[Lustre-discuss] Recovering from failed OST

Alex Lee alee at datadirectnet.com
Fri Sep 12 00:59:46 PDT 2008


I have a system with a failed OST. Resolved any hardware issues and 
e2fsck the OST before mounting. e2fsck seem to have quite of bit of 
repair done.
I remounted the OST and everything looked fine. I did not run lfsck yet.

Would lfsck restore any files that were in the OST's /lost&found to the 
lustre's fs /lfs/lost&found?

Also after the system was restored, another OST went in to a read only 
mode because the fibre connection to the disk died. Currently the disk 
is down.

Users started reporting files that some of the files were 0byte size. I 
checked them out and while some were the downed disk, there were files 
NOT mapped to the failed OST.
Files that were 0byte size are now coming back with a "No such file or 
directory".

I'm really worried now that I might be losing data or the MDS is 
starting to disconnect the inodes.

Can anyone help? What can I do about this situation other then bring 
back the disconnected OST...

Thanks,
-Alex




More information about the lustre-discuss mailing list