This script had been perpetuated as a service alerting script for years in the environs I work in, edited and passed on as working. It reads:

if [failure code here]; then
# return failure to nagios nrpe daemon
else
# return success to nrpe
fi

The test for a failure failed to match, even when the test was looking for a jboss instance not present on the server. Because the code falls through to success, everything looks just fine, all of the time.

The test for failure being incorrect struck me first when I looked through the script. The more basic flaw in logic hit me after. I think it would be true in an alerting script you would NEVER drop through a loop to a final success. The test would be for success, the fall-through to failure. This would have prevented a false sense of security, and the failure to detect success would have been dealt with immediately.

Something like:

if [success code here]; then
# return success to nagios nrpe daemon
else
# return failure nrpe
fi

Obvious. But an epiphany anyway.

–doug