(This came up as in-reply-to to the previous "mds crashing" thread --
it's better to start threads with a fresh message)
On 28/05/2015 16:58, Peter Tiernan wrote:
Hi all,
I have been testing cephfs with erasure coded pool and cache tier. I
have 3 mds running on the same physical server as 3 mons. The cluster
is in ok state otherwise, rbd is working and all pg are active+clean.
Im running v 0.87.2 giant on all nodes and ubuntu 14.04.2 .
The cluster was working fine but when copying a large file on a client
to cephfs, it froze and now mdss keep crashing with:
0> 2015-05-28 16:50:58.267112 7f0282946700 -1 mds/MDCache.cc: In
function 'virtual void C_IO_MDC_TruncateFinish::finish(int)' thread
7f0282946700 time 2015-05-28 16:50:58.243904
mds/MDCache.cc: 5974: FAILED assert(r == 0 || r == -2)
any ideas?
You're getting some kind of IO error from RADOS, and the CephFS code
doesn't have clean handling for that in many cases, so it's asserting out.
Enable "debug objecter = 10" on the MDS to see what the operation is
that's failing, and please provide the whole section of the log leading
up to the crash rather than just the last line.
Cheers,
John
_______________________________________________
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
