Many Pixie contributors and maintainers are at KubeCon EU this week. Stop by our Project Pavilion booth and say hi! 👋

Want to add canary analysis 🐤 to your release process, but don't want to instrument your app? Check out this lightning talk showing @argoproj rollouts canary analysis with open source app metrics (HTTP error rate, latency) magically traced by Pixie. ✨ youtu.be/OHU7dGcRZvg

Have you ever wanted to observe the #gRPC messages in your #K8s app? gRPC can be hard to sniff because of HTTP/2 header compression. Luckily, #eBPF can help! 🐝 See how @orishuss and @AvivZohari add gRPC-C tracing to Pixie using eBPF uprobes. 📝 blog.px.dev/grpc-c-tracing/

Filtering #O11y data using Pod labels can help you debug your #K8s apps more quickly. Big thanks to @TharinduWije90 for this contribution to Pixie. 👏 Check out how we use Pixie to inspect the request & response bodies for all HTTP errors to pods labeled `environment=dev` 👇

How to support replay testing in more complex scenarios? Check out these resources: engineering.fb.com/2021/10/2… blog.twitter.com/engineering…

After testing for breaking API changes, let's see how performance is impacted with the latest set of changes. Uh oh 😱 Good thing we caught this in CI/CD and not in prod!

Specify your replay tests as a GitHub Action, which will automatically replay prod traffic as you test. In the case of this demo, it will rerun the curl commands we generated from Pixie data.

It's a full house!

Some great info about Pixie's protocol tracer. Because it uses eBPF, Pixie can trace both encrypted and unencrypted traffic, regardless of the language the code is written in! Unencrypted traffic is traced directly with kprobes. Encrypted traffic is traced with uprobes. ✨

Traffic replay allows you to tap into production traffic and re-use it later in your CI/CD systems. By taking regular snapshots, you can run your tests with realistic, up-to-date traffic no matter what is currently happening in prod. 🔥