tcpip occasionally dies without explanation 1

[SteveB42]

Hello all,

I have this IBM 950 server that is our /home server. Periodically (and usually late at night) the tcp/ip daemon just 'disappears.' Preventing remote log on and kicking off anyone who is logged on. I have to go to the local terminal and execute rc.tcpip and it starts up just fine... sometimes the rc.nfs daemon needs to be restarted as well. Has anyone ever seen this? I have poked around and can't find a causal relationship at all. Not an emergency by any means, but if you have any ideas, they would be appreciated. Thanks a lot.

Steve

 

[gheist]

If this is public-access server somebody might be doing something nasty.

You are denying a relationship you just found:
Periodically (and usually late at night)

- a cron daemon runs processes usually late at night - could that be a problem?

 

[SteveB42]

No, no public access... and the cron jobs don't seem to be the culprit... again, not a big deal, but would like to find out what this buggaboo really is...

 

[gheist]

what messages are found in syslog and errpt?

 

[rishard]

check your /etc/inittab entry for rc.tcpip. this should restart you tcpip daemon if it dies.
this is what the entry should look like :
rctcpip:2:wait:/etc/rc.tcpip > /dev/console 2>&1 # Start TCP/IP daemons

hope it sheds more light

ciao

 

[Yegolev]

Shouldn't the 3rd field be "respawn"? Like so:

srcmstr:2:respawn:/usr/sbin/srcmstr

My various /etc/rc.tcpip lines are set to "wait" as well, but ours do not restart themselves as far as I know.

 

[Guest_imported]

Hi,

I have an AIX box running on 4.3.3

Are there any files that I can set default router and hostname.

THX

Steve

 

[Yegolev]

I would use SMIT to configure TCP/IP:

smit ethernet

smit tcpip

Or you can add routes at the cmd line:

route add default xxx.xxx.xxx.xxx

To remove, use:

route delete xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx

The last 32-bit word is optional but sometimes necessary if you have similar routes defined. It is a lot safer to do route work via smit. Be careful.

If you just want to look, use the netstat command.

 

[gheist]

and you can use "hostid" utility to set permanent hostname in AIX

 

[MoshiachNow]

Also,

1. Check errpt for "APPLICATION ERRORS" time related to the tcpip failures.Normally this will assosiate with some "core" file on the system,who's reason can be determined by running the following on the core file:

lquerypv -h core 6b0 64

Look for the program name in column 5.

2.Secondly - I would run

entstat enX

on the rellevant interface to check for errors.

3.errpt |grep en

To look for enX problems

4.Also,a script can be written to poll every minute the system processes for tcpip processes presense ,reporting what applications are currently running the minute the tcpip failes.
Sometimes can give an idea on what caused the problem
"Long live king Moshiach !"

 

[MoshiachNow]

P.S.

The above mentioned script can even send you a mail with thefound results and greps from the needed logs... "Long live king Moshiach !"

 

[SteveB42]

Hi Lev,

Have that script handy?

Steve

 

[MoshiachNow]

#!/bin/ksh
#This scripts catches "inetd" daemon presense every 10 seconds ,creates log and notifies via email if inetd disappears in “ps –ef”
#Needs to be added to /etc/inittab ,to be run always after reboot
#-------_test.sh------------------------------------------------------------------------------------------
#Put it in an script called "tcp-gone.sh" in /scitex/version/scripts (chmod 711 tcp-gone.sh)
#Can cause it to be performed every minute as following :
#crontab -e
#as root, and appending the following line to the file that will come into vi:
#* * * * * /scitex/version/scripts/ tcp-gone.sh
#BUT – then remove the LOOP !!!!!

DATE=`date`
LOGFILE="/scitex/version/logs/tcp-gone.log"
LOGFILE1="/tmp/tcp-gone.log"

if [[ -z $LOGFILE ]] ;then
touch $LOGFILE
fi

while true
do
tail -5000 $LOGFILE > $LOGFILE$$
mv $LOGFILE$$ $LOGFILE
chmod 666 $LOGFILE
ps -ef |grep inetd |grep –v grep
if [[ $? != 0 ]] ;then
echo "----------------------------------">> $LOGFILE
echo $DATE >> $LOGFILE;
echo " ----------Starting (ps -ef) ---------------------- " >> $LOGFILE
ps -ef >> $LOGFILE;
echo " " >> $LOGFILE
echo " ----------Starting (swho) ---------------------- " >> $LOGFILE
/usr/local/es/swho >> $LOGFILE ;
echo " " >> $LOGFILE
echo " "
tail -500 $LOGFILE > $LOGFILE1
mail -s tcp-gone <email-address> <$LOGFILE1
exit 1
fi
sleep 10
done




&quot;Long live king Moshiach !&quot;

 

[avadh]

hello,
instfix -i | grep ML ->
what level of AIX you are running.
What level is bos.tcp filesets at.

 

[SteveB42]

Thank you Levw, that script was very helpful. I modified it considerably... set it to run every 10 minutes via crontab and it has already worked once. Again, thanks!