[Linux-cluster] How can i Monitor Red Hat Cluster suite with 2 nodes

Ralph.Grothe at itdz-berlin.de Ralph.Grothe at itdz-berlin.de
Fri Sep 28 08:18:36 UTC 2012


_______________________________

	From: linux-cluster-bounces at redhat.com
[mailto:linux-cluster-bounces at redhat.com] On Behalf Of Ben
.T.George
	Sent: Thursday, September 27, 2012 11:37
	To: linux-cluster
	Subject: [Linux-cluster] How can i Monitor Red Hat
Cluster suite with 2 nodes
	
	
	HI
	
	How can i Monitor Red Hat Cluster suite with 2 nodes. i
am looking for web based monitoring tools.
	
	regards,
	Ben
	



Hi Ben,


have you had a search at the Monitoring Exchange website yet?

It is a kind of repository / exchange platform for third party
Nagios, Icinga, Shinken etc. plug-ins for virtually every
conceivable kind of service check.

I have just looked and were output several plug-ins for RHCS
clusters.

For instance this plug-in written in Python: 
https://www.monitoringexchange.org/inventory/Check-Plugins/Softwa
re/Red-Hat-Cluster-Suite-Check

I haven't deployed it myself, but just now had a short glimps at
it.
Since I am more into Perl I some time ago installed another
similar Plug-in from that source, and had written my own stuff,
which I haven't shared since I don't want to support it or being
held liable for the code.
But I know sufficient Python to read what it's doing.
Basically, it simply parses the clustat output dumped in XML
(i.e. using "-f").
Therefore the plug-in needs to pull in an XML parser from another
Python module.
My older Perl variant simply parsed the usual textual output of
clustat.
With my own stuff I made use of the abundance of features of the
check_multi Plug-in (of course, also written in Perl).
Also, as its name implies, it collects several service checks on
a Nagios monitored host (such as cluster a node or cluster
service (i.e. a VIP)) to reduce the overhead of connecting and
querying a daemon like nrpe for each check individually.

I would suggest to give the several third party Nagios plug-ins
from Monitoring Exchange a try and stay with the one that meets
your requirements best.
Without having looked at all the others I would assume that they
generally rely on either a Perl, Python or Ruby interpreter on
the monitored host.
These (maybe apart from Ruby) are usally already installed on a
RHEL host per default.
Sometimes, if a plug-in requires more exotic modules/libraries 
(see the list of modules following the "import" or "use"
statements in the plug-in's header, but usually the interpreter
will complain about missing modules when invoked on the plug-in)
you need to install these on the monitored host as well.

Have you had a look at the NRPE section in the Nagios
documentation yet?
If you installed and configured your Nagios web interface
correctly you will always have this documentation readily
available with no need for an Internet connection.
If not, I should advise you to read it to find out how to set up
nrpe on your cluster nodes (only a matter of minutes).

Another option to have a Nagios plug-in executed on a remote host
would be to make use of the check_by_ssh plug-in that comes with
the standard Nagios plug-ins.
This requires the distribution of public RSA or DSA keys in the
$HOME/.ssh of an account on your cluster nodes that has
sufficient privilages to execute the system commands of the
plug-in.
If elevated privileges are required for that, usually sudo is an
option.

Yet another way to get check results of plug-ins to the Nagios
server would be, as Michael already referred to, to use SNMP.
That's usually fiddling with the snmpd on the cluster nodes to
extend it to run some script or the like, I guess.
But I cannot further comment on that as in our network
environment SNMP for us mere OS sysadmins is frowned upon, or
rather made impossible (viz. by firewalls).
Honestly, I find this a bit schizophrenic since our network admin
folks make ample use of SNMP to monitor their network gear.
But in a company the choice of monitoring tool and which plug-ins
to use for it often is dictated by policy.


Hope that would help a little

Ralph











	





More information about the Linux-cluster mailing list