Since the introduction of Enterprise Manager 12c, folks have been asking for a list of best practices. I know a lot of you have been waiting for this post!
1.Use previously deployed, older hardware for your Enterprise Manager deployment on 13c.
Enterprise Manager is a simple, single service system. There is no need for adequate resources and ability to scale. In fact, I’ll soon be posting on my blog about building an EM13c on a Raspberry Pi 3.
2. Please feel free to add new schemas, objects and ETL’s to the Oracle Management Repository, (OMR.)
This database doesn’t have enough to do with metric collections, data rollup, plugin, metric extensions and notifications.
3. Turn on the standard statistics jobs and baseline collection jobs on the OMR.
The OMR has its own version of the stats job, but running two jobs should make it run even better and even though baselines aren’t used, why not collect them, just in case?
4. Set the EM13c to autostart, but set the listener to stay down.
The Oracle Management Service, (OMS) shouldn’t require the listener to connect to the OMR when starting, after all.
5. If there is a lot of garbage collection, just add more memory to the java heap.
If we give it more memory, then it will have less to clean up, right? More is better and there isn’t any way to find out what it should be set to anyway.
6. If you want to use the AWR Warehouse, you should use the OMR database for the AWR repository, too.
It shouldn’t make a difference to network traffic, datapump loading or resource workloads if they share a box. These two databases should work flawlessly on the same hardware, not to worry about network traffic, etc.
7. If you have a lot of backlog for job processing on your EM13c, you should trim down the worker threads.
Serializing jobs always speeds up the loading of data.
8. Sizing an Enterprise Manager EM13c is a simple mathematical process, which I’ve displayed below:
(If I didn’t mention it earlier, there will be a quiz at the end of this post…)
9. Never apply patches to the Enterprise Manager tiers or agents.
Each release is pristine and bugs don’t exist. It will only require more work in the way of applying these patches and downtime to your EM13c environment.
10. Patch any host, database or agent monitored by the Enterprise Manager manually.
Patch plans and automation of patching and provisioning is a terrible idea and the only way a DBA can assure if something is done right is if they do it manually themselves. Who needs a good night’s sleep anyway?