[Lustre-discuss] Clearing NOT HEALTHY status from health_check

[Andreas Dilger]

On Nov 11, 2008  16:58 +0900, Alex Lee wrote:
> Is there anyway you can clear the NOT HEALTHY status from 
> /proc/fs/lustre/heath_check without rebooting or remounting the OSTs?
> 
> Right not its stuck in NOT HEALTHY status because of a LBUG trigger. I 
> found the bug and want to clear the status to see if it gets triggered 
> again.

If you hit an LBUG your system is in an indeterminate state and should
be rebooted at the earliest convenience.  In some lucky cases the LBUG
will not interferewith anything other than the one process that hit it,
but in most cases it will eventually cause other threads to be blocked
or hit the same LBUG.

There is intentionally no way to clear this flag without rebooting.


Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.