[Linux-cluster] umount failed - device is busy

[Herta Van den Eynde]

Lon Hohberger wrote:
> On Mon, 2005-10-10 at 22:06 +0200, Herta Van den Eynde wrote:
> 
> 
>>>Did you try enabling force unmount in the device/file system
>>>configuration?
>>>
>>>-- Lon
>>
>>Thanks for the explanation, Lon.  Yes, the devices are configured for 
>>"Force Unmount".
>>With the device unmounted on all of the nfs clients I even tried to 
>>'umount -f' manually, but I got the same result.
> 
> 
> Odd.  Well, "umount -f" actually doesn't do what most people think it
> does.
> 
> The "force unmount" option looks for and kills any user-land process
> holding a reference on the file system using "kill -9".
> 
> So, if you're getting EBUSY on unmount even though force-unmount is
> working (confirmed by you looking at lsof/fuser), chances are good that
> there's a kernel reference on the file system.
> 
> It could be something NFS related - try "service nfs stop" and see if
> you can umount the file system.
> 
> -- Lon
> 
Unfortunately, this is a production cluster which serves well over 
100,000 users (e-learning environment for our university, a dozen 
associated colleges, and a few hundred K-12 institutions) and I only 
have 4 hour maintenance windows on the 7th of each month, so stopping 
all of nfs is not an option today.  :-(
One of the cluster services is used for admin purposes, and that's the 
only one I can currently use (within limits) to test suggestions.

FWIIW, I don't think the force unmount works.  True, lsof/fuser don't 
report processes against the filesystem, but "df" and "mount" show that 
  it's still there, and I can write to and read from it after I try a 
"umount -f".

Kind regards,

Herta


Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm