Ceph » CephFS

On Sat, Oct 21, 2017 at 5:42 PM, Xiaoxi Chen <superdebuger@gmail.com> wrote:
> Hi , I hit this on one of my MDS rank, it looks to me like the
> export_dir message is way much too big , the hux dump take more than
> 20GB(stop by no more disk space...).
>
>
> The only crazy thing in the FS is we have 6 mdses and 5000+ export binding.f
>
>
> 2017-10-21 00:47:18.731219 7f27463e0700  0 --
> 10.148.245.141:6800/4152861050 >> 10.148.245.48:0/3250358670
> conn(0x558aeb4ac800 :6800 s=STATE_OPEN pgs=44 cs=1 l=0).fault server,
> going to standby
> 2017-10-21 00:47:44.500382 7f27463e0700 -1 failed to decode message of
> type 1110 v1: buffer::end_of_buffer
> 2017-10-21 00:47:44.503923 7f27463e0700 -1 *** Caught signal (Aborted) **
>  in thread 7f27463e0700 thread_name:msgr-worker-1
>
>  ceph version 12.2.0 (32ce2a3ae5239ee33d6150705cdb24d43bab910c) luminous (rc)
>  1: (()+0x59b344) [0x558ad5d94344]
>  2: (()+0x11390) [0x7f274910d390]
>  3: (gsignal()+0x38) [0x7f274809d428]
>  4: (abort()+0x16a) [0x7f274809f02a]
>  5: (__gnu_cxx::__verbose_terminate_handler()+0x16d) [0x7f27489e084d]
>  6: (()+0x8d6b6) [0x7f27489de6b6]
>  7: (()+0x8d701) [0x7f27489de701]
>  8: (()+0xb8d38) [0x7f2748a09d38]
>  9: (()+0x76ba) [0x7f27491036ba]
>  10: (clone()+0x6d) [0x7f274816f3dd]
>  NOTE: a copy of the executable, or `objdump -rdS <executable>` is
> needed to interpret this.