[Gluster-users] Directories got the same gfid after rename

[huangql]

Dear all,

      I got a problem that some directories got the same gfid after doing actions as follow:

      # cd /testfs/mydir

      # ls -l   
       total 36
-rw-r--r-- 1 root    root    0 Apr 20 00:09 bb
drwxrwxrwx 5 protein cms  4096 Apr 21 13:33 aa
     drwxr-xr-x 2 root    root 4096 May  9 22:36 cc
     drwxr-xr-x 3 root    root 4096 May 13 23:12 dd
drwxr-xr-x 4 protein cms  4096 May 10 15:33 test

      # mv test test_old

      # mkdir test

By doing those, we got the same gfid of "test" and "test_old" on server side. At this time, it's confused for client when accessing "test" and "test_old".

For this problem, I generated a unique gfid for "test" and reset it manually. It's neccessary to restart client and server to clean inode table to take effect, otherwise, gfid self-heal will always using gfid got from inode to heal it.

Does someone get this problem? Who knows to clean inode table online without restart all clients and servers to keep from affecting storage service?


Thank you in advance.


Cheers,
Qiulan  



2013-05-15 



huangql 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130515/d56ee07f/attachment.html>