AIX

Hello,

I'm having a problem starting the cron daemon automatically from inittab, let me provide the details below:

We are having five equally installed machines. One of them was upgraded in the past, one we upgraded recently, both from 5300-05-06 to 5300-07-01-0748. On the upgraded machine cron is not starting automatically from inittab, otherwise it is running perfectly fine. While trying to figure out the root of the cause, I tried running dummy script from inittab - it didn't start, too (and ran with no problem on another machine). Strange enough, the other system processes are starting just fine. Then I installed the last system backup on another machine and upgraded to TL6 first. Again, it didn't start on TL6. I compared all the installed packages and inittab head to head on the two 5300-07 machines (one working and one - not) - and they were equal. As far as I know, the problem existed on the first machine, but was fixed by a former SA. Unfortunately, due to poor documenting at his side, we don't know how and now we are trying to solve it (sounds familiar, doesn't it?). I googled and didn't find anything that suits.

For now, I'm using really dirty hack by starting cron from the rc.tcpip script, but this is certainly not an acceptable long-term solution. By Murphy's law, there are important scripts running from cron almost every minute.

I'll be really grateful if you can help me track the problem, please tell me if you need more specific data. Thanks in advance.

I don't use AIX, but as a shot in the dark, try:
telinit q
to make init reread /etc/inittab

I guess you checked the errpt?
Also checked "alog -f /var/adm/ras/bootlog -o"?

Since you want to put it into rc.tcpip, I guess that rc.tcpip is executed by inittab flawlessly, starting the stuff listed in there?

Maybe edit the line in inittab to:

Code:

cron:23456789:respawn:/usr/sbin/cron >> /tmp/cronlog.tmp 2>&1

and check if it writes anything to that file about the problems it has.

Thank you for your replies.

Yes, I checked the boot log and errpt, and yes, I tried to start with piping stdout, stderr to a file, it is not being even created, leave alone populated. As zaxxon said, other system services are being started fine by it, which is really strange. Maybe it's worth to mention that I tried either by viing the file directly or with the *itab commands. On other machines it works either way (though I know IBM recommend not editing it directly).

Of course, I executed "init q" every time I changed something.

Hmmm.. maybe check the inittab with "od -c" or "vi :set list" if there are some unwanted characters in that line of cron?
If inittab should be absolutely the same on a different box, copy them to one machine and try a cksum to check if they are the same.

Sorry for the delay, didn't have Internet access.

I checked it both ways, tried changing it - made no difference. Obviously, the file itself is intact.

I assume cron is not running - no sign of it in ps output.
This indicates inittab is stalled, probably on a wait command.
Either comment out everything in the inittab, from the last job showing in the console log until the cron entry, and then ad them back one at a time until you find the one blocking the inittab, or add an entry between each line that writes an update to a log file and then the log file should show where inittab stalls.
Or try checking the ps output (ps -ef) for any inittab wait processes still running, then kill them to see if the inittab continues and starts cron.

HTH