You can most definitely have two OEM agents on the same host...uploading to different OMS's. The most important thing is that they need to be configured using unique ports, e.g. one using the default 3872 and another using 3874.
What are the requirements of the "hoster" though? I would guess they're only using OEM to monitor at an infrastructure level and if so, just ensure this is configured under an operating system account that won't have access to anything besides the inventory.
Hope this helps
Yes I was thinking that while pushing the agent from the New OEM, they can specified another directory.
Using a different port, I would have pass this one ...
What they want to do ?
I guess they want to monitor OS critical points such as Storage / CPU / Memory / SWAP.
and DB ones : locks, audits, long SQLs, tablespace issues, database status etc ...
Well we will see how to test that on UAT.
Thanks for the answer , have a good day
just use a different installation and a different set of ports for the 2nd OEM agent.
Thanks Markus, we will move with this solution.
crossing the fingers , that it won't mess
Yes, very easily done with (as mentioned by other posters) different agent ports, and this (largely) operates without issue.
I've come across some sites with different support providers, necessitating that 2 agents be installed and currently have this setup in my present site.
It's not ideal, but sometimes necessary. Ideally you have 1 EM and create other users and have different groups depending on whom is accessing EM, but in the event that you want 2x OEMs / 2x agents per node, this does work.
Best to do with a response file too, so that you can have 2x custom homes with the agent password, agent home, agent instance. Just the same as DB binaries, you can install as many homes from the agent files.
One thing to look out for -
Everything will be submitted twice against each DB, depending on how you have notifications, alerts and templates setup.
E.g. some collection schedules will execute jobs to collect data at largely the same interval. You've 2x EMs monitoring the same dictionary tables / submitting jobs from 2x EM rather than one.
As such, some bugs (like v$backup piece queries with EM13c against legacy 11G RDBMS) can get executed twice, doubling the pain :-/
Same with troubleshooting, you can see the two agent JVMs hitting CPU quite hard if two agents are simultaneously trying to generate AWR or ADDM reports, one for the local DBA and one for the hosting provider.