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.

Using Groups with Dependencies

We have an Orion instance in Chicago that monitors network and server hardware at our India offices in Bangalore.

The circuit is managed by a redundant pair of Cisco ISR 4451 routers.

When this circuit has severe degredation or drops due to vendor issues, our Bangalore managed nodes all bark.

In order to address this, I added the Bangalore nodes to a group (as the child) and then added both representations of the internet circuit from each of the ISR routers to a group (as the parent).

The ISR circuit group is set to reflect worst child status, so it should go critical if either managed interface in the HA pair goes down or packet loss is critical.

Since I'm not too familiar with using groups as dependencies, I wanted to get advice on whether or not this should work as intended.

  • On the surface it looks like it should work like gangbusters and is a clean way to cover the dependency alerting.

    Would it be possible to add in a Layer 3 interface somewhere on a Bangalore switch, attach the same level of alerting to that interface, "admin down" the interface and see if you get the intended results?  Seems to me the safest way to test the alerting capability without actually dropping critical services/nodes.  Of course topology and equipment used will be the deciding factor to running a test like this.

  • I'm pretty sure you want to have the circuit group reflect "best child status" so it only goes down if all objects are down.

    As a HA Pair, you don't want one circuit being down to suppress legitimate alerts on remote devices that are still reachable.

  • Sorry for the delay in response.. I believe you are correct, so I'll get back in there and take a fresh look.

    The testing option by Brandon should be useful too; providing I can get the network engineers to participate.