Following up on an old thread, for the record...
On Mon, Oct 16, 2017 at 03:30:52PM -0700, Gregory Farnum wrote:
On Mon, Oct 16, 2017 at 5:58 AM, 陶冬冬 <tdd21151186@xxxxxxxxx> wrote:
Dear Cephers,
ceph version: 10.2.5
log below here:
0> 2017-10-16 03:30:18.346892 7fa278797700 -1 os/filestore/LFNIndex.cc: In function 'int LFNIndex::list_objects(const std::vector<std::basic_string<char> >&, int, long int*, std::map<std::basic_string<char>, ghobject_t>*)' thread 7fa278797700 time 2017-10-16 03:30:18.342894
os/filestore/LFNIndex.cc: 443: FAILED assert(long_name == short_name)
...
In this case, the LFNIndex is trying to convert the naive Long File
Name of an object into names short enough to play nicely with local
filesystems. It shortens them via hashing and has a bunch of extension
tricks to deal with collisions. The particular assert is that, if the
name is short enough to not need hashing, then the long and short
names should be identical. Apparently they aren't? I'm surprised you
found an issue here, though — this is pretty old and stable code. Has
anything strange happened to your cluster? Or are you running some
custom code?
-Greg
See Also: https://tracker.ceph.com/issues/22531
I was also suffering this problem on a pg repair (ceph 10.2.10 in my case).
That was because I was following the procedure at:
https://ceph.com/geen-categorie/ceph-manually-repair-object/
...and at the point of "move the bad object to another location" I did a "mv
${obj} ${obj}.bad", thus creating in Josh's words in the tracker item above
"a file that does not obey the internal naming rules of filestore".
To resolve the issue I moved the bad object elsewhere entirely (to save it
for further examination) and restarted the osd.
Cheers,
Chris
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
