We are planning to use the default MM ( by changing the Agent expression) in CA APM 10.7 in all the App servers in our environment as a starting point of monitoring.I know that the default MM consists of alerts and metric groupings specific to dotnet, java, docker, spring boot, struts, expressjs etc. I also know that if i deploy this MM in java environment then only java related alerts and metric groupings will be populated and rest of the things will not populated.
I just want to know how it is going to impact my MoM, collector and agent servers. Is this approach will add more resource overhead on my APM clusters. It may not be a best practice approach but i need a confirmation that this approach will not make my APM cluster unstable.
A Management Module (MM) is just a way to manage and provide a better visual interpretation of the metric data already captured so there is no additional impact on agent servers that have already sent the data to the Collectors.
On a cluster only the MOM is processing the MM (MMs should not be deployed on Collector).
The impact on Collector is MOM retrieving the data from it to show in any MM metric grouping, dashboard etc., so similar to retrieving data via metric browser etc.
There will be some additional load on MOM and I would suggest you introduce your agent regex changes on a gradual basis to avoid initial large metric groupings and then see how you go. Also if you start to introduce calculators that will also increase the workload.
This subject is also related to sizing MOM correctly. Some relevant links here:
CA APM Sizing Recommendations - CA Application Performance Management - 10.5 - CA Technologies Documentation
How to perform an APM Cluster Performance Health C - CA Knowledge
What do the fields in Perflog.txt mean? - CA Knowledge
APM Health Tip - How to improve the performance of - CA Knowledge
What is the work being done by the harvesting proc - CA Knowledge
Hope that helps.
It would also be interesting to hear of any real world experiences/advice that other Community users can provided.
If you have not created any custom Management Module for the environment, the default Management Module is definitely a good starting point. By default, it should include all agents within the domain that it is associated to, using the Management Module Agent Expression (.*)\|(.*)\|(.*)
All Management Modules should only be deployed on the MOM, and not the Collectors, so it shouldn't impact performance on Collector or the application servers (agents).
In terms of overhead, I believe it depends on the scale/load of the monitoring, but it shouldn't be significant, as the Default Management Module only comes with Alerts, Dashboards and Metric Groupings, it does not contain any resource-intensive calculators. Having said that, the recommendation is to always implement it in a UAT environment that mimics the load of a Production server, assess and tune the settings accordingly. Elements you can fine tune in the Default Management Module are for example the alert notification intensity and the scope of the metric groupings.
Hope this helps.
using a Template MM to quickly generate dashboard and alerts is a best practice in APM 9.x and previous versions approach. In this way you can deploy these objects reducing the time consuming that is typical of MM implementation.
I’m envolved in a huge APM project (3700 Java Agents), started 10 years ago, so now we worked a lot with MM templates (and we are still working now with these objects, but we are moving to the new Team center approach).
We built one template for each kind of AS and we deployed a huge number of MM, using copy and paste technique and modifying the Agent expressions defined in the “root” of MM.
Under the MOM load point of view my suggestion is that you have to limit the number of calculators and you have to use, in the MG, narrow metric expressions.
In my experience I have to say that moving toward the APM new versions (we started with 8.x and now we are on 10.5.2) the behavior of APM infrastructure regard the management of a large number of MM objects has improved (you can see this improvement when you save a new object: with the older version we had few seconds with no data on dashboard, now the MOM, starting with 10.5.x, responds very quickly, without HW changes on the APM infrastructure).
Thanks a ton for your Insight. I will be grateful to you if you can share MM template of WebLogic and WebSphere with me so that I will be able test those in my UAT environment and built on top that if required.
I am not sure whether you have seen my previous reply. I will be grateful to you if you can share MM template of WebLogic and WebSphere with me so that I will be able test those in my UAT environment and built on top that if required.
sorry for the delay. In this period I have no access to APM infrastructures, so I have to search something in my archives.
I'll keep you informed.