nick's site

centos 5’s implementation (probably many OSs too) of xinetd puts a connection limit per source IP, by default.  this causes the service to become unresponsive when there are dozens of queries a second.  in my situation, it caused nagios’s NRPE agent to become unresponsive.  not being able to monitor hosts causes alerts.  alerts cause me angry.  from syslog/messages:

Jun 27 22:13:28 nms1 xinetd[2592]: FAIL: nrpe per_source_limit from=123.456.798.123
Jun 27 22:13:28 nms1 xinetd[2592]: FAIL: nrpe per_source_limit from=123.456.798.123
Jun 27 22:18:28 nms1 xinetd[2592]: FAIL: nrpe per_source_limit from=123.456.798.123
Jun 28 01:38:56 nms1 xinetd[2592]: FAIL: nrpe per_source_limit from=123.456.798.123
Jun 28 01:38:56 nms1 xinetd[2592]: FAIL: nrpe per_source_limit from=123.456.798.123
Jun 28 16:50:00 nms1 xinetd[2592]: FAIL: nrpe per_source_limit from=123.456.798.123

we didn’t encounter this until we hit like, 100 checks every three minutes.

to fix this, update the service’s xinetd config. in my case, /etc/xinetd.d/nrpe


service nrpe
{
option=var
per_source = UNLIMITED
instances = UNLIMITED
moreoption=bla
}

restart xinetd.