[Nagiosplug-help] check_nrpe plugins work fine for a while then start throwing "NRPE: unable to read output"

[Matthias Eble]

Hi Kyle,

Am Mittwoch, den 20.05.2009, 15:56 -0700 schrieb Kyle Smith:
> We have some plugins written in Ruby (boss loves it, I am learning
> it :-)) that work fine with a freshly started nrpe daemon. Start up
> nrpe on the client, force checks from the gui, see a green nagios
> console, nice. But after a while in the console under "Status
> Information" we start seeing "NRPE: unable to read output" rather than
> our normal success messages. Bounce the nrpe on the client, force the
> check through the gui, and things start working again for a while. 

"NRPE: Unable to read output" is returned if the checkcommand's stdout
is empty. Have you tried to capture stdout *and* stderr to a logfile or
so? For example using a wrapper script. Maybe this can help finding the
root cause. 

> Another data point - We have one system where things were working fine
> but we needed to start passing params to nrpe. I recompiled after
> running ./configure --enable-command-args and setting
> dont_blame_nrpe=1 in the nrpe.cfg. Start up a new nrpe and it works
> fine, can pass params from the server no problem...but the a check
> that was working fine for weeks starts throwing "NRPE: unable to read
> output" after 3 or 4 successful invocations. Arg!! I had to rollback
> to the previous nrpe and check_nrpe.

Do all other affected systems also have args enabled?

> I am trying to set nrpe debug on in nrpe.cfg on the client and then
> bounce nrpe but i don't see any more messages in /var/log/system.log
> as compared to when I run with nrpe debug turned off. It's like
> turning on debug has no effect. 

Maybe a dumb question.. Is your syslog configured to print debug
messages?

Matthias