Nagios XI correlation dependency between two hosts

johnjones · 05-09-2016, 10:58 AM

I have two hosts. If host1 goes down and host2 is up the Nagios must not create a critical alert as the host2 is still up. Instead it must create a warning alert for the host1 so that there is no need for immediate action. The same must be done when the host2 is down and host1 is up. Can anyone please help how this can be done on Nagios XI??

TenTenths · 05-10-2016, 03:15 AM

Off the top of my head I can't think of an easy way to do this in Nagios, possibly something to do with dependencies?

What I'd do though is write a custom check script that takes two hosts as parameters, the first being the main host to check and the second being its slave.

Run your check and if the main host fails but the second host is up then return the warning alert.

So the pseudocode (I'm too lazy to write the whole script here!) would be something like

Code:

twincheck.sh hosta hostb

Code:

#!/bin/bash
HOSTA=$1
HOSTB=$2
check_host_ping -h ${HOSTA} [plus whatever other parameters for timeout, etc.]
AUP=$?

if ${AUP} = 0; then
  # A is up, so we're fine here, even if B is down.
  # We don't need to check B because if it's down
  # then our check of B A will catch it.
  echo "OK"
  exit 0
fi

if ${AUP} != 0; then
  # A Isn't up, let's check our twin
  check_host_ping -h ${HOSTB} [plus whatever other parameters for timeout, etc.]
  BUP=$?
fi

if ${BUP} = 0; then
  # A is down, but B is up, so we exit with a warning
  echo "Down but twin up"
  exit 1
fi

if ${BUP} != 0; then
  # A is down and B isn't in an "OK" state
  # So we exit with a CRITICAL
  echo "Down and twin down"
  exit 2
fi

MadeInGermany · 05-12-2016, 02:12 PM

This script is a bit closer to the requirement:

Code:

#!/bin/bash
err=0; errhost=""
for arg
do
  check_host_ping -h "$arg" || {
    ((err++)); errhost="$errhost $arg"
  }
done
case $err in
0)
  echo "OK"; exit 0
;;
1)
  echo "WARNING: host $errhost is down"; exit 1
;;
*)
  echo "CRITICAL: hosts $errhost are down"; exit 2
;;
esac

You can also look at host and service dependencies, but this can only suppress notifications (or suppress further service checks).