The OEM 12c agent installation process includes automatic target discovery for anything oracular on the host. It’s a great feature. Those targets and their relationship to the agent are recorded in the repository database in sysman tables EM_MANAGEABLE_ENTITIES and EM_MANAGEABLE_ENTITIES_HIST tables.
You know better than to edit these tables directly but ‘ll warn you anyway: Never edit sysman data directly. Always use the console or EM CLI.
Anyway, the results of auto-discovery can be promoted (to become managed targets in OEM), deleted from the repository, or put on the Ignore List via the Setup | Add Target | Auto-Discovery Results page on the Agent-based Targets tab.
The Ignore List removes the targets from subsequent discovery and OEM monitoring and metrics collection. Target limbo.
Here’s how it can bite you:
We discovered a cluster running RAC one-node. In addition to the HA services, the Oracle database, and all the Oracle homes we discovered an instance named orcl_1. OEM recorded the name of this target as orcl_oracl_1. We put orcl_orcl_1 on the Ignore List and promptly forgot about it.
We’re running tests to determine whether to purchase Lifecycle Management Pack to help us with (among other things) patching clusters. It didn’t work at first.
OEM was trying to blackout all targets associated with the cluster – including instance orcl_orcl_1 from the Ignore List. We hadn’t promoted it, we weren’t monitoring it, but it still exists as a target known to the agent because the target exists in the SYSMAN schema. OEM stumbled on an unconfigured target that couldn’t be handled by the logic in release 188.8.131.52.
LMP patching completed without errors after we deleted orcl_orcl_1, instead of parking it in the Ignore List. Now OEM had one instance to blackout instead of two.
Bottom Line: Delete unmanaged targets and stick them in the Ignore List.
Target auto-discovery in OEM 12c is trivial, so keep it clean kids!