This discussion has been locked. The information referenced herein may be inaccurate due to age, software updates, or external references.
You can no longer post new replies to this discussion. If you have a similar question you can start a new discussion in this forum.

NPM dependencies and alerts with a ring network topology.

Hi everyone. I've run into some unexpected alerting behavior on my network and need a bit of insight on how I can effectively configure dependencies for my particular situation.

The network I'm monitoring has a number of distribution switches that are connected in a ring. Thus, each distro switch is connected to two other switches to the "left" and "right" of it.  [CORESW]---[DistroSW1]---[DistroSW2]---[DistroSW3]---[DistroSW4]---[DistroSW5]---[CORESW]. So if there is a break in the link between any two switches, none of them lose connectivity to the core. In addition, each distro switch has a number of access switches that all have a single connection to their respective distro switch.  Essentially it's a ring of hubs each with their own spokes. 

Note: All dependencies currently configured in my NPM instance are auto0generated.

Right now, we alert only on Node Down status because we want to suppress alerts for nodes that are only unreachable due to their parent being down.  This seems to work well for access switches when their respective distro switch goes into a Down state.  All of the dependent access switches become unreachable as expected.  The problem is that in some scenarios when we would want a distro switch to show as Down, it goes into Hardware Status Unknown and the access switches all go into a Down state instead of unreachable. Case in point: A single distro switch lost power, no alert was triggered on the switch itself, but the 5 access switched connected to it triggered Node Down alerts.

I analyzed the dependencies and found something interesting.
Solarwinds auto mapped the Distro to Distro dependencies in only one direction around the ring. [CORESW]-->[DistroSW1]-->[DistroSW2]-->[DistroSW3]-->[DistroSW4]-->[DistroSW5]-->[CORESW] so it seems as if Solarwinds saw that it could not poll [DistroSW3] but it could poll one of its children [DistroSW4]. I think this is why it considered [DistroSW3] to be in an Unknown state and the other children (the access switches) as Down.

With all of that background out of the way, here's the question I'd like to answer:
1. What's the best way to configure dependencies between the distro switches which are arranged in a ring? I'm considering making each Distro both a parent and child of its adjacent neighbors (bidirectional dependencies), but how would Solarwinds interpret a single node going down that is both a parent and a child to 2 nodes that are up? Is there a best practice for this kind of arrangement?

2. What would be the behavior if [DistroSW1] and [DistroSW4] lost power? Ideally, I would want them to go into a DOWN state and Distro 2 and 3 to be Unreachable. How can I set up Solarwinds to achieve the desired behavior in both of these scenarios?

I wasn't able to find any discussions addressing a scenario like this so I'm wondering if anyone here has a similar scenario or understands the behavior of Solarwinds well enough to offer some advice.