[lustre-discuss] reasonable runtime for e2fsck

Dilger, Andreas andreas.dilger at intel.com
Sat Aug 13 01:02:42 PDT 2016


I wouldn't consider the lack of an lfsck run a reason to rush into an upgrade.  In most cases this will only result in some orphan OST objects consuming space for a while, or MDT files returning ENOENT that can be deleted with "unlink".

As for upgrading, there isn't any reason to stop at the different versions along the way.

Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Intel High Performance Data Division

On 2016/08/12, 02:32, "lustre-discuss on behalf of E.S. Rosenberg" <lustre-discuss-bounces at lists.lustre.org<mailto:lustre-discuss-bounces at lists.lustre.org> on behalf of esr+lustre at mail.hebrew.edu<mailto:esr+lustre at mail.hebrew.edu>> wrote:

Yeah I stopped this and am strongly considering upgrading the servers to CentOS 6.7 + lustre 2.8, I hope to first test it on a test environment....
Did anyone run this type of upgrade? Should it work? Or should I take it in steps (2.6, 2.7 and only then 2.8)?
Thanks,
Eli

On Fri, Aug 12, 2016 at 9:49 AM, Dilger, Andreas <andreas.dilger at intel.com<mailto:andreas.dilger at intel.com>> wrote:
Running "e2fsck --mdsdb" will take _much_ longer than a regular e2fsck, because the mdsdb database is sparsely written.  I would recommend not running the old e2fsprogs-based lfsck, since Lustre 2.5 has enough functionality to repair the Lustre-specific parts of the local filesystem (after a regular e2fsck), and Lustre 2.7 will also fix the MDT-OST consistency (orphan objects and dangling MDT layout references).

Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Intel High Performance Data Division

On 2016/08/11, 20:19, "lustre-discuss on behalf of E.S. Rosenberg" <lustre-discuss-bounces at lists.lustre.org<mailto:lustre-discuss-bounces at lists.lustre.org> on behalf of esr+lustre at mail.hebrew.edu<mailto:esr+lustre at mail.hebrew.edu>> wrote:

Sorry about spamming the list but I realize it may be better that subjects be split into threads....
I started e2fsck --mdsdb 6 hours ago on an MDT that is 1T in size, am I being unreasonable if I think it should have been done by now?
What type of runtimes have you seen?
I shudder to think how long this is going to take on the OSTs if this is normal runtime....

Thanks,
Eli

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20160813/367b218e/attachment.htm>


More information about the lustre-discuss mailing list