I recently put 10.2 onto one of my desktop boxes, and now I have a
bizarre problem: on boot, when chown root:utmp /var/run/utmp is called,
there's a delay of about 1-2 minutes. I have no idea what the box is
doing, but eventually the chown terminates and rc.S goes on its merry
way. The other bizarre component is that, after it's booted, the same
chown works fine with no delays. Weird, eh?

This machine is set up with nss_ldap to do authentication against an
LDAP server, which I thought might be the problem, except that a) my
10.0 box configured similarly doesn't have the problem, b) using numeric
ids didn't solve the problem, and c) the network interfaces aren't up,
so it should simply return immediately rather than waiting for the LDAP
server.

I don't suspect a hardware problem on the drive, since a) it works fine
after boot is complete, b) none of the tools I've tried detect one, c)
the machine has never exhibited other symptoms of hardware failure, and
d) the problem is clearly repeatable.

Google searches didn't turn up much. What are some other possiblilites
for what this problem could be?

--keith


--
kkeller-usenet@wombat.san-francisco.ca.us
(try just my userid to email me)
AOLSFAQ=http://wombat.san-francisco.ca.us/cgi-bin/fom
see X- headers for PGP signature information

Keith Keller wrote:

> I recently put 10.2 onto one of my desktop boxes, and now I have a
> bizarre problem: on boot, when chown root:utmp /var/run/utmp is called,
> there's a delay of about 1-2 minutes. I have no idea what the box is
> doing, but eventually the chown terminates and rc.S goes on its merry
> way. The other bizarre component is that, after it's booted, the same
> chown works fine with no delays. Weird, eh?

compare nsswitch.conf between those two machines, see if it's the same.

On 2006-01-05, Dominik L. Borkowski <dom@vbi.vt.edu> wrote:
> Keith Keller wrote:
>
>> I recently put 10.2 onto one of my desktop boxes, and now I have a
>> bizarre problem: on boot, when chown root:utmp /var/run/utmp is called,
>> there's a delay of about 1-2 minutes. I have no idea what the box is
>> doing, but eventually the chown terminates and rc.S goes on its merry
>> way. The other bizarre component is that, after it's booted, the same
>> chown works fine with no delays. Weird, eh?
>
> compare nsswitch.conf between those two machines, see if it's the same.

Nice idea, but I copied the 10.2's nsswitch.conf from the 10.0 box.
It might just be possible that a difference between the nss_ldap libs
might be an issue, though: on the 10.0 box IIRC I've got nss_ldap-226,
but on the 10.2 box I have nss_ldap-244 (I think). I'll compile the
older version on the 10.2 box and see if that fixes anything. (There's
precedent for this: the 10.0 box has 226, which was not the latest
version at the time, because the latest version at the time didn't work
at all. So it's definitely possible that things broke between
versions.)

--keith


--
kkeller-usenet@wombat.san-francisco.ca.us
(try just my userid to email me)
AOLSFAQ=http://wombat.san-francisco.ca.us/cgi-bin/fom
see X- headers for PGP signature information

Keith Keller <kkeller-usenet@wombat.san-francisco.ca.us> wrote:
> on boot, when chown root:utmp /var/run/utmp is called, there's a delay
> of about 1-2 minutes.

Are you really sure that is causing your delay?

> I have no idea what the box is doing, but eventually the chown
> terminates and rc.S goes on its merry way. The other bizarre component
> is that, after it's booted, the same chown works fine with no delays.
> Weird, eh?

My guess is that rc.S gets delayed by something else after the chown. To
verify what is causing your delay you could add some trace echos to your
rc.S, something like:

-8<----------------------------------------------------
touch /var/run/utmp
echo before chown `date +%H:%M:%S`
chown root.utmp /var/run/utmp
echo after chown `date +%H:%M:%S`
chmod 664 /var/run/utmp
echo after chmod `date +%H:%M:%S`

if [ "$ROOTTYPE" = "umsdos" ]; then # we need to update any files added in DOS:
echo "Synchronizing UMSDOS directory structure:"
echo " /sbin/umssync -r99 -v- /"
/sbin/umssync -r99 -v- /
fi

# Setup the /etc/motd to reflect the current kernel level:
# THIS WIPES ANY CHANGES YOU MAKE TO /ETC/MOTD WITH EACH BOOT.
# COMMENT THIS OUT IF YOU WANT TO MAKE A CUSTOM VERSION.
echo "`/bin/uname -sr`." > /etc/motd

# Now, this looks suspicious!
echo before PNP `date +%H:%M:%S`
# Configure ISA Plug-and-Play devices:
if [ -r /etc/isapnp.conf ]; then
if [ -x /sbin/isapnp ]; then
/sbin/isapnp /etc/isapnp.conf
fi
fi
echo after PNP `date +%H:%M:%S`
-8<----------------------------------------------------

regards Henrik
--
The address in the header is only to prevent spam. My real address is:
hc7(at)uthyres.com Examples of addresses which go to spammers:
root@variousus.net root@localhost

On 2006-01-05, Keith Keller <kkeller-usenet@wombat.san-francisco.ca.us> wrote:
>
> Nice idea, but I copied the 10.2's nsswitch.conf from the 10.0 box.
> It might just be possible that a difference between the nss_ldap libs
> might be an issue, though: on the 10.0 box IIRC I've got nss_ldap-226,
> but on the 10.2 box I have nss_ldap-244 (I think). I'll compile the
> older version on the 10.2 box and see if that fixes anything.

This was indeed the issue! Reverting back to nss_ldap-226 eliminated
the delay. Thanks for the nudge, Dominik. (Now, if I could nudge
myself to file a bug report with PADL....)

--keith


--
kkeller-usenet@wombat.san-francisco.ca.us
(try just my userid to email me)
AOLSFAQ=http://wombat.san-francisco.ca.us/cgi-bin/fom
see X- headers for PGP signature information

I've already solved the problem, but I thought I'd address Henrik's
comments anyway.

On 2006-01-05, Henrik Carlqvist <Henrik.Carlqvist@deadspam.com> wrote:
> Keith Keller <kkeller-usenet@wombat.san-francisco.ca.us> wrote:
>> on boot, when chown root:utmp /var/run/utmp is called, there's a delay
>> of about 1-2 minutes.
>
> Are you really sure that is causing your delay?

Yes, I'm positive that was the issue, because I did some trace echos in
rc.S, similar to what you described (except without the datestamps,
which would have been smart). My first attempts didn't even look at
the lines concerning utmp, because I thought it was incredibly unlikely
that the problem would be there. It's just a simple chown, right?
Eventually I did narrow it down to the chown, but it was a slow process.
("It couldn't be utmp, right? Maybe it's this umsdos stuff? No? Ok,
maybe the sysV init? No? Golly, there's nothing left...is it really
utmp? Could the touch be the problem? No? chown?!? Wow, it's really
chown! WTF?!?" And, of course, a reboot each time, since that's the
only time the problem manifested itself.

The moral of the story? If you're sprinkling echos into your shell
scripts to debug them, be liberal with your sprinkling. It doesn't take
a lot of time, and might save you time down the road, especially when
dealing with rc.S or rc.M.

The other moral (for those who care) is that even a member of the pack
of wild dogs can forget to post relevant information. In this case, I
should have mentioned how I came to the conclusion that it was hanging
at the chown of utmp; because I didn't, I sowed doubt that that was
really the issue.

--keith

--
kkeller-usenet@wombat.san-francisco.ca.us
(try just my userid to email me)
AOLSFAQ=http://wombat.san-francisco.ca.us/cgi-bin/fom
see X- headers for PGP signature information

Keith Keller <kkeller-usenet@wombat.san-francisco.ca.us> writes:

> I recently put 10.2 onto one of my desktop boxes, and now I have a
> bizarre problem: on boot, when chown root:utmp /var/run/utmp is called,
> there's a delay of about 1-2 minutes. I have no idea what the box is
> doing, but eventually the chown terminates and rc.S goes on its merry
> way. The other bizarre component is that, after it's booted, the same
> chown works fine with no delays. Weird, eh?
>

Sorry Keith, I accidently sent this already to you by e-mail... Anyway,
here's one for the news as well.

I have seen the same and if memory serves the delay is caused by the
fact that when the chown is issued like "chown root.utmp ...", i.e.
using the BSDish dot convention, the system will still use ldap
regardless of the fact that both the user and group are already found in
the local passwd and group files. If you change the command to "chown
root:utmp ..." the delay will go away -- assuming your setup resembles
ours.

--
Sami

Sami Nuuttila <sami.nuuttila@utu.fi> trolled:
> Keith Keller <kkeller-usenet@wombat.san-francisco.ca.us> writes:

> > I recently put 10.2 onto one of my desktop boxes, and now I have a
> > bizarre problem: on boot, when chown root:utmp /var/run/utmp is called,
> > there's a delay of about 1-2 minutes. I have no idea what the box is
> > doing, but eventually the chown terminates and rc.S goes on its merry
> > way. The other bizarre component is that, after it's booted, the same
> > chown works fine with no delays. Weird, eh?
> >

> Sorry Keith, I accidently sent this already to you by e-mail... Anyway,
> here's one for the news as well.

There is no reason to ever apologize to Ms. Weller.

cordially, as always,

rm