Nagios Support Forum

I'm trying my best to grasp the concept of parent/child relationship and dependencies.
My goal is to get minimize the number of notifications sent out in case of something
going down or becoming unreachable.

My environment is redundantly setup like this;
Internet -> 2 redundant routers -> 2 node FW Cluster -> 2 redundant Switches -> 3 ESX
HostsVMware Cluster

My Nagios server is physical and is connected to both the switches and so is a couple
of other servers. All other servers are virtual guests in VMware.

Now how should I design the parent/child relationships? I suppose I should go from
Nagios perspective of things? All VM guests should have the 3 ESX hosts as their parent.
The ESX hosts should list the 2 switches as parent and these should have the
2 firewalls as parents. Last but not least the firewalls should have the routers as parents.
Reason for me being uncertain is because the map only shows one parent relation
despite me listing for example both switches (perhaps this is by design or am I missing
something).

Also what should I use dependencies for? Is this for stuff not shown by parent/child relations
of how traffic flows. Like for example my SharePoint server is dependent upon my SQL server.

Thanks in advance

Now how should I design the parent/child relationships? I suppose I should go from
Nagios perspective of things?

Yes, you should setup parent/child relationships from the Nagios perspective. This will reduce your "unwanted" alerts, when children are unreachable.

Unfortunately, the hypermap cannot show parent/child relationships properly, when there is more than one parent. Is seems, this is a bug in hypermap, and I will be reporting this on our bug tracker as soon as I can. You can use the Network Status Map instead - it will show these relationships properly.

Hope this helps.

I just posted a bug report on our bug tracker. You can view it here:

http://tracker.nagios.com/view.php?id=308

Thanks so much for the answer and submitting of the bug.
I do appreciate it very much.

How 'bout the dependencies question, have I gotten that
right. That this is to be used for dependencies not handled
by the parent/child relations. I mean this is for dependencies
where traffic doesn't pass through another host but merely
where a process is dependant on a process on another host,
as in my example with Sharepoint depends on the availability
of my MS SQL server. From the users perspective they do not
talk with the MS SQL server. Tehy talk to the Sharepoint and
the Sharepoint server in its turn talk to the MS SQL and then
send the reply back to the client.
Neither the Sharepoint nor the MS SQL is a parent/child to
one another, they're both at the same level. Having the ESX
servers as their parents. Am I correct in assuming this is a
example for setting up a dependency for the Sharepoint
server listing the MS SQL?
Another example would be with monitoring VMWare. This is
done in my case by getting info out of the Virtual Center
server. Now Virtual Center is not a parent nor child to the
ESX boxes. It's a physical box connected to my switches.
In this case I assume that the Virtual center server should
be listed as a dependecny of the ESX hosts?

Thanks

This is correct, dependencies can be service to service, host/service, etc.

Basic rule of thumb is what you had outlined in the first post

iprovidence wrote:Is this for stuff not shown by parent/child relations
of how traffic flows. Like for example my SharePoint server is dependent upon my SQL server.

This is an excellent example. If the dependency would be covered by the host parent/child relationship there is no need to add it, but if not you can add one.

Thanks so much for helping me
understand this.