AW: Cascading Services/Service hierarchy

[Robert Nelson]

> Just so I am clear on this. The status is not propogated from 
> a service to the dependant service. For example, we have a 
> web service that is dependant on the network service. If the 
> network service goes down, the web service is logially not 
> accessible. You can tell Nagios not check the Web services at 
> this point because it makes no sense to if the network is 
> down. However, the status of the Web service is unchanged. I 
> would have to explicitely check the Web service to see if it 
> were down or set the status of the Web service directly, 
> using something like send_nsca.  

If you use simple host parent/child relationships, it works in this
manner:

Host A (Switch) is tested by accessing the SNMP and HTTP Services.

Host B (web server) is tested by accessing the HTTP service

Service A.SNMP and A.HTTP fail. Nagios checks Host A and finds that it's
dead.

Obviously, host B should be down as well as the HTTP service on it.
Nagios runs the check anyway, marks it as down.

Two things will be recorded by nagios, a HOST DOWN message for host A
and a HOST UNREACHABLE message for host B. In your case, you would
probably only get the HOST DOWN notification and not see the HOST
UNREACHABLE (configurable, of course). However, Nagios notes both in its
logs.

At the end of the month when you spit out the downtime report for the
web server, you will see 80% uptime, 18% unreachable, and 2% down. You
will then know that 2% of the time, the service itself was down
(restarts, etc.). 18% of the month, something in front of it died that
prevented it from being accessed. The other 80% of the time, it was
still accessible.

If your relationship is between various hosts, then you might be
satisfied with the host parent directives. If you are monitoring
services on the same host, then you will need to use service
dependencies. Configuration of the parent directives is much easier than
service dependences, and IMHO would be better if the situation allowed
it. KISS, and all that.

An alternative is to set up two hosts with one service each, where each
host is the same IP/hostname, and use the host parent directive. We use
this when we do port forwards to devices in the RFC1918 address space,
so we know when we have a WAN connectivity problem or some sort of LAN
disfunction. Probably not as helpful if you've got a mail server that
depends on an SSH tunnel, or something else.

Rob Nelson
Network Engineer
Windchannel Communications
919-538-6326 


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM.
Deadline: Sept. 24. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Nagios-users mailing list
Nagios-users at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nagios-users
::: Please include Nagios version, plugin version (-v) and OS when reporting any issue. 
::: Messages without supporting info will risk being sent to /dev/null