README.monitoring

Monitoring for Eucalyptus
=========================

This document explains how to integrate Eucalyptus within a running nagios
and ganglia installations. You will then be able to monitor basic status
of a running Eucalyptus system. We'll start from Nagios, and you will see
ganglia instructions following.

You should check http://open.eucalyptus.com for newer version of these
scripts, up-to-date information and help.

NAGIOS
======

Prerequisite
============

Have a Eucalyptus (1.6 or higher)  installation running. You should have
one CLC and at least one Walrus, SC, CC and NC. For simplicity in this
document we expect to install nagios on the CLC machine and that all the
machines are 'visible' by the CLC machine. 

Install nagios on the CLC machine (for example for ubuntu/debian install
nagios3) and read the package instructions. You should be able to login to
the nagios graphical interface (usually http://localhost/nagios3) and see
your host informations there.

Setup
=====

We will use the passive cheks capability of nagios, and to do so you need
to check that the the named pipe exists. Usually is
	/var/lib/nagios3/rw/nagios.cmd
but you can check in nagios.cfg which file that is supposed to be. Nagios
should create it at startup when you change
	check_external_commands=1
into nagios,cfg.

To create the configuration for the eucalyptus machine, you can use the
script nagios.sh. For example
	nagios.sh -setup -nodes "two three four" -cc "one" -cloud one -walrus "one" > eucalyptus.cfg
assumes that hosts 'one' has the cloud manager, walrus and the cc, while
machines two, three and four are the nc. Inspect the generated file
(eucalyptus.cfg) and it all looks good, you can move it to the nagios
configuration directory (usually /etc/nagios3/conf.d) and restart nagios.

Running
=======

Nagios should now display the Eucalyptus machines and services, but there
won't be any data reporting yet. To do so you can now run the same
script, in the example 
	nagios.sh -nodes "two three four" -cc "one" -cloud one -walrus "one"
(usually the script runs in a cron job). Give sometime to nagios to
process the information (more so if you just started it) and you should
now be able to monitor the status of Eucalyptus hosts and services.

NOTE: some of the eucalyptus hosts may show a status of PENDING in some
nagios screens (Host detail in particular). This is no really true since
the tests are taken, but nagios report them as such becasue some tests
aren't under nagios control. You will see the correct values in the others
screens.

GANGLIA
=======

Prerequisite
============

Have a working ganglia installation. In particular you should one of more
gmond running on the machines you want monitored (nc, sc and walrus), and
at least one gmetad and one webfrontend to collect and to show all the
data. You should be able to see ganglia's graphs (usually at
http://localhost/ganglia).

You need to have a working Eucalyptus installation (1.6 or higher).

Setup
=====

You may want to customize ganglia (gmetad in particular) to be able to
group all the nodes under a 'cluster' to be able to see the aggregation of
the resources. In particular on nodes with multiple NICs you may have to
specify by hand which interface to bind to.

Eucalyptus (1.6 or higher) should already be setup and working.


Running
=======

To start pushing data into ganglia you can use the script ganglia.sh. You
will need to specify which type of machine you are monnitoring (nc for
node controller, sc for storage controller and walrus) and where
Eucalyptus is installed. For example
	ganglia.sh -type nc -d /opt/eucalyptus

You may want to run the above command line from a cronjob.