Re: OMS and Repository Preferences
Date: Thu, 12 Apr 2012 08:48:28 -0600
Cloud (and Grid) Control are designed to use any number of independent OMSs against one OMR (repository).
As a rule, I always set up one OMS on the same machine as the OMR. It provides me access whether physical networking is up or not. If the customer wants/needs multiple OMS, then I will install the required number on other machines, turn off (emctl stop oms) on the repository machine for use only in troubleshooting or in emergency as a console, and not publish that URL.
I also tend to have specific machine(s) dedicated to OMS (count based on load) which are connected to the agent and do not (normally) have hooman console interaction. Again, the URL is not published to admins. I separate the console load from the agent connectivity machines because I want smaller machines with slightly more predictable load behaviour. It also allows me to mix n match OSs.
In larger shops, I have other OMSs set up specifically for em console. Since these run using ADF, the memory and CPU stress can be significant, especially in the older 10g and 11g variants, and especially when heating up the servlets for display. I monitor the memory for those OMSs carefully and add new OMS no separate machines based on use and number of admins who need to access the environment.
I have been working very hard to increase the number of admins from just DBAs, as the OEM core is really designed to be used as an SNMP management center, not just a DBA pretty picture place. With multiple OMSs, I can end the storage adn OS guys to one OMS, the prod DBAs to a second OMS and the dev/test DBA to a third OMS, and turn on/off access as needed. There is no restriction mechanism, other than which URL I send to whom, but when they realize that going to a different console than the one I assign will take extra milli-bleems to make the image available, because it's not normally served from that OMS, they tend to stay where they are assigned.
On 12/04/2012 8:15 AM, Kellyn Pot'vin wrote:
> I know it's common for many folks to separate their OMS from their repository for Enterprise Manager- I haven't been one to do this in the past due to, OK, just going to say it, flaky network connectivity in previous environments that would cause *false* communication alerts with the OMS components to/from the repository. I now have new clients, new systems to work on and at least two EM12c environments to build out in the near future.
> So here's my question:
> What is your preference when building an Enterprise Manager environment? Do you house the OMS and the repository on the same host or separately and I want to know WHY you build it one way or the other, not just "It just should be that way.. " :)
> Kellyn Pot'Vin
> Senior Technical Consultant