logstash-forwarder init script for CentOS 5 and 6

[Rumbles]

I have had absolutely no luck getting the init script provided here to work with CentOS 5 or 6.

Luckily I found someone else had written another script which works well with CentOS 6 and mostly works with CentOS 5. The issue I have with CentOS 5 is when I reboot the server (even though I have chkconfig on set for the init script) the logstash-forwarder does not fire up at all (after reboot if you check the status you get the response "logstash-forwarder dead but pid file exists"). I have put a question on Server Fault to see if anyone there can help me, as I know this isn't probably the best place to ask about this finer differences between CentOS 5 and 6 init scripts, but when I went in your IRC channel I was advised to raise an issue to at least make you aware of the problem I have encountered.

Just to note, I have changed the init script slightly to stop the logstash-forwarder logs going to messages as I am using forwarder to ship those logs, but the issue was apparent before I made these changes also. Any help, suggestions or debugging tricks would be appreciated. Thanks

[driskell]

The problem could be that when using -log-to-syslog, which it seems you are, the syslog service must be started or forwarder exits failing to attach to syslog. You can catch this with a stdout/stderr redirect to a file in the init script.

Though with 99 startup priority is expect syslog to have been started. But maybe the redirecting will catch another error.

[driskell]

I say seems you are but I meant you probably aren't heh if not sending to messages.

[Rumbles]

Initially I was sending to messages, but now I have removed that part of the init script and I am still having the same issue

[driskell]

Did you redirect output to a file and check contents after a reboot?

[Rumbles]

I didn't I just removed the -log-to-syslog part, I just tried turning back on the log-to-syslog part and found that I got this slightly wrong. The service is running after I reboot, but the pid file is blank when the service is fired up, so when I check the status it says dead because there is no pid to check against.

I'm in the centos irc channel now trying to figure out why the pid file is created blank....

[driskell]

Try this script, maybe a problem with the one you found.

#196

[driskell]

Problem with the script you used above is probably "pidofproc" requires full path to the binary ($DAEMON) and not the basename ($NAME) - so it can't find the binary.

[Rumbles]

Thanks, I have tried the new init script and my service starts on boot, and I can check it with the service command. Which is great. The only oddity is if I restart the service it doesn't confirm the service has been started. If I stop it I get the normal "Stopping logstash-forwarder [ ok ]" message but on the starting message I get no ok at the end.... not a massive problem to be honest, but just thought I would note it. I will mention it on the Issue. Thanks for the help!

[Rumbles]

I'll mark the issue closed, thanks for the help again!