Some Operations Manager environments are just too big too upgrade overnight. I am talking about a bunch of management groups with a dozen or more management servers, all utilizing a shared datawarehouse.
The Microsoft documentation isn’t that descriptive about what is supported if you can’t upgrade your entire infrastructure in one go. Here are my findings:
The moment you upgrade the RMS of the first management group, your datawarehouse gets upgraded. However, unlike your Operational Database, your datawarehouse remains backwards compatible with 2007-management groups. This is quite logical due to the fact that reports in the current management packs are compatible on both versions of OpsMgr, which utilize the datawarehouse to get historical data.
The bottom-line: both versions of OpsMgr can store and retrieve data from an upgraded datawarehouse without a hitch.
Mixed role compatibility
Lets say you have a management group with 300 gateways, 3000 agents and 30 management servers. This situation requires a planned approach, to prevent an overload on various resources. But is it possible to split up the upgrade of a single management group? And if so, in what order should it be split?
Well, the answer is yes. The rule of thumb is: upgrade downstream (in regards to the flow of monitoring data).
– 2012-agent –> 2007-management server: supported
– 2012-gateway –> 2007-management server: supported
– 2012-agent –> 2007-gateway: supported
– 2012-management server –> 2007-RMS: supported, but don’t keep this configuration for too long. Your operational database is semi-upgraded at this point, which means that should something crash you’ll have to perform a restore. This is the moment where you should start upgrading your agents and gateways (if not already upgraded) and then move straight on to your RMS-upgrade.
– Anything 2007 to anything 2012: not supported (except for management packs )
I find that we are provided with lots of flexibility regarding upgrade possibilities. So don’t rush your upgrade to the point that you either have too much downtime, or forget something and can start all over again. Make a phased approach, and the transition to 2012 will be much more relaxed and efficient.