check_ganglia_metric 2011.09.09

Introduction

check_ganglia_metric is a Nagios plugin that allows you to trigger alerts on any Ganglia metric. This plugin was heavily inspired by Vladimir Vuksan’s check_ganglia_metric.php, but it comes with a number of improvements.

Installation

# pip install check_ganglia_metric

…or:

# easy_install check_ganglia_metric

Ganglia Configuration

Unless your Nagios server and Ganglia Meta Daemon are running on the same host, You probably need to edit your gmetad.conf to allow remote connections from your Nagios server.

To allow connections from nagios-server.example.com:

trusted_hosts nagios-server.example.com

To allow connections from all hosts (probably a security risk):

all_trusted on

Testing on the Command Line

First, let’s see if check_ganglia_metric can communicate with the Ganglia Meta Daemon:

$ check_ganglia_metric.py --gmetad_host=gmetad-server.example.com \
  --metric_host=host.example.com --metric_name=cpu_idle
Status Ok, CPU Idle = 99.3 %|cpu_idle=99.3%;;;;

The “Status Ok” message indicates that check_ganglia_metric is working. If you’re having trouble getting this to work, try again with verbose logging enabled (--verbose) in order to gain better insight into what’s going wrong.

Now let’s try setting an alert threshold:

$ check_ganglia_metric.py --gmetad_host=gmetad-server.example.com \
  --metric_host=host.example.com --metric_name=cpu_idle --critical=99
Status Critical, CPU Idle = 99.6 %|cpu_idle=99.6%;;99;;

We told check_ganglia_metric to return a “Critical” status if the Idle CPU was greater than 99. The “Status Critical” message indicates that it worked. Note that check_ganglia_metric uses NagAconda to parse ranges and thresholds according to the official Nagios plugin development guidelines.

To see a complete list of command line options with brief explanations, run check_ganglia_metric with the --help option.

Nagios Configuration

First, create a command definition:

define command {
  command_name  check_ganglia_metric
  command_line  /usr/bin/check_ganglia_metric.py --gmetad_host=gmetad-server.example.com --metric_host=$HOSTADDRESS$ --metric_name=$ARG1$ --warning=$ARG2$ --critical=$ARG3$
}

Now you can use the above command in your service definitions:

define service {
  service_description  CPU idle - Ganglia
  use                  some_template
  check_command        check_ganglia_metric!cpu_idle!0:20!0:0
  host_name            host.example.com
}

This will work fine until something goes wrong with check_ganglia_metric (e.g. the cache file can’t be read/written to, the Ganglia Meta Daemon can’t be reached, etc.). At that point, every service that relies on check_ganglia_metric will fail, possibly inundating you with alerts. We can prevent this through the use of service dependencies.

The first thing we need is a command definition for checking the age of a file:

define command {
  command_name  check_file_age
  command_line  /usr/lib/nagios/plugins/check_file_age -f $ARG1$ -w $ARG2$ -c $ARG3$
}

Next, we define a service which checks the age of check_ganglia_metric’s cache file. Note that in order to be truly effective, this service needs to be checked at least as (preferably more) frequently than all the other checks that rely on check_ganglia_metric:

define service {
  service_description  Cache for check_ganglia_metric
  use                  some_template
  check_command        check_file_age!/var/lib/nagios/.check_ganglia_metric.cache!60!120
  host_name            localhost
  check_interval       1
  max_check_attempts   1
}

And finally, we set up the actual service dependency. Note that I’ve enabled use_regexp_matching in Nagios, which allows me to use regular expressions in my directives. By sticking “- Ganglia” at the end of every service that relies on check_ganglia_metric, I can save myself a lot of effort:

define servicedependency {
  host_name                      localhost
  service_description            Cache for check_ganglia_metric
  dependent_host_name            .*
  dependent_service_description  .* \- Ganglia$
  execution_failure_criteria     c,p
}

Now if something goes wrong with check_ganglia_metric, only one alert will be sent out about the cache file, and all dependent service checks will be paused until you fix the problem that caused check_ganglia_metric to fail. Once the problem is fixed, you’ll need to update the timestamp on the cache file in order to put the “Cache for check_ganglia_metric” service back into an OK state (which will allow dependent service checks to continue):

$ touch /var/lib/nagios/.check_ganglia_metric.cache

Tips and Tricks

It’s possible to get a complete list of available hosts and metrics by enabling “more verbose” logging (-vv). Since the metric_host and metric_name options are required, you have a little bit of a “chicken and egg” problem here, but that’s OK. Just supply some dummy data. The plugin will error out at the end with a “host/metric not found” error, but not before it dumps its cache:

$ check_ganglia_metric.py --gmetad_host=gmetad-server.example.com \
  --metric_host=dummy --metric_name=dummy -vv

Known Issues

• Doesn’t work with Python 2.4

Author

Michael Paul Thomas Conigliaro <mike [at] conigliaro [dot] org>