[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: NFS random "permission denied"

On Thu, 22 Mar 2007, Jim Popovitch wrote:

> On Thu, 2007-03-22 at 10:58 +0000, Dan Wade wrote:
> > Hi -
> >
> > Forty NFS client workstations running Debian etch and sarge, am-utils
> > automounter, a dozen active users.  Server is Debian sarge on Dell
> > PowerEdge 2850, PERC, Raid-5, stock kernel 2.6.8-2-686-smp.  Runs just
> > fine for about 3 months, then starts randomly saying "permission
> > denied".  nfs-kernel-server restart doesn't fix it.  Reboot has worked
> > on both of two occasions, but 3 months later it starts "permission
> > denied" again.  If there's an error in /etc/exports, I can't see it.
> >
> > Googling found a couple of similar reports, but no obvious course of
> > action other than checking /etc/exports.  If I've got an error there, I
> > can't see it.  Sorry, can't publish - security.  In any case the
> > symptoms don't mesh with a syntax error.  What should I do?
> Question:  Do the logs, or manual poking around, show any read or write
> problems with one or more of the exports?

No nasties in the logs.  I can create/read/delete files.  The exported
reiserfs partitions seem to be free of hardware errors and showed no
errors while I watched it reboot about three months ago.  All the blue
lights are on.  lost+found directories remain empty.  But takes a lot of
time to read 4 233G partitions, and it slows the server horribly.  I
don't know a good way to slowly scan the disk or slowly read all the
files (some quite large) looking for a bad bit, without hogging

> Any one making changes to /etc/hosts.(allow|deny)?

aide says no, maintenance script says no.

I should expand a bit, I get the impression that "permission denied"
relates to mount requests, not to specific files.  The snag is that it
rarely does it when I go looking for it, but I did get it to refuse me
"mount server:/export/partition" just once.  I get user reports like "no
home directory, logging in to /", and from simulationists whose
long-running programs have silently stopped updating their files (though
a biggish network fryup will do that also).  And there are samba users
who report "no network drive", then later say "it's come back".  To some
extent the automounter timeout is saving our bacon - the frustrated user
logs off for 5+ minutes, amd does a umount and when they try again, they
often get lucky.

> -Jim P.

I have the stock 2.4 kernel lined up (Michael Loftis' suggestion) for a
scheduled reboot early on Monday.  Hope it works on the rather new
hardware, "permission denied" is getting more frequent - and my normally
tolerant users are stocking-up with squashy tomatoes and eggs...

- Dan

Reply to: