list> >
>&, const std::string&, int)' thread 7f871b7ae7
00 time 2014-04-01 15:46:34.448320
mds/CDir.cc: 1474: FAILED assert(r == 0 || r == -2 || r == -61)
.....
2014-04-01 15:46:34.557614 7f871b7ae700 -1 *** Caught signal (Aborted) **
in thread 7f871b7ae700
ceph version 0.78 (f6c746c314d7b87b8419b6e584c94bfe4511dbd4)
1: /usr/bin/ceph-mds() [0x827909]
2: /lib64/libpthread.so.0() [0x34c360f710]
3: (gsignal()+0x35) [0x34c3232925]
4: (abort()+0x175) [0x34c3234105]
5: (__gnu_cxx::__verbose_terminate_handler()+0x12d) [0x34c3abea5d]
NOTE: a copy of the executable, or `objdump -rdS <executable>` is
needed to interpret this.
Thanks again!
----- Message from Kenneth Waegeman <Kenneth.Waegeman at UGent.be> ---------
Date: Tue, 01 Apr 2014 14:36:20 +0200
From: Kenneth Waegeman <Kenneth.Waegeman at UGent.be>
Subject: Re: ceph 0.78 mon and mds crashing (bus error)
To: Joao Eduardo Luis <joao.luis at inktank.com>
Cc: ceph-users <ceph-users at lists.ceph.com>
> ----- Message from Joao Eduardo Luis <joao.luis at inktank.com> ---------
> Date: Tue, 01 Apr 2014 13:33:05 +0100
> From: Joao Eduardo Luis <joao.luis at inktank.com>
> Subject: Re: ceph 0.78 mon and mds crashing (bus error)
> To: Kenneth Waegeman <Kenneth.Waegeman at UGent.be>, ceph-users
> <ceph-users at lists.ceph.com>
>
>
>> On 04/01/2014 01:17 PM, Kenneth Waegeman wrote:
>>> Hi all,
>>>
>>> We have installed Ceph 0.78 on 3 hosts running SL6, each serving OSD and
>>> MON daemons, 2 of them running MDS in active/backup. Since today the MON
>>> and MDS daemons are crashing every time (after a short while). Rebooting
>>> the nodes didn't help. The cluster was running for 1-2 weeks without
>>> this problem. This is what's in the log files:
>>>
>>
>> Something is going wrong on leveldb. What leveldb version are you using?
>
> This is leveldb-1.7.0-2.el6.x86_64
>>
>> -Joao
>>
>>>
>>>
>>> 2014-04-01 12:48:09.442819 7fa39cb1d700 0 log [INF] : mdsmap e67: 1/1/1
>>> up {0=ceph001.cubone.os=up:active(laggy or crashed)}
>>> 2014-04-01 12:48:09.442897 7fa39cb1d700 1
>>> mon.ceph001 at 0(leader).paxos(paxos active c 504009..504758) is_readable
>>> now=2014-04-01 12:48:09.442898 lease_expire=2014-04-01 12:48:14.442854
>>> has v0 lc 504758
>>> 2014-04-01 12:48:15.340419 7fa39d51e700 1
>>> mon.ceph001 at 0(leader).paxos(paxos updating c 504009..504758) is_readable
>>> now=2014-04-01 12:48:15.340420 lease_expire=2014-04-01 12:48:14.442854
>>> has v0 lc 504758
>>> 2014-04-01 12:48:15.340438 7fa39d51e700 1
>>> mon.ceph001 at 0(leader).paxos(paxos updating c 504009..504758) is_readable
>>> now=2014-04-01 12:48:15.340439 lease_expire=2014-04-01 12:48:14.442854
>>> has v0 lc 504758
>>> 2014-04-01 12:48:15.340443 7fa39d51e700 1
>>> mon.ceph001 at 0(leader).paxos(paxos updating c 504009..504758) is_readable
>>> now=2014-04-01 12:48:15.340444 lease_expire=2014-04-01 12:48:14.442854
>>> has v0 lc 504758
>>> 2014-04-01 12:48:22.260674 7fa39d51e700 1
>>> mon.ceph001 at 0(leader).paxos(paxos active c 504009..504759) is_readable
>>> now=2014-04-01 12:48:22.260676 lease_expire=2014-04-01 12:48:27.260625
>>> has v0 lc 504759
>>> 2014-04-01 12:48:22.260688 7fa39d51e700 1
>>> mon.ceph001 at 0(leader).paxos(paxos active c 504009..504759) is_readable
>>> now=2014-04-01 12:48:22.260689 lease_expire=2014-04-01 12:48:27.260625
>>> has v0 lc 504759
>>> 2014-04-01 12:48:22.260703 7fa39d51e700 1
>>> mon.ceph001 at 0(leader).paxos(paxos active c 504009..504759) is_readable
>>> now=2014-04-01 12:48:22.260704 lease_expire=2014-04-01 12:48:27.260625
>>> has v0 lc 504759
>>> 2014-04-01 12:48:22.292363 7fa39cb1d700 0 log [INF] : mon.ceph001
>>> calling new monitor election
>>> 2014-04-01 12:48:22.292428 7fa39cb1d700 1
>>> mon.ceph001 at 0(electing).elector(43) init, la2014-04-01 12:48:27.450606
>>> 7fa39d51e700 -1 *** Caught signal (Bus error) **
>>> in thread 7fa39d51e700
>>>
>>> ceph version 0.78 (f6c746c314d7b87b8419b6e584c94bfe4511dbd4)
>>> 1: /usr/bin/ceph-mon() [0x85e619]
>>> 2: /lib64/libpthread.so.0() [0x34c360f710]
>>> 3: (memcpy()+0x15b) [0x34c3289aab]
>>> 4: (()+0x3d4b2) [0x7fa3a0ae54b2]
>>> 5: (leveldb::log::Writer::EmitPhysicalRecord(leveldb::log::RecordType,
>>> char const*, unsigned long)+0x125) [0x7fa3a0acbe85]
>>> 6: (leveldb::log::Writer::AddRecord(leveldb::Slice const&)+0xe2)
>>> [0x7fa3a0acc092]
>>> 7: (leveldb::DBImpl::Write(leveldb::WriteOptions const&,
>>> leveldb::WriteBatch*)+0x38e) [0x7fa3a0ac0bde]
>>> 8:
>>> (LevelDBStore::submit_transaction_sync(std::tr1::shared_ptr<KeyValueDB::TransactionImpl>)+0x2b)
>>> [0x7c63eb]
>>> 9: (MonitorDBStore::apply_transaction(MonitorDBStore::Transaction
>>> const&)+0x183) [0x53c6b3]
>>> 10: (Paxos::begin(ceph::buffer::list&)+0x604) [0x5a2ca4]
>>> 11: (Paxos::propose_queued()+0x273) [0x5a3773]
>>> 12: (Paxos::propose_new_value(ceph::buffer::list&, Context*)+0x160)
>>> [0x5a3960]
>>> 13: (PaxosService::propose_pending()+0x386) [0x5ad946]
>>> 14: (Context::complete(int)+0x9) [0x580609]
>>> 15: (SafeTimer::timer_thread()+0x453) [0x768a33]
>>> 16: (SafeTimerThread::entry()+0xd) [0x76abed]
>>> 17: /lib64/libpthread.so.0() [0x34c36079d1]
>>> 18: (clone()+0x6d) [0x34c32e8b6d]
>>> NOTE: a copy of the executable, or `objdump -rdS <executable>` is
>>> needed to interpret this.
>>>
>>> --- begin dump of recent events ---
>>> -10000> 2014-04-01 12:37:41.891458 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 --> mon.2 10.141.8.182:6789/0 -- paxos(lease lc
>>> 504309 fc 503758 pn 0 opn 0) v3 -- ?+0 0x6246e00
>>> -9999> 2014-04-01 12:37:41.891476 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 <== mon.0 10.141.8.180:6789/0 0 ==== log(1 entries)
>>> v1 ==== 0+0+0 (0 0 0) 0x29f4380 con 0x2050840
>>> -9998> 2014-04-01 12:37:41.891487 7fa39cb1d700 1
>>> mon.ceph001 at 0(leader).paxos(paxos active c 503758..504309) is_readable
>>> now=2014-04-01 12:37:41.891488 lease_expire=2014-04-01 12:37:46.891457
>>> has v0 lc 504309
>>> -9997> 2014-04-01 12:37:41.893979 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 <== mon.2 10.141.8.182:6789/0 672234191 ====
>>> paxos(lease_ack lc 504309 fc 503758 pn 0 opn 0) v3 ==== 80+0+0
>>> (2650854818 0 0) 0x6246e00 con 0x2051760
>>> -9996> 2014-04-01 12:37:42.087358 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 <== mon.2 10.141.8.182:6789/0 672234192 ====
>>> forward(pg_stats(1 pgs tid 13980 v 0) v1 caps allow profile osd tid
>>> 57315 con_features 4398046511103) to leader v2 ==== 916+0+0 (3952263404
>>> 0 0) 0x6247580 con 0x2051760
>>> -9995> 2014-04-01 12:37:42.087395 7fa39cb1d700 1
>>> mon.ceph001 at 0(leader).paxos(paxos active c 503758..504309) is_readable
>>> now=2014-04-01 12:37:42.087396 lease_expire=2014-04-01 12:37:46.891457
>>> has v0 lc 504309
>>> -9994> 2014-04-01 12:37:42.890767 7fa39d51e700 5
>>> mon.ceph001 at 0(leader).paxos(paxos active c 503758..504309)
>>> queue_proposal bl 16145 bytes; ctx = 0x1fa6e90
>>> -9993> 2014-04-01 12:37:42.891438 7fa39d51e700 1 --
>>> 10.141.8.180:6789/0 --> mon.2 10.141.8.182:6789/0 -- paxos(begin lc
>>> 504309 fc 0 pn 1800 opn 0) v3 -- ?+0 0x369d000
>>> -9992> 2014-04-01 12:37:42.891520 7fa39d51e700 5
>>> mon.ceph001 at 0(leader).paxos(paxos updating c 503758..504309)
>>> queue_proposal bl 1025 bytes; ctx = 0x1fa0af0
>>> -9991> 2014-04-01 12:37:42.891527 7fa39d51e700 5
>>> mon.ceph001 at 0(leader).paxos(paxos updating c 503758..504309)
>>> propose_new_value not active; proposal queued
>>> -9990> 2014-04-01 12:37:42.892656 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 <== mon.2 10.141.8.182:6789/0 672234193 ====
>>> paxos(accept lc 504309 fc 0 pn 1800 opn 0) v3 ==== 80+0+0 (2064056753 0
>>> 0) 0x369d000 con 0x2051760
>>> -9989> 2014-04-01 12:37:42.893660 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 --> mon.2 10.141.8.182:6789/0 -- paxos(commit lc
>>> 504310 fc 0 pn 1800 opn 0) v3 -- ?+0 0x6247580
>>> -9988> 2014-04-01 12:37:42.894019 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 --> 10.141.8.180:6789/0 -- log(last 1348) v1 -- ?+0
>>> 0x41ddcc0 con 0x2050840
>>> -9987> 2014-04-01 12:37:42.894038 7fa39cb1d700 1 --
>>> 10.141.8.180:6789/0 --> mon.2 10.141.8.182:6789/0 -- paxos(lease lc
>>> 504310 fc 503758 pn 0 opn 0) v3 -- ?+0 0x6246e00
>>>
>>> 2 of the 3 MONS are down, the MDSs are down, but the OSDs are still
>>> running..
>>> Does someone knows what is happening here?
>>>
>>>
>>> Thanks!
>>>
>>> Kind regards,
>>> Kenneth
>>>
>>>
>>> _______________________________________________
>>> ceph-users mailing list
>>> ceph-users at lists.ceph.com
>>> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
>>
>>
>> --
>> Joao Eduardo Luis
>> Software Engineer | http://inktank.com | http://ceph.com
>
>
> ----- End message from Joao Eduardo Luis <joao.luis at inktank.com> -----
>
> --
>
> Met vriendelijke groeten,
> Kenneth Waegeman
>
> _______________________________________________
> ceph-users mailing list
> ceph-users at lists.ceph.com
> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
----- End message from Kenneth Waegeman <Kenneth.Waegeman at UGent.be> -----
--
Met vriendelijke groeten,
Kenneth Waegeman
