Is there a way to force APM to calculate Differential metrics for a particular metric? i.e. I wish to see differential analysis metrics for some metrics on a EPAgent.
I could do this in previous versions of APM with heuristics, copying metrics to Frontends/Apps.
Yes you definitely can as long as data being used is suitable for differential analysis.
Simply create a Metric Grouping for required metrics, create a new differential control to use this grouping and define a alert on top of it.
See https://docops.ca.com/ca-apm/10/en/using/configure-and-monitor-differential-analysis#ConfigureandMonitorDifferentialAnalysis-HowtoConfigureDifferentialAnalysis for more details.
I had the same idea about the EPA vmstat/netstat and the backends.
So I am curious about what do you mean by suitable or non-suitable?
Also have you seen somewhere some a notes about sizing guidelines for the collectors specifically in relations to the amount of metrics that are being "differentially analyzed"?
Take a look at this excellent blog to understand DA better and also my comment about suitability of metric for DA:
Differential Analysis Revealed
There is a clamp on number of metrics that are analyzed by DA and you can increase\decrease as needed depending on collector resources
Configure and Monitor Differential Analysis - CA Application Performance Management - 10.0 - CA Technologies Documentati…
Hope that helps.
I found this really helpful.
In the documentation examples were always using Frontends/BT metrics, so I didn't know if that was a requirement for these metrics. I needed to create a lab for this, but I'm quite short of time lately, and I know community helps
I used this on Heuristics, by copying metrics I send to EPAgents, but related to average response times for processes and ws calls that regular Agents could not analyze.
Fred, if you need any details to explore this, tell me and I can post further info.
Thanks. For now I am just playing around a little bit every day and learn so I can set interesting defaults for my companies v10.1+ upgrade. My wish is for one alert with some information about where to look So I wanted to add the hosts' vmstat. Looking at the metrics many of them are fixed (e.g. total memory) and others are probably not interesting but for the few odd cases. So for now just added only EPA vmstat|Cpu:(.*). I still need to add a couple others.
I am concerned that I am overenthusiastic about this technology and will add too much useless or miss out and that my current VMs hosting the managers v9.7 will fall over after upgrade.
KulbirNijjer I read so many of your posts about the managers that my head is spinning, what about setting up the differential analysis for the Enterprise managers themselves? (and then later show them in some "admin" appmap)?
I think there is a need for one community conversation about "best practices" and some webinars (there is too much to read:-) to learn from each other about how to get to the KPIs of an application and then set these for the analysis. The next step would be obviously how to script that process (Mike Sydor had a good piece with the ABA)