Since then I deleted the pool.
But I now have another problem, in fact the "opposite" of the previous :
now I never deleted files in clients, data objects and metadata are
still in pools, but directory is empty for clients (it is another
directory, other pool, etc. from previous problem).
Here are logs from MDS when I restart it about one of the files :
2015-03-16 09:57:48.626254 7f4177694700 12 mds.0.cache.dir(10000a95e05)
link_primary_inode [dentry #1/staging/api/easyrsa/vars [2,head] auth
NULL (dversion lock) v=22 inode=0 | dirty=1 0x6ca5a20] [inode
10000a95e11 [2,head] #10000a95e11 auth v22 s=0 n(v0 1=1+0) (iversion
lock) cr={29050627=0-1966080@1} 0x53c32c8]
2015-03-16 09:57:48.626258 7f4177694700 10 mds.0.journal
EMetaBlob.replay added [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v22 s=0 n(v0 1=1+0) (iversion lock)
cr={29050627=0-1966080@1} 0x53c32c8]
2015-03-16 09:57:48.626260 7f4177694700 10 mds.0.cache.ino(10000a95e11)
mark_dirty_parent
2015-03-16 09:57:48.626261 7f4177694700 10 mds.0.journal
EMetaBlob.replay noting opened inode [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v22 dirtyparent s=0 n(v0 1=1+0) (iversion
lock) cr={29050627=0-1966080@1} | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:48.626264 7f4177694700 10 mds.0.journal
EMetaBlob.replay sessionmap v 21580500 -(1|2) == table 21580499 prealloc
[] used 10000a95e11
2015-03-16 09:57:48.626265 7f4177694700 20 mds.0.journal (session
prealloc [10000a95e11~3dd])
2015-03-16 09:57:48.626843 7f4177694700 10 mds.0.journal
EMetaBlob.replay for [2,head] had [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v42 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:48.629319 7f4177694700 10 mds.0.journal
EMetaBlob.replay for [2,head] had [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v99 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:48.629357 7f4177694700 10 mds.0.journal
EMetaBlob.replay for [2,head] had [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v101 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:48.636559 7f4177694700 10 mds.0.journal
EMetaBlob.replay for [2,head] had [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v164 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:48.636597 7f4177694700 10 mds.0.journal
EMetaBlob.replay for [2,head] had [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v166 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:48.644280 7f4177694700 10 mds.0.journal
EMetaBlob.replay for [2,head] had [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v227 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:48.644318 7f4177694700 10 mds.0.journal
EMetaBlob.replay for [2,head] had [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v229 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:51.911267 7f417c9a1700 15 mds.0.cache chose lock
states on [inode 10000a95e11 [2,head] /staging/api/easyrsa/vars auth
v229 dirtyparent s=8089 n(v0 b8089 1=1+0) (iversion lock) |
dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:51.916816 7f417c9a1700 20 mds.0.locker
check_inode_max_size no-op on [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v229 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:51.958925 7f417c9a1700 7 mds.0.cache inode [inode
10000a95e11 [2,head] /staging/api/easyrsa/vars auth v229 dirtyparent
s=8089 n(v0 b8089 1=1+0) (iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
2015-03-16 09:57:56.561404 7f417c9a1700 10 mds.0.cache unlisting
unwanted/capless inode [inode 10000a95e11 [2,head]
/staging/api/easyrsa/vars auth v229 dirtyparent s=8089 n(v0 b8089 1=1+0)
(iversion lock) | dirtyparent=1 dirty=1 0x53c32c8]
What is going on ?
On 03/16/2015 02:18 AM, Yan, Zheng wrote:
> I don't know what was wrong. could you use "rados -p data ls" to check
> which objects still exist. Then restart the mds MDS with debug_mds=20
> and search the log for name of the remaining objects.
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
