New automount (NFS) permission issues
I am running debian/testing across a number of machines, all mostly up
to date (usually any given machine is no more than a week behind).
Some time ago, maybe a couple of months, I started noticing some
problems with my automounted NFS mounts, and wondering if anyone else
has noticed something similar to the problems I'm about to describe.
It's possible that I have something misconfigured that has worked for
years, but now something is more strict. But I've been unsuccessful
in finding anything in searches. And I'm not sure if this might be an
autofs, kernel, or some other issue. Any suggestions on what kind of
debugging to turn on would be helpful as well.
The effect I'm seeing is: Immediately after boot, a normal user can
access any NFS directory causing it to automount. After some amount
of time though, after the mount is automatically unmounted, normal
user can no longer do that, and instead root has to be used to access
the directory (causing automount to remount).
My configuration has not changed for a couple of years and has worked
fine until recently.
Each machine has a number of directories in /export that are exported
via NFS. Most of these are configured for mounting into /share,
though /home is as well. A typical exportfs entry looks like this:
/export/images
192.168.1.0/255.255.255.0(rw,async,no_subtree_check,root_squash)
The automount configuration is served up via ldap, and my configs
there look like:
dn: ou=auto.share,ou=automount,dc=mrc,dc=home
ou: auto.share
objectClass: top
objectClass: automountMap
dn: cn=/share,ou=auto.master,ou=automount,dc=mrc,dc=home
objectClass: top
objectClass: automount
automountInformation: ldap:ou=auto.share,ou=automount,dc=mrc,dc=home
--timeout=600
cn: /share
dn: cn=images,ou=auto.share,ou=automount,dc=mrc,dc=home
objectClass: automount
automountInformation: thune:/export/images
cn: images
This seems to happen on both local (direct mounts) and remote (NFS mounts).
Neither [autofs reload] nor [autofs restart] seems to help the
situation once is gets into this state.
One thing I've been doing as a temporary work around has been opening
extra shells and dropping them into the directories I want to keep
mounted. Sort of defeats the whole purpose of automounting, but such
is the life of hacks. Also, I've recently been seeing problems with
util-linux's flock(1) in my home directory. I use it as a
serialization tool:
$ cat bin/ser
#!/bin/bash
flock $0 "$@"
And that has suddenly started failing with:
flock: /home/nexus/bin/ser: Input/output error
Which was quite scary at first (crap, /home is dying!) but is turning
out to actually just be a constant NFS glitch. Not sure if related to
the automount issue, but would not be surprising.
I've scoured all of the stuff in /var/log with no luck.
Does anyone see anything I have misconfigured above? Or extra configs
I need to share?
Any ``Yes, you need to read X as it explains a recent change?''
Any recommend flags to turn on for additional debugging?
Anyone else seeing similar issues?
Thanks,
mrc
Reply to: