Wacky conf file issue

[wurm]

I am really, really stumped on this one. You will either think I'm crazy or a liar, but here's what is happening, if you think you can face the risk of losing your mind, too.

I use zabbix_agentd on AIX 5.3. I had put together an install script so others on my team could easily install the agent on their hosts. I wanted to test my install script, so I killed the agent, backed up all of the files normally used by zabbix, removed the zabbix user, etc. Ran the script, but the agent would not start... it would just hang using an entire CPU. I would put back all of the original files and it would start up fine. After a day of trial and error, I tracked it down to the conf file. Now check this out:

If I place a copy of my original conf file (from when it was working) in /etc/zabbix, the agent hangs using an entire CPU. If I move the original conf file into /etc/zabbix or link to the original, it works fine.
Permissions are identical between the two conf files and the directories before and after. Even aclget shows no differences.
'diff' shows no difference between the conf files.
'sum' shows no difference.
I have copied, linked, moved, renamed and otherwise switched the conf files back and forth dozens of times with the same result. Only one file and one file only will allow the agent to start.

Am I insane?? Am I an idiot?? Half my kingdom and my daughter's hand in marriage to anyone that can tell me what's going on?? Well, a hearty 'thank you' at any rate...

Help me unix gurus!

[Alexei]

Check permissions and ownership of agent's log files. In 1.4.1, ZABBIX daemons may possibly eat 100% of CPU time in case if a daemon does not have write permission to its log file. This is already fixed in pre 1.4.2.

[wurm]

Log file permissions are ok. When using the 'magical' conf file, the agent starts fine and puts entries in the log. No log entries are made when using the copy of the conf file.

I forgot to mention I am using Zabbix 1.4.

Also, I'm pretty sure the agent can find the conf file, because I tried to start it with an empty /etc/zabbix directory, and it gave me an error message that it couldn't find the conf file. When it hangs, I don't get that message.

Thanks,

[Alexei]

Did you make a diff on that box? Perhaps DOS/Unix style end of line difference?

You may also check IPCS resources after killing ZABBIX agent. Run 'ipcs' and use 'ipcrm' to remove it if there are any.

Also, try running zabbix agent from a command line:

./zabbix_agentd -c /etc/zabbix/zabbix_agentd.conf #old config
./zabbix_agentd -c <path to new one>/zabbix_agentd.conf #new config

Will it start?

[wurm]

Thank you! You have saved my sanity! ipcs revealed a semaphore in use by one of the previous but deleted 'zabbix' user id's that had been created during the testing of the install script. After removing it, zabbix_agentd began to behave normally again.

A hearty THANK YOU, Alexei!

Wurm