Key Requests

[BrunoBaptista]

Hi everyone,

I hope you're all doing well.

I came across something that raised a question for me.

I ran some tests while creating a Dashboard for a few endpoints. Since I was using metrics rather than spans, I noticed that some endpoints seemed to use timeseries {... in their queries.

Why do some endpoints appear without the need to be marked as Key Requests, while others require it?

The only difference I was able to identify between the ones that appear and the ones that don't is that the endpoints that already appear in the Dashboard do not go through Kubernetes, while the ones that don't appear do go through Kubernetes.

Does anyone know why this happens or have any insight into this?

I've attached two examples for reference.

[sujit_k_singh]

Hi @BrunoBaptista 

This comes down to how Dynatrace detects the service, not whether it's on Kubernetes. Dynatrace has two ways of building the per-endpoint rows that feed timeseries/dashboards:

SDv2 → all endpoints show up automatically, no marking needed.
Classic SDv1 without Enhanced endpoints → only endpoints manually marked as Key Requests get their own timeseries data. Everything else falls into an aggregate bucket with no per-endpoint metrics.

So the split you're seeing (Pix.Recorrente.Api showing up fine vs. spi-liq-api needing Key Requests) is most likely because those two services are being monitored under different detection modes.

Thanks,

Sujit

[Fin_Ubels]

You can also enable automatic endpoint detection for SDV1. This allows you to start moving away from manually marking key requests and reaping the benefits of being able to granularly measure all of your endpoints. You can find more details in the docs here for SDV1 Enable endpoint metrics for SDv1 services — Dynatrace Docs and here for SDV2 Customize endpoint detection in Service Detection v2 — Dynatrace Docs

[sujit_k_singh]

Hi @Fin_Ubels 

That's a great point!

[BrunoBaptista]

Thank you for your response.

In a dev environment, I enabled SDv1. However, I’m now having trouble finding the response time metric.

For example, I assume it is different from the one I used previously, which was something like service.keyRequest.response.client.

According to the documentation, it should be dt.service.request.response_time. However, when I search for it in Data Explorer or in a DQL query, I can’t find the same metric.

Would you know why this might be happening?

[Julius_Loman]

@BrunoBaptista the metrics are standard service metrics such as dt.service.request.response_time , look for the endpoint.name dimension - it's the request name.

Documentation: https://docs.dynatrace.com/docs/shortlink/built-in-metrics-on-grail#services 

[BrunoBaptista]

Thank you again for your response, and sorry for the question.

However, I searched for the metrics mentioned in the documentation, and my endpoints are still not appearing.

Using the builtin:service.response.server metric, I can find my service, but it does not allow me to filter by endpoint, which is what I need.

I have not been able to find a metric that shows the response time for a single endpoint.

I'm not sure if I have done something wrong or if there is another issue involved.

I configured SDV1 for only one specific Kubernetes cluster for testing purposes. Please find attached a screenshot of the configuration.

My Dynatrace environment is also running version 1.45.0.

Thank you again for your help.

[Julius_Loman]

@BrunoBaptista you are referring to builtin:service.response.server which is a classic metric. Those endpoint metrics exist only in Grail. Thus you need to use dt.service.request.response_time in Notebook or Dashboard app (or any other app using Grail metrics).

It won't be available in Data Explorer or classic dashboards.