Skip to main content

OpenTelemetry

info

This lab requires a user with the platform-administrator role!

In order to make a system observable, it must be instrumented. Language specific implementation of OpenTelemetry provides a way to instrument your application without touching your source code. In this lab we're going to instrument a Java application.

Prerequisites​

Make sure the following Apps are enabled on the platform (requires a user with the platform-administrator role):

  • Tempo on Platform
  • Otel on Platform
  • Loki on Platform
  • Grafana for the Team

Enable tracing for Istio and ingress-nginx-platform:

  1. Go to Apps.

  2. Click on Istio.

  3. Click on the configuration icon.

  4. Click on the Values tab.

  5. In the values click Edit and set the following values:

tracing:
  enabled: true
  samplingRatio: "100"
note

Setting the samplingRatio to 100 is only done for this lab to make sure all traces are sampled. In a production environment the ratio would normally be set to 0.01.

  1. Click Submit.

  2. Repeat step 2 to 6 for the ingress-nginx-platform app.

note

Sometimes instrumenting the nginx-ingress controller requires a restart of the pods. Restart the ingress-nginx-platform-controller Pod if you don't see any traces from nginx.

Build an image from source code​

For this lab we're going to use the Spring PetClinic Sample Application. Follow these steps to build an image from source code.

Add the code repository​

  1. In the left menu in the Console, click on Code Repositories.

  2. Click on ADD CODE REPOSITORY.

  3. Provide a name for this code repository. The name will be used to identify the code repository in App Platform. You can only use the name once within a Team. We are going to use the name petclinic for this code repository.

  4. Select GitHub for the Git service.

  5. Fill in the following Repository URL: https://github.com/spring-projects/spring-petclinic.git

  6. Click Add Code repository

Create the container image​

  1. In the left menu of the Console, click on Container Images.

  2. Click on CREATE CONTAINER IMAGE.

  3. Select the BuildPacks build task.

  4. Select the petclinic repository from the Repository drop-down list.

  5. Select the Reference main.

  6. Optionally you can change the Image name and tag. This is the name and tag used to store the image in the private Harbor image registry.

  7. Click Create Container Image.

Create a workload from the developer catalog​

  1. Go to Container Images in the left menu and click on the Add to clipboard of the petclinc image in the Repository column. Remember that the tag is main.

  2. Go to Workloads in the left menu and click on Create Workload.

  3. Select Quickstart-K8s-Deployment-Otel from the catalog.

  4. Add the Name petclinic for the workload.

  5. Leave the Auto image updater to Disabled.

  6. In the workload values, change the following parameters:

image:
  repository: <paste from clipboard>
  tag: main

Because we're deploying a Java application here, maybe also change the resources:

resources:
  limits:
    cpu: 500m
    memory: 1Gi
  requests:
    cpu: 300m
    memory: 256Mi
  1. Click Submit.

Expose the service​

  1. In the left menu click Services then click on CREATE SERVICE.

  2. Select the petclinic service from the Service Name drop-down list:

  3. Click Create Service.

See traces​

To be able to see traces, we'll first need to generate some requests. Click on the URL of the petclinic service and generate some requests.

To see traces, you'll first need to find a TraceID of a trace. Go to Apps in the left menu and then click op Loki. Select the label App and select value petclinic.

Click on a log entry of a request. Note that the requests are logged by the Istio Envoy proxy. You will now see a link to the full trace in Grafana Tempo. Click on it.

note

If you don't see any traces, then check and see if the pod runs the ghcr.io/open-telemetry/opentelemetry-operator/autoinstrumentation-java:1.26.0 container. Sometimes the pod starts before the Instrumentation resource has been created. If this is the case, restart the Pod.

Team apps

And when you click on the Node graph, you’ll see the complete flow:

Team apps