Best practice for config files

[lraymond]

ok I had some help from the board with some initial things and now I'm running but ready to roll this out to more servers and just wanted some help/direction as to the best way to go about it. I have 3 windows servers which are in one windows config, a host-group, all separate IP's and things are working perfect.

I have added a test remote linux server with a public IP, NRPE and monitoring some basics like load, processes and there working perfect. That box is under one config (linux-remote). My next step is to monitor my tomcat servers (I have 5) which are behind a loadbalancer. The LB has some 1-1 virtual services so from the nagios server I can issue;
/usr/lib/nagios/plugins/check_nrpe -H n.tc1.domain.com -p 5667
NRPE v2.12 is what get's returned.

I can do the same for 5668 which goes to server 2 (all the servers are listening on 5666 and the LB is switching). Now finally the question(s).

I don't think I can define a port at the host level can I? Basically I want a tomcat.cfg with something like this;

define host{
use linux-remote
host_name tc1
address 1.1.1.1
port 5667
}

By doing that I can then just define;
define service{
use generic-service,srv-pnp
host_name tc1
service_description Server Load
check_command check_nrpe!check_load
}

A basic service. I know I can define the port in the service, but then if I had 25 servers behind the LB, I would have 25 * number of services which would be HUGE config wise. So I was looking to make a hostgroup of each group and give them one service but not sure if it's possible.

That is the basic set of what I want. If I can't and need to define each service port, should I still keep under one file, or make a file for each server?

Thanks

[jsmurphy]

In your commands.cfg look for your check_nrpe definition... it will look similar to:

Code: Select all

$USER1$/check_nrpe -H $HOSTADDRESS$ -u -t 30 -c $ARG1$ $ARG2$

so the easiest way to fix this would be to change your service definition to:

Code: Select all

define service{
use generic-service,srv-pnp
host_name tc1
service_description Server Load
check_command check_nrpe!check_load!-p 5668
}

Basically we are inserting the -p 5668 switch into where ARG2 is, which is used to tell check_nrpe which port to connect to. If however you are using this same service on all the tomcat servers instead of 1 service per server, you could accomplish this using host macros.

Code: Select all

define host{
use	 linux-remote
host_name	 tc1
address	1.1.1.1
_tcport 5668
}

define host{
use	 linux-remote
host_name	 tc2
address	2.2.2.2
_tcport 5669
}

define service{
use generic-service,srv-pnp
host_name tc1,tc2
service_description Server Load
check_command check_nrpe!check_load!-p $_HOSTTCPORT$
}

More information on custom macros: http://nagios.sourceforge.net/docs/3_0/ ... tvars.html and http://nagios.sourceforge.net/docs/3_0/macros.html

[lraymond]

That is EXACTLY what I was looking to do. I didn't want 50 services, I simply wanted each host to have one little variable and use the same service. Each NRPE host is listening on the default 5666, but to test each, I have the loadbalancer listen on multiple, so tc1 will hit the public IP port 5667 which sends to server 1 port 5666, tc2 same IP port 5668, etc.

Will test this and close if all good.

Thanks.

[slansing]

Great! Let us know if it worked out.