How oscarlucas22/demo-pi works
Project Type and Purpose
• This is a Kubernetes–centric tutorial/documentation repo for installing and using Linkerd as a Service Mesh, plus observability components (Prometheus, Grafana) and a demo app (Emojivoto).
• The “imagenes” folder holds screenshots; there’s no application source code here.
Key Components
a. Kubernetes Cluster
• Namespaces:
– linkerd (control plane & extensions)
– emojivoto (demo app)
• kube-apiserver, etcd, kubelet, controller-manager (abstracted as “K8s control plane”).
b. Linkerd Service Mesh
• Control Plane (in linkerd ns):
– linkerd-controller (linkerd-web, identity, destination, proxy-injector, etc.)
– linkerd-viz extension (metrics-collector, dashboard)
• Data Plane (injected sidecar proxies next to each pod)
c. Demo Application (in emojivoto ns)
– web-svc → web pods (HTTP)
– vote-bot, vote-front, emoji-svc, results-svc (gRPC/HTTP flows)
d. Observability Stack
– Prometheus (scrapes metrics from Linkerd and app proxies)
– Grafana (visualizes Prometheus metrics)
e. User & CLI
– Developer’s local machine with linkerd CLI & kubectl
– Browser accessing dashboards and emojivoto UI
Relationships & Data Flows
• Developer runs linkerd install → CLI → applies Control Plane MANIFESTS to cluster.
• K8s Control Plane schedules Control Plane & extension pods.
• Developer injects Linkerd proxies into Emojivoto deployments → data-plane proxies wrap each app container.
• User HTTP/gRPC traffic to Emojivoto:
Browser → Kubernetes Ingress/Service → Linkerd inbound proxy → app container → outbound proxy → downstream pods.
• Linkerd proxies emit mTLS-encrypted traffic and telemetry.
• Prometheus scrapes /metrics endpoints on Linkerd control plane and sidecar proxies.
• Grafana queries Prometheus API for dashboards.
• CLI linkerd check commands validate both planes.
Architectural Patterns & Principles
• Sidecar Proxy Pattern: Linkerd injects a proxy container per pod (data plane).
• Control-Data Plane Separation: Control plane components are centralized; data-plane logic lives alongside app pods.
• Namespace Isolation: Separate namespaces for mesh control, observability, and applications.
• Git-Ops / Declarative Infrastructure: Use of kubectl apply on generated YAML manifests.
Final Recommendations
• Be as granular as possible: each namespace, each component, each network flow.
• Use consistent iconography (Kubernetes logo for cluster, Linkerd logo for mesh).