Friday, March 16, 2018

Let’s talk about #opsmgr Sizing, baby

The OpsMgr 2012 Sizing Helper is an interactive document designed to assist you with planning & sizing deployments of System Center 2012 Operations Manager. It helps you plan the correct amount of infrastructure needed for a new OpsMgr 2012 deployment, removing the uncertainties in making IT hardware purchases and optimizes cost. A typical recommendation will include minimum hardware specification for each server role, topology diagram and storage requirement.

[source]

IMG_20180316_151010_BokehOne of the issues I have encountered most frequently in the years I have supported customers running System Center Operations Manager is poor performance of their environment because it is under-resourced and over-utilized. And when we have the discussion about resources, I am often told that they used the Sizing Helper tool, so it must be right.

The Sizing Helper really is just that, a helper, a guide. A starting point. It is a guide on the basic hardware to provision for the number of agents you intend on monitoring, without any consideration for additional workloads, number of management packs or environment health. And this is something that is so environment specific, that the calculator will never be able to cater for each scenario.


The Management Pack Lifecycle recommends that you review each management pack you want to import in a test environment before you implement into production exactly for this reason. Your process should look like this:

  1. Download new management pack (or updated version of existing management pack) as well as the management pack documentation.
  2. Read the management pack documentation.
  3. Read it again, this time taking note of the requirements for this management pack, such as run-as account requirements, additional resource requirements, configurations to be made on the target, etc.
  4. Baseline the performance and capacity of your test environment.
  5. Take note of the current workflows on the management servers. You can use Dirk Brinkman’s script to help, if unsure.
  6. Import the management pack into your test environment and configure as per the management pack documentation.
  7. Compare the performance and workflows of the environment before and after import. You will also need to take into consideration how many agents you are monitoring in your test environment vs in production, and identify the load per agent/monitored entity.

I would also recommend you run the management pack in your test environment for at least a two week period, and perform stress testing on the monitored entities. Generate failures, generate event storms, put it through its paces as much as possible to ensure you can gauge, as much as possible, what the impact on your production environment will be.

You will also be able to tune the noise out quite quickly this way, and ensure that, when you import the management pack into production, you are only receiving the alerts you want, you are not over-collecting performance counters you will never use, and you can keep your SCOM environment healthy.

It may sound like a long process, but it will save you a lot of time in the long run.

No comments:

Related Posts with Thumbnails