System won't boot multi user

[Geoff Wild]

I have a B2000 Workstation with HP-UX 11i - latest patches.

The other day, I made a kernel parameter change - MAX_DBC_PCT from 50 to 10.

The system hangs at:

/sbin/auto_parms: DHCP access is disabled (see /etc/auto_parms.log)

If I wait a while, then press the power button, it comes back - "command respawning to quickly"....samd

I tried booting the previous kernel - same thing - so it can't be the parameter change?

The only other change to my system - 2 weeks ago I installed BIND 9.2.0 for testing...

I can boot into single user mode...

What I have tried:

commenting start up of BIND in /etc/rc.config.d/namesrvs

commenting out samd in /etc/inittab

There's nothing strange in any log file....all file systems can be mounted....

Looking for suggestions....

Thanks...Geoff

[Pete Randall]

Geoff,

It sure sounds like we should be suspecting BIND! In the meantime, have you tried going to multi-user mode from single-user mode: "init 3"?


Pete

[Jeroen Peereboom]

Geoff,

As a workaround to get your system up and running I would add the hostname (short and FQDN) to /etc/hosts and make sure that files are used for DNS lookups first (in /etc/resolv.conf).

samd can respawn if network config is not correct.

JP.

[Geoff Wild]

Pete - yes I did init 3 - same thing.

Jeroen - Yes, short and FQDN in /etc/hosts

nsswitch.conf:

hosts: files [NOTFOUND=CONTINUE] dns

Rgds...Geoff

[Geoff Wild]

Playing around with the init levels - I can go from s to 1 to 2 with no issues....

I even put samd back - and init 2 works!

Now - just to figure out why init 3 doesn't....

Rgds...Geoff

[Sridhar Bhaskarla]

Hi Geoff,

As a temporary fix, try moving /sbin/auto_parms as /sbin/auto_parms.old and see if the system boots. Then you can troubleshoot from there.

-Sri

[Pete Randall]

Geoff,

Pursuing Jeroen's line of thought, do you have a legitimate /etc/hosts file and does it contain entries for the host in question? You might want to try taking the dns entry out of nsswitch.conf (or rename nsswitch.conf).


Pete

[Geoff Wild]

Here's what I can do:

ISL > hpux -is

# init 2

login as root

# init 3

Voila - Gnome login screen appears....

But trying a reboot, (default run-level is 3) - it hangs...

Stranger and stranger....

Rgds...Geoff

[Alex Glennie]

when you say hang ..... what exactly do you mean ? If gnome is running are you aware there was a problem with the gdm binary ...

http://ximian.com/hpux-updates.html & http://forums1.itrc.hp.com/service/forums/parseCurl.do?CURL=%2Fcm%2FQuestionAnswer%2F1%2C%2C0x58458b82cc91d711abdc0090277a778c%2C00.html&admit=716493758+1077551381640+28353475

[Geoff Wild]

Alex - yes - I know about the gdm issue with Gnome.

hang - the system just sits there - doing nothing - last thing on console is:

"/sbin/auto_parms: DHCP access is disabled (see /etc/auto_parms.log)"

It seems to hang at the inittab - IE it never gets to the rc start up scripts....

Rgds...Geoff

[Vasikaran Venkatesan]

try this!.
Press ^\ (control + backslash) (when it is staying at /sbin/auto_parms.....)
It may escape from where it is stuck. R u having any CA application loaded in the system ?.

[Volker Borowski]

Hi Geoff,

that sounds to be a timing problem between level 2 and 3. When you do it the login-way, enough time has passed (may be a network-timeout) to make things work.
If it is done too quickly, it gets locked up.

Check the links in /sbin/rc2.d and /sbin/rc3.d !

If you find nothing obvious, try to confirm the timing issue by creating a script
/sbin/rc3.d/S0000000_sleep
that just sleeps for a while.

If you succeed, you know you have to look for a startscript in rc2.d that is not waiting long enough.

Good luck
Volker

[Geoff Wild]

^\ (cntrl back slash) does the trick.....that is handy - but it would be nice if I didn't have to...

What is CA?

Rgds...Geoff

[Pete Randall]

Geoff,

In my world, CA is Computer Associates, though I'm not sure what their software might have to do with this problem!


Pete

[Geoff Wild]

Yeah - that's what I thought CA was - nope - nada....

Small issue with the ^\

The GUI logon does not run....

Back to inittab...

[Sridhar Bhaskarla]

Hi Geoff,

Did you try renaming auto_parms?. I got into trouble quite sometime back and I remember renaming this file (I don't use DHCP and ofcourse the system was already configured) that allowed me to login. You could always move it back after finding the problem by manually running it later.

-Sri

[RAC_1]

Hoe about renaming auto_parms and then botting and running it manuualy.

This may give some idea on what is going wrong.

[Geoff Wild]

I tried the auto_parms - boots - with a lot of complaining (like /usr not mounted....) so a lot of stuff does not run....

/usr mounts at the end of all the rc...

Running auto_parm manually - produces nothing....

The log says - auto_parm can only be run from the transition from run level S....

Rgds..Geoff

[Hrvoje Dubravica · Accepted solution]

Hi Geoff,

I had same problem, it is not conected with any kernel paramters or DHCP or anything from /etc/inittab.

Boot engine in single user mode, go in /etc/rc.config.d/ and find script(s) or executable(s) that someone put there and move it somewhere else. In this directory you can only have text files with variables!!!

After that shutdown -ry now.

Hope this help.
Kind regards,

Hrvoje

[Pete Randall]

Geoff,

Hrvoje's idea occurred to me yesterday but I never got a chance to mention it. When you're faced with unexplainable startup behaviour like this, it's alway a good idea to look at the contents of your /etc/rc.config.d and your /etc/rc?.d directories. Anything in there gets sourced during startup. Quite often you'll find that someone has backed up a file before changing it, giving you a filename.old or a filename.save or some such thing that gets sourced along with (and most likely afterward) the real one. This can lead to some very strange symptoms.


Pete

[Vasikaran Venkatesan]

Geoff - Good it worked. This is stuck in the place most likely because of some script not exiting properly and the boot up is waiting for it to complete.

If there is some application like from CA (like the CA unicenter scripts) - seems they are parsed before the actual runnning before executing them in rc2.d or rc3.d; IF parsing gives an error, it hangs there -

Solution: Remove the latest added startup script - and see if it boots up fine.

[Geoff Wild]

Hrvoje was correct.

We had one of our "external" consultants come in here working on an AssetCentre project - he used my machine to test.

He put a startup script in /sbin/rc3.d

Here's the script (I moved it to little "s" so it wouldn't get sourced):

s989AssetCentre.rc -> /opt/AssetCentre/gather_hp.sh

#!/bin/ksh -p

# 22 January 2004

unset LD_LIBRARY_PATH
unset LPATH
unset SHLIB_PATH
export PATH=/usr/bin

cd /var/opt/AssetCentre
/opt/cfg2html/cfg2html_hpux.sh -FLSacdeklx > /dev/null 2>&1
# -F disable: Fibre channel
# -L disable: Screen tips inline
# -S disable: Software
# -a disable: Applications
# -c disable: Cron
# -d disable: Diagnostics (cstm #)
# -e disable: Enhancements
# -k disable: Kernel
# -l disable: LVM
# -x don't create background images

/opt/AssetCentre/hp.sh

Of course, he didn't test it with a boot - only from the command line....

Thanks for all the answers...

I now will find out why he was running it at boot - when I run the cfg2html from cron once a month....and he setup the asset centre on to run from cron as well...

Rgds...Geoff