We have a fairly large nagios setup, running the standard Fedora nagios packages. E.g. nagios-2.9-1.fc6, nrpe-2.7-3.fc6.
The strange thing is for some reason, the our x86_64 based clients are unable to run the check_log NRPE client. The master is reporting, "NRPE: Unable to read output".
Note that in our setup, the nagios server is running on a i686, FC6. On our i686 client machines, everything works fine. I.e., all NRPE clients work.
On the x86_64 client machines, all NRPE commands work: LOAD, PROCS, FREE_SPACE, etc., but NOT check_log. We have checked permissions, etc. On the client machines:
su - nagios
$ sudo /usr/lib64/nagios/plugins/check_log -F /var/log/messages -O ~/check_log.messages.old -q "some error string"
Log check ok - 0 patter matches found
However, when the NRPE server tries to do this, we see the "NRPE Unable to read output" error. It is also clear that the client check_log is not even running because the -o output file is not being updated.
One further note, our nrpe client is run under xinetd. Again, all other NRPE commands are working.