Ceph » CephFS

    -5> 2013-04-16 12:27:41.205931 7ffc064f9700  5 client.4112 put_cap_ref dropped last FILE_CACHE ref on 10000000005.head(ref=5 cap_refs={1024=0,2048=0,4096=0,8192=0} open={3=1} mode=100664 size=83
88608 mtime=2013-04-16 12:27:15.077370 caps=pAsxLsXsxFsxcrwb(0=pAsxLsXsxFsxcrwb) objectset[10000000005 ts 0/0 objects 1 dirty_or_tx 0] parents=0x20e1fd0 0x7ffbf4008390)
    -4> 2013-04-16 12:27:41.205949 7ffc064f9700 10 client.4112 put_inode on 10000000005.head(ref=5 cap_refs={1024=0,2048=0,4096=0,8192=0} open={3=1} mode=100664 size=8388608 mtime=2013-04-16 12:27:1
5.077370 caps=pAsxLsXsxFsxcrwb(0=pAsxLsXsxFsxcrwb) objectset[10000000005 ts 0/0 objects 1 dirty_or_tx 0] parents=0x20e1fd0 0x7ffbf4008390)
    -3> 2013-04-16 12:27:41.205968 7ffc064f9700 10 client.4112 _async_invalidate 0~0 done
    -2> 2013-04-16 12:27:41.205972 7ffc064f9700 10 client.4112 _async_invalidate 0~0
    -1> 2013-04-16 12:27:41.205985 7ffc064f9700 -1 put_cap_ref Fc went negative on 10000000005.head(ref=4 cap_refs={1024=0,2048=0,4096=0,8192=0} open={3=1} mode=100664 size=8388608 mtime=2013-04-16 
12:27:15.077370 caps=pAsxLsXsxFsxcrwb(0=pAsxLsXsxFsxcrwb) objectset[10000000005 ts 0/0 objects 1 dirty_or_tx 0] parents=0x20e1fd0 0x7ffbf4008390)
     0> 2013-04-16 12:27:41.206953 7ffc064f9700 -1 client/Inode.cc: In function 'bool Inode::put_cap_ref(int)' thread 7ffc064f9700 time 2013-04-16 12:27:41.206021
client/Inode.cc: 126: FAILED assert(cap_refs[c] > 0)

 ceph version 0.60-473-g4a84ddb (4a84ddbd30ef3cc105053f272819c2f1ea2be0d2)
 1: (Inode::put_cap_ref(int)+0x229) [0x4e6039]
 2: (Client::put_cap_ref(Inode*, int)+0x38) [0x495c88]
 3: (Client::_async_invalidate(Inode*, long, long, bool)+0x2d0) [0x4967c0]
 4: (Finisher::finisher_thread_entry()+0x1c0) [0x4f5d70]
 5: (()+0x7e9a) [0x7ffc111a9e9a]
 6: (clone()+0x6d) [0x7ffc0fbcccbd]
 NOTE: a copy of the executable, or `objdump -rdS <executable>` is needed to interpret this.

This was during a personal teuth run while trying to test and reproduce the MPI ior issues for #3637 so I've got proper archives (with core dump!), and I'm saving the log as well on mira042 in case of accidents.

I'm setting it to high right now, but I'm not sure how important it actually is yet.