Need deployment advice for multiple servers

[BanditBBS]

Ok, I'm going to try and explain this as good as possible.

We will be deploying a total of 5 XI servers and one Fusion(Mainly just for screens in NOC) server.

One of the XI servers will be monitoring all of the network switches and routers.
One of the XI servers will be monitoring all of our customer facing web servers/sites
One of the XI servers will be monitoring the linux servers in one of our distribution centers and other misc devices
The other 2 servers aren't important for the example I am going to give....

I was all ready to not do distributed monitoring until I just had a thought....parenting and dependents.

How can / What is best method to deploy like I wanted but still able to have dependency setup so if a switch goes down, the one server monitoring linux servers at the dc won't alert us because the switch there is unavailable. Is this even possible with how I was going to set it up? If not, then what do you suggest? We really don't want to get alerted for items that are unreachable due to a network outage or other issues, but I don't see how I can do it with how I stated we were going to deploy.

Thanks,
Jim

[scottwilkerson]

Jim,

I would add the relevant host's that you need to establish parent/child relationships (with a simple ping check), but I would add just the host.

This way you won't eat up too many resources, but will be able to take advantage of unreachable states.

[BanditBBS]

Scott,

That would let us see what is unreachable, sure. But the other XI server that is doing all the checks for that host would still alert that it is down. If this isn't doable it isn't a show stopper. Our network has so much redundancy built in, it should never happen. ****knocking on some wood****

We also own Tivoli Netcool and may be eventually rolling everything up into it from nagios, so if that happens, then this wouldn't be an issue anyway. I just wouldn't mind getting rid of Netcool and Alarmpoint(xmatters) and replacing it all with Nagios. In order to do that though, I would need to figure out this root cause analysis or whatever you want to call what I am trying to figure out here.

Jim

[scottwilkerson]

Jim,

I think I need to be a little more clear on what I was trying to convey.
What I meant when I said I would add the Host with just a ping check, was as follows.

Host = All switch/router's needed for the parent child relationship.

You would add this host to the XI servers that aren't monitoring the switch/routers, just as a simple host alive check.

Then you can establish your parent/child relationships.

Yes, this is slightly redundant as these same switches/routers will be being checked by the XI server tasked to do so, but simple ping checks are light weight, and you could disable notifications on the extra checks if you needed to.

[BanditBBS]

Ok, now I get it. That does make much more sense. I thought you were saying to do it the opposite way

That would work for what I was wanting.