we plan the monitoring based on SCOM 2016. Our IT and organizational infrasructure is complex. There exist many division/sub-divisions which have their own server responsibilities.
We have Windows, Linux Computers, network components, etc. Independent of this complexity, we have following aims in the Management Packs authoring:
1) We want to create Management Pack for each division/sub-division. We want to be flexible in our deployment without to take impact on the other divisions/sub-divisions.
2) We want to create independent of operation system version (2008, 2012, R2, 2016, etc.) one basic Management Pack for Windows and one basic Management pack for Linux. These base OS Mamanegment Packs will include base operation system-related services (Windows Update, etc.), system availability in network (ping, rdp, etc.), system resources (CPU, memory, etc.).
3) It is also very important that the server responsibilities in each division/sub-divisions will monitor only and only their own computers, get alerts only and only from their own computers, etc.
What is the best strategy for Management Pack authoring for that without thousands of management packs, thousands of groups, thousands of notification groups, etc.?
I’m not sure I see the benefit of creating an “all-in-one” OS MP. There are some differences between each version of Windows…some perfmon counters are named slightly differently and some services are different. So you would need to enable/disable groups of monitors/rules via overrides and you would still have to target those overrides at custom groups. I don’t see the point of going through all of that effort.
If you are concerned about the manageability of “thousands” of notification groups (would there really be that many?), there are some things you can do to automate the process to make it easier. For example, you could enable dynamic group population based on reg-key values that you define. That would essentially automate the population of notification groups.