Bug #16923
Updated by Jason Dillaman over 7 years ago
ceph version 10.2.0 (3a9fba20ec743699b69bd0181dd6c54dc01c64b9) 1: (()+0x24c4bb) [0x7fdebcbde4bb] 2: (()+0x7cf54) [0x7fdebca0ef54] 3: (()+0x7dadb) [0x7fdebca0fadb] 4: (()+0x7e827) [0x7fdebca10827] 5: (()+0x65d46) [0x7fdebc9f7d46] 6: (()+0x66218) [0x7fdebc9f8218] 7: (rbd_aio_write()+0x29) [0x7fdebc9ec629] 8: (()+0xde816) [0x7fded0108816] 9: (()+0xde910) [0x7fded0108910] 10: (()+0xb69b0) [0x7fded00e09b0] 11: (()+0xbbc44) [0x7fded00e5c44] 12: (()+0xbbc44) [0x7fded00e5c44] 13: (()+0xbc8fc) [0x7fded00e68fc] 14: (()+0xfc36a) [0x7fded012636a] 15: (()+0x49800) [0x7fdeb84ac800] The current assumption is that this was a race with a transient hiccup in the image header's watch registration. The aio_write resulted in an attempt to (re-)acquire the exclusive lock and the assertion test for ensuring a valid watch header failed (since the watch was most likely being re-established in the background).