Hi, After an OSS crashed I ran fsck and all but one OST returned quickly after fixing a few errors. It''s been duplicated multiply claimed blocks for a few days now. Seems it''s a very slow and CPU bound operation. Are there other ways to fix or replace this OST? I''m running on RHEL 4 w/ latest Lustre e2fsprogs. Thanks for any suggestions! Dan
On Wed, 2010-04-14 at 09:59 -0700, Dan wrote:> Hi,Hi,> It''s been duplicated multiply claimed blocks > for a few days now.Are you saying that an fsck has been running for a few days, complaining all that time about multiply claimed blocks? That seems a long time. How big is the OST? Do the block numbers seem to be changing or is there any chance it''s looping? b. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 197 bytes Desc: This is a digitally signed message part Url : http://lists.lustre.org/pipermail/lustre-discuss/attachments/20100414/5c87cb9d/attachment.bin
On 2010-04-14, at 09:59, Dan wrote:> After an OSS crashed I ran fsck and all but one OST returned quickly > after fixing a few errors. It''s been duplicated multiply claimed > blocks > for a few days now. Seems it''s a very slow and CPU bound operation. > Are there other ways to fix or replace this OST? > > I''m running on RHEL 4 w/ latest Lustre e2fsprogs. Thanks for any > suggestions!Do you mean RHEL5.4? The Lustre-patched e2fsck has a feature that can simply discard inodes that have shared blocks. This is a potential concern in secure environments when it is unclear which inode the original file data belongs to. A new e2fsck option: -E shared=preserve|lost+found|delete Select the disposition of files containing shared blocks. "preserve" is the old behavior which remains the default. "lost+found" causes files to be unlinked after cloning so they will be reconnected to /lost+found in pass 3. "delete" skips cloning entirely and simply deletes the files. You probably want to use the "-E shared=delete" option. Cheers, Andreas -- Andreas Dilger Principal Engineer, Lustre Group Oracle Corporation Canada Inc.