One last netbooting issue

[Mark Street]

This problem seems to be related to runlevel 0 rc scripts.  The order with 
which your network, dhcpcd, nfs go down is probably what is hanging the system.

Is this a RedHat box?  In /etc/rc.d/rc0.d/ you will notice that all 
services are KILLED one by one is specific order according to their 
importance to system stability for an orderly shutdown.  You could tweak 
this order so your important services do not get shutdown.  The last run 
control script to run in run level 0 is the killall script which you are 
describing below.  Worst case if you turn it off you will probably have 
some stale lock files in /var/run/subsys which may have to be cleaned up on 
boot.

My RH 7.3 box /etc/rc.d/rc0.d directory listing, note the order in which 
services are terminated - K05-K96.  Note dhcpcd is killed early on, also 
note the S00killall script that runs just before the S01halt script.  You 
could probably do the same thing for run level 6.

  ls /etc/rc.d/rc0.d/
K05anacron     K25sshd        K50tux      K73ypbind   K91isdn
K05keytable    K30qmail       K50xinetd   K74apmd     K92ip6tables
K10xfs         K34yppasswdd   K60atd      K74nscd     K92ipchains
K12mysqld      K35dhcpd       K60crond    K74ntpd     K92iptables
K12postgresql  K35smb         K60lpd      K74ypserv   K95kudzu
K15gpm         K35vncserver   K61ldap     K74ypxfrd   K96irda
K15httpd       K40mars-nwe    K65identd   K75netfs    K96pcmcia
K20nfs         K44rawdevices  K65kadmin   K80random   S00killall
K20rstatd      K45arpwatch    K65kprop    K86nfslock  S01halt
K20rusersd     K45named       K65krb524   K87portmap
K20rwalld      K50snmpd       K65krb5kdc  K88syslog
K20rwhod       K50snmptrapd   K72autofs   K90network


S00killall
#!/bin/bash

# Bring down all unneeded services that are still running (there shouldn't
# be any, so this is just a sanity check)

for i in /var/lock/subsys/*; do
         # Check if the script is there.
         [ ! -f $i ] && continue

         # Get the subsystem name.
         subsys=${i#/var/lock/subsys/}

         # Bring the subsystem down.
         if [ -f /etc/init.d/$subsys.init ]; then
             /etc/init.d/$subsys.init stop
         elif [ -f /etc/init.d/$subsys ]; then
             /etc/init.d/$subsys stop
         else
             rm -f $i
         fi
done


At 03:26 PM 5/23/2002 -0700, Lincoln Peters wrote:

>I have gotten the netbooting system at RCHS to the point where only one 
>problem remains that I don't already have the answer to.
>
>When a netbooting client shuts down, I need to prevent it from terminating 
>dhcpcd, because if it does so at any point prior to shutdown, the system 
>hangs.  I fixed the symlinks that cause the network to shut down on any 
>runlevel (0, 1, or 6), but when it gets to "Sending all processes the TERM 
>signal", it terminates dhcpcd and the machine crashes (just like if I had 
>unplugged the hard disk on a typical computer).  Is there any way to 
>prevent the system from sending the TERM signal to dhcpcd specifically, or 
>maybe just prevent the "Sending processes the TERM signal" event from 
>happening at all?


Yeah, turning off the S00killall script in runlevel 0