Profiling

Requirements

Prerequisites here
eBPF in this context only runs on Kubernetes with modern Linux kernel based containers.
Note that eBPF integration w/ profiling is currently in a beta state and is subject to change.

Official Documentation here

Please read the official documentation thoroughly before proceeding.

Instructions

This workshop is designed to add the eBPF agent to an existing installation of the Coralogix OpenTelemetry helm chart and then deploy simple applications in Java to demonstrate the spans and APM / profiling capabilities available with eBPF.

Install the OpenTelemetry Collector on your k8s cluster

Easy Coralogix instructions for Complete Observability are here

Clone Repo

git clone https://github.com/coralogix/workshops

Change to Profiling Directory

cd workshops/workshops/otel/profiling

Step 1 - Override the OpenTelemetry Collector with Continuous Profiling

Edit override-otel.yaml and update the globals for your domain and cluster name then:

source 1-deploy-override.sh

Step 2 - Deploy Example Applications

We will reuse a zero-code instrumentaiton example from the autoinjection workshop.

This application has a client and server in the deployment, and the server has random slowdowns to study with profiling.
Consult the README (also found here workshops/workshops/otel/autogenerators/java/microservices/README.md) for more details on the architecture and frameworks used.

cd workshops/workshops/otel/autogenerators/java/microservices
source deploy-java-autogen.sh

Study The Results

Coralogix->APM shows the working service catalog.

Click on the cx-java-server service and select Profiles to see the profiles. Consult the official documentation for more details.

Step 3 - Cleanup

Delete the app examples:

source delete-java-autogen.sh

Return to the profiling workshop and roll back the otel collector

cd workshops/workshops/otel/profiling
source 2-rollback-otel-collector.sh