This document was translated by GPT-4

1. Introduction

This is an example of a microservice application developed using Spring Boot to illustrate the AutoTracing capability of DeepFlow.

2. Deploying the Spring Boot Demo

The demo we used comes from this GitHub repository (opens new window). It has a simple distribution tracing chain: foo_svc -> bar_svc -> loo_svc.

And this is the command to quickly deploy the Demo in K8s:

kubectl apply -f https://raw.githubusercontent.com/deepflowio/deepflow-demo/main/DeepFlow-EBPF-Sping-Demo/deepflow-ebpf-spring-demo.yaml

The original demo in the GitHub repository used Jaeger for primary tracing. In order to demonstrate AutoTracing capabilities, we removed Jaeger from the deployment script mentioned above.

3. Exploring Distributed Tracing

You can proceed to Grafana and open the Distributed Tracing Dashboard by selecting namespace = deepflow-ebpf-spring-demo. Then, choose an invocation to trace. Here's how it looks:

eBPF Sping Demo

The tracing data in DeepFlow includes three types of Spans that track the entire trajectory of a request:

• N: Spans extracted from network traffic using BPF
• S: Spans extracted from system or application function calls using eBPF
• A: Spans collected internally from the application using OTel

The image above shows the first two types. The third one can be displayed upon integrating OpenTelemetry.

You can also view the effect of tracing by visiting DeepFlow Online Demo (opens new window). The corresponding topology diagram of the call chain flame graph in the image above is as follows:

To summarize this Demo,

• Zero code injection: No manual code injection is required for the entire tracing process, nor is there a need for any TraceID/SpanID injection into the HTTP Header.
• Multilingual: Supports tracing of Java applications and C (curl) basic services.
• Full-link: Utilizes eBPF and BPF to automatically trace 18 Spans in this Trace, including 6 eBPF Spans and 12 BPF Spans.
• Full stack: Supports tracing of network paths between two Pods on different K8s Nodes, even if the path goes through tunnel encapsulations, such as Span 2-5 (IPIP tunnel encapsulation).