AIX 4.2 login problem

[xMaximex]

Hi,


I'm on a project to keep a AIX 4.2 server alive for at least 6 month
before it got changed for a new one. The problem on this server is that
about once a day it become impossible de logon on a telnet session and
even on the console. Last week we were able to connect to telnet but we
got no login prompt at all. This week when the problem occurs, we see
the login prompt but after I enter the username nothing happen. The
only way to fix it is to reboot the server. I found a way to have a
command prompt even if the server is f**ked up. I set a cronjob that
open a xterm on a remote X server every 1 minutes, so I can execute
some commands. I see with ps -ef that every time I try to logon by
telnet or on the console there's a process that stay here : "login -h
x.x.x.x -p". I don't see anything in the error log.


What could cause that problem ? I'm currently monitoring the "netstat
-m" memory use ..


Thanks and sorry for my bad english.

 

[p5wizard]

I would also monitor paging space, but sth like that would also show up in the error log. mbuf shortage should also show up in the errlog - but on 4.2 maybe it doesn't...

HTH,

p5wizard

 

[Mag0007]

make sure your networking guys aren't doing anything, like maintenece

 

[xMaximex]

I already monitored swap space and no, the network guy isn't doing any maintenance

 

[p5wizard]

run a vmstat 10 w/ output to a file for a day and try and correlate cpu usage/paging with your problem occurance.

put this in a script time_vmstat
while true
do
date +'%Y%m%d %T'
sleep 300
done &
vmstat 10

and run the script:
time_vmstat >/tmp/vmstat.log
make sure there's enough space in /tmp - choose another location if necessary.

(on newer AIX versions you can do vmstat -t to include a timestamp, but I don't think it's available on 4.2)

remember to kill the vmstat script after you're through.

Also: a ps -el minute by minute may help to identify the processes that are eating up your CPU (just a guess)





HTH,

p5wizard