Everyone can contribute! Learn DevOps and Cloud Native in our cafe ☕
Technology is moving fast in the DevOps and Cloud Native community.
Join the conversation and add your thoughts, tips, experiences, stories.
"Everyone Can Contribute" is inspired by GitLab's mission.
52. #EveryoneCanContribute Cafe: Learned at KubeCon EU, feat. Cilium Tetragon first try
At first, Michael shared the insights from eBPF day, and highlighted Tetragon now being open source. Niclas mentioned that they use Cilium in production.
Isovalent open-sourced Tetragon as a new Cilium component that enables real-time, eBPF security observability and runtime enforcement. Recommend watching the eBPF day keynote at KubeCon EU, where Thomas Graf also explains the basics and future of eBPF in Cloud Native.
Spontaneous let’s try Tetragon
From talking about Tetragon, it was not far of using the Civo Kubernetes cluster already running and deploy Tetragon.
civo kubernetes create ecc-kubeconeu
civo kubernetes config ecc-kubeconeu --save
kubectl config use-context ecc-kubeconeu
kubectl get node
helm repo add cilium https://helm.cilium.io
helm repo update
helm install tetragon cilium/tetragon -n kube-system
kubectl rollout status -n kube-system ds/tetragon -w
kubectl create -f https://raw.githubusercontent.com/cilium/cilium/v1.11/examples/minikube/http-sw-app.yaml
After inspecting the raw JSON logs, Michael used
jq for better formatting. We continued using the tetragon-cli binary on macOS to observe and filter more event types.
kubectl logs -n kube-system ds/tetragon -c export-stdout -f | jq
tar xzf tetragon-darwin-amd64.tar.gz
chmod +x tetragon
kubectl logs -n kube-system ds/tetragon -c export-stdout -f | ./tetragon observe
We wondered about a Homebrew formula for the CLI, feature proposal here.
One way to see something is to execute a command inside a container.
kubectl exec -ti tiefighter -- /bin/bash
# figure out which distribution and package manager
apk add curl
Next to the default demo cases, Michael deployed and showed his KubeCon EU demo, a C++ application which leads memory when DNS resolution fails. Together with kube-prometheus we inspected the Prometheus graph interface, querying for
git clone https://gitlab.com/everyonecancontribute/observability/cpp-dns-leaker.git && cd cpp-dns-leaker
kubectl apply -f https://gitlab.com/everyonecancontribute/observability/cpp-dns-leaker/-/raw/main/manifests/cpp-dns-leaker-service.yml
kubectl logs -f deployment.apps/cpp-dns-leaker-service-o11y
We’ve talked shortly about the application’s code, endless loop, allocating !MB memory, and freeing after operations. The DNS handle error function continues on error, but does not free the buffer. This creates the memory leak to observe.
We simulated chaos by scaling the Core DNS replicas to zero in the running Kubernetes cluster. Alternatively, deploy Chaos Mesh and inject DNS failures.
kubectl scale --replicas 0 deploy/coredns -n kube-system
kubectl scale --replicas 2 deploy/coredns -n kube-system
TCP connection observability was next.
kubectl apply -f https://raw.githubusercontent.com/cilium/tetragon/main/crds/examples/tcp-connect.yaml
We could not make the file handle demo work, probably a kernel specific limitation in Civo. We will research async.
kubectl apply -f https://raw.githubusercontent.com/cilium/tetragon/main/crds/examples/sys_write_follow_fd_prefix.yaml
kubectl logs -n kube-system ds/tetragon -c export-stdout -f | ./tetragon observe --namespace default --pod xwing
kubectl exec -it xwing -- /bin/bash
Even though, impressive first demo with Tetragon. We will continue to evaluate demo cases and how it fits in production observability, maybe at Kubernetes Community Days Berlin in 2 weeks where Michael is giving a talk.
Learned at KubeCon: More Updates
- CloudNative Nordics summary video
- My Cloud Native Developer Diary: KubeCon EU by Edidiong Asikpo
- Daniel Bryant: My top five takeaways from #KubeCon Twitter thread
- LitmusChaos at KubeCon EU 2022
- How what we learned at KubeCon EU 2022 will impact our product roadmaps
OpenTelemetry announced GA for metrics at KubeCon EU, which means that the APIs are stable, and we can look into the collector, auto-instrumentation, and much more. A deep dive into OpenTelemetry metrics touches on the getting started questions, provides the architecture, tools/frameworks to use, and much more. Fantastic article!
The KubeCon EU community vote in TAG Observability is very interesting: Add profiling as OpenTelemetry supported event type
Jaeger Tracing can now accept the OpenTelemetry protocol directly, allowing trace data sent directly: “With this new capability, it is no longer necessary … to run the OpenTelemetry Collector in front of the Jaeger backend.”
Bumblebee also brings in a new perspective, helping to build, run and distribute eBPF programs using OCI images. Another great example is Parca for Profiling: at eBPF day at KubeCon EU, the change from C to Rust for more programming safety was a super interesting talk.
Niclas mentioned the [service mesh vs. eBPF](topic and blog post)
Michael shared insights into the developer learning Observability talk story at KubeCon EU (slides).
In between, Michael shared the background story about the DevOps Twins, and how https://devops-twins.com/ came to life. Or why GitLab DevRels wear the same shoes - more KubEcon stories in Michael’s blog post.
The next meetup happens on July 12, 2022.
We will meet on the second Tuesday at 9am PT.