I think this is a pretty common scenario - hopefully it helps someone else too.
Basic Multihomed Edge to Core switch connection:
Core A Interface 1 -> User C Interface 1
Core B Interface 1 -> User C Interface 2
I made a couple groups representing these interfaces. I made manual dependency: User C Group (IF 1&2) the Parent of Core Group (A&B IF 1). *Yes - ignore VPC's and such, trying to keep this simple
I put User Switch C into maintenance mode to reboot it. I didn't get an Alerts related to User C, which is good, HOWEVER, I still got the Alerts for IF down from each Core.
I assumed as Core IF's are children of User IF's (via groups), I wouldn't get any Alerts from the Cores? Then I started thinking (I know, scary...) - If the suppression logic depends on an Alert, and there is no Alert because of maintenance, maybe this won't work?
This seems like a common thing, no? Interested how others address this. I'm sure I could create a bunch of groups with all related objects/entities in each, and then maintenance the entire group? That's a lot of work...
Also interested in any white papers that get into the weeds about how auto-dependencies are built and how all this suppression logic is tied together, how long event A waits for event B, Best Practices to ensure this works as designed, etc. When view auto-dependencies, I expected to see at least the basic node level parent-child relationships, but not much there. You know, like for a remote site the WAN router is the parent of everything at that site; such that if the WAN link fails I only get one Alert and not 100...
TIA
Gary
