cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

This product reached the end of support date on March 31, 2021.

Calculation Rate - What am I doing wrong?

vlu
Guide

Hi all, I appreciate any help. Getting very frustrated.

So I have two measures (1) web request count split by jvm (2) web request count split by application.

The calculation rate is web request count split by jvm / web request count split by application, I should get a percentage less than 100%. But I'm getting 100% across the board.

Thoughts?

7 REPLIES 7

richard_boyd22
Inactive

Can you describe in plain words what you're trying to achieve?  I don't think what you're attempting will work but we may be able to figure out another way to get the data you need.

vlu
Guide

Hi Richard,

This is an easy way to find out load distribution across JVMs. DT allows to create web request count measures for Application and Agent.

Example:

  1. Web request count for Application XYZ = 100
  2. Web request count for agent1 = 40
  3. Web request count for agent2 = 25
  4. Web request count for agent3 = 10
  5. Web request count for agent4 = 25

Based on the above example, agent1 is taking 40% (40/100) of the requests; whereas agent3 is only taking 10% (10/100). This is no secret sauce to calculate load distribution.

What I'm trying to accomplish is web request count for agents / web request count for application. See attached screen shot.

2015-04-14_12-28-24.png

 

Thanks.

What you will have to do instead is have one measure which splits on both, then create a chart.  Add the measure to the chart, choosing Agent in the dropdown when you add it.  Then in the chart controls (upper right), click on the 4th button for the chart splitting and change this to Application.  Then you can change the chart type to stacked percentage bar chart to see the trend over time or to pie chart to see the summary data.

Hope that helps,

Rick B

Boyd,

Great minds think alike. Yes I do have another measure that splits on both and charting the way you mentioned. Unfortunately, this has limited capability. Using the above example, in a perfect world, each JVM should be evenly loaded 25%. What I'm attempting to do is to trigger an incident if an agent consistently taking more than 50% or less than 5%.

dave_mauney
Dynatrace Champion
Dynatrace Champion

Has anyone used the "Tier Subpaths" measures for tracking load balancing?  I have used the count measures split by agent with meter charts in the past, and it seems to work well.

David, how did you address the peak hour vs off hour count? More importantly, did you create an incident to trigger alert?

dave_mauney
Dynatrace Champion
Dynatrace Champion

Hi Vinh,

I was only going for a visualization where I could easily see if any server was taking more or less load during a given time period, so I did not address the peak hour or incident issues you mentioned.

In the past, with a different APM tool, I created alerts based on the percentage of the cluster being used per process.  This involved creating a table of tiers, assigning each tier to a cluster, etc.  Using percentages should handle both peak and non-peak.  Maybe you could accomplish the percentage calculation by using an Agent split over an Agent Group split as a rate measure?  I have not tried this, but it seems like it might be feasible.

HTH,

dave