Service Dependencies "wildcarding" error ?

Carroll, Jim P [Contractor] jcarro10 at sprintspectrum.com
Wed Oct 16 21:29:11 CEST 2002


Hmm.  Alright, my response:
 
- if I were to install netsaint_statd, that would a) increase the complexity
of the installations, b) check for the existence of the NRPE daemon without
validating whether the daemon is even listening to the port
- increasing complexity where we don't have root on a number of our hosts,
as well as requiring yet another firewall rule to be pushed... I'm not too
keen on that
- I'm a big fan of the KISS principle
- assuming I stick with the KISS principle, then sticking with the NRPE
paradigm across all platforms (which includes NSClient for Windows) will
improve the happiness quotient at the end of the day
 
But you're making some good arguments in favour of sticking with one NRPE as
the 'master' service for that host.  I don't mind using check_tcp, because
it at least validates that *something* is listening on the desired port.  I
was just now thinking about creating a dummy test which always returns 0
(eg, /bin/true), but then again, it's a bit silly to repeatedly test for
that.  (Or is it?)  Now I just need to decide which of the many monitored
NRPE services should be The One.
 
Perhaps the better solution is to make a request that a future release of
NRPE allow for a rudimentary handshake verification, a built-in test, if you
will...?  The test could be called check_null.
 
After some more pondering,  a variation on this theme would involve
/bin/echo, thusly (in nrpe.cfg):
 
command[check_serial]=/bin/echo "42"
 
This could serve the function of 'master'.  It could also serve as a way of
tracking which version (or date-last-modified) of nrpe.cfg that you have on
that host.
 
Food for thought.  Don't mind me, I'm just thinking out loud.  :)
 
jc

-----Original Message-----
From: Paulo Pinto [mailto:japh at ip.pt]
Sent: Wednesday, October 16, 2002 1:06 PM
To: Carroll, Jim P [Contractor]; nagios-users at lists.sourceforge.net
Subject: Re: [Nagios-users] Service Dependencies "wildcarding" error ?


Ok, my mistake.
 
I've just upgraded from 1.0b5 to 1.0b6 and the wildcards on service
dependencies work as expected.
 
To Jim Carroll:
 
If you want to be ... hmm .. a bit paranoid about the tests, I'd suggest you
to run something like NSClient (windows) or netsaint_statd (*nix) on the
client machines and then check for the availability of your NRPE daemon
(using check_named_proc or similar).
 
On the other hand, to make things a *LOT* easier, just choose one of your
NRPE checked services, nominate it as your 'master' service and make all the
other NRPE checked services dependent of it. At least, if the daemon fails,
you'll get only one notification.
 
I wouldn't go for check_tcp ... I personally don't like the idea :)
(sometimes, a returned banner isn't a proof that the service is healthy ...
)
 
Hugz all.

----- Original Message ----- 
From: Carroll, Jim P [Contractor] <mailto:jcarro10 at sprintspectrum.com>  
To: 'Paulo Pinto' <mailto:japh at ip.pt>  ; Scott
<mailto:lists.scott at themagicbox.net>  ; nagios-users at lists.sourceforge.net
<mailto:nagios-users at lists.sourceforge.net>  
Sent: Wednesday, October 16, 2002 4:26 PM
Subject: RE: [Nagios-users] Service Dependencies "wildcarding" error ?

Paulo, that looks like an excellent scenario to streamline!
 
At some point I'll be trying my hand at service dependencies.  Here's the
scenario I hope to streamline:
 
- host1 is running NRPE
     - nrpe.cfg has several tests configured
- most other hosts will be similarly configured with NRPE
- some hosts will be running a superset of the NRPE definitions
 
Risk in this scenario:  if the NRPE daemon is down and Nagios starts
checking those services, naturally it will fail for each and every NRPE
service test defined.
 
Goal:  If NRPE is down, be notified that NRPE is down.  Silence on all the
NRPE tests.
 
Actually, I wouldn't mind some input on one aspect of setting this up.  I'm
torn between making one of the NRPE tests the service depended on by the
other NRPE tests (for a given host), versus just using check_tcp to verify
that the NRPE daemon is up and listening.  Thoughts anyone?
 
jc

-----Original Message-----
From: Paulo Pinto [mailto:japh at ip.pt]
Sent: Wednesday, October 16, 2002 7:07 AM
To: Scott; nagios-users at lists.sourceforge.net
Subject: Re: [Nagios-users] Service Dependencies "wildcarding" error ?


Hi.
 
Imagine the following scenario:
 
- 2 web servers, with 50 Vhosts each.
- 1 Alteon webswitch, virtualizing the 50 Vhosts
- Nagios is monitoring the 150 webservices. A failed service gets notified
by SMS. 
 
Now, without service dependencies, if a webserver goes down, you get 50 SMS.
If BOTH webservers go down, you get 150SMS. 
 
WITH service dependencies, you can make your webtests depend on the
RealServers status on the Alteon... if 1 webserver goes down, you get 1 SMS,
if both go down you got 2 sms.
That's why :-)
 
Anyway, I've already seen in the changelog that wildcarding was addedd on
1.0b6 (I'm running beta5) ... I'll upgrade it and see if it works :-)
 
Hugz all.
 

----- Original Message ----- 
From: Scott <mailto:lists.scott at themagicbox.net>  
To: Paulo Pinto <mailto:japh at ip.pt>  ; nagios-users at lists.sourceforge.net
<mailto:nagios-users at lists.sourceforge.net>  
Sent: Wednesday, October 16, 2002 11:32 AM
Subject: Re: [Nagios-users] Service Dependencies "wildcarding" error ?

how canb a service be dependant on all services like this? wouldnt you just
have a host dependancy and be done with it?

----- Original Message ----- 
From: Paulo Pinto <mailto:japh at ip.pt>  
To: nagios-users at lists.sourceforge.net
<mailto:nagios-users at lists.sourceforge.net>  
Sent: Tuesday, October 15, 2002 11:00 AM
Subject: [Nagios-users] Service Dependencies "wildcarding" error ?

Hi all.
 
I'm trying to setup service dependencies.
 
In my case, I want ALL services in host A to be dependent of service SERVICE
on host B.
 
So, I've configured dependencies.cfg like this:
 
define servicedependency{
        host_name                       HOST-B
        service_description             SERVICE
        dependent_host_name             HOST-A
        dependent_service_description   *
        execution_failure_criteria      n
        notification_failure_criteria   u,c
        }

 
Guess what:
 
$ /path/to/nagios/bin/nagios -v /path/to/nagios/etc/nagios.cfg
 
(...)
Error: Dependent service specified in service dependency for service '*' on
host 'HOST-A' is not defined anywhere!
Error: Dependent service specified in service dependency for service '*' on
host 'HOST-A' is not defined anywhere!
(...)
 
I'm using Nagios 1.0b5, btw
 
Any hints ?
 
Hugz,
 
Japh
 
 
 


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://www.monitoring-lists.org/archive/users/attachments/20021016/b2f2aa61/attachment.html>


More information about the Users mailing list