Re: "device busy" reading file from NFS

[Paul Goyette <paul%whooppee.com@localhost>]

On Tue, 28 Apr 2009, John D. Baker wrote:

> In other communication, this problem seems known as PR bin/41259.
>
> I've been conducting some tests and clients running -current with amd(8)
> automounted high-traffic filesystems DO NOT exhibit this problem.  There
> are the occasional "nfs server 'foo' not responding"/"nfs server 'foo'
> is alive again" messages, but never a "device busy" message.
>
> So, whatever broke in -5.0 seems fixed (or maybe masked by something
> else?) in -current.

Well, maybe I can add (or subtract) another datapoint...

I've had a couple of mysterious nfs "incidents" lately, running on 
-current from 5.99.8 thru 5.99.11 inclusive.  Symptoms appear more or 
less at random.  The only thing I've seen that's common is some weird 
.nfs* file showing up in a client's view of a directory.  Neither the 
client nor the server can umount the file system unless using -f.  And 
even if the client does umount -f, the server still cannot umount it, 
even if ALL clients hav umounted it.

This has shown up a couple times on the nfs volume I use as my "build" 
volume (multiple */{src,obj,dest,rel} plus an rsync'd cvsroot), and it 
has caused more than a few cvs updates to just hang.

Generally, rebooting the nfs server fixes things, but that's a rather 
"intrusive" fix.  Unfortunately I haven't found a smaller hammer.  :)


-------------------------------------------------------------------------
|   Paul Goyette   | PGP DSS Key fingerprint: |  E-mail addresses:      |
| Customer Service | FA29 0E3B 35AF E8AE 6651 |  paul at whooppee.com   |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer |                          | pgoyette at netbsd.org  |
-------------------------------------------------------------------------