kubernetes /

Kubernetes

Live map
View repository
Snapshotmaster · a231bf3

A control-plane loop where every component observes API objects and steadily reconciles the cluster toward the user's declaration.

Click a building or route
Choose a path

A declaration becomes child Pods, a scheduling decision, and finally CRI container operations on a Node.

System map8 components · 10 connections
130%
Kubernetes: desired state becomes running containers architectureBegin with an API write. The control plane stores and enriches desired state; node agents then turn an assigned Pod into runtime and network actions.HTTP request carrying …resource key, serializ…resource events and re…ReplicaSets, Pods, sta…PodSpec without nodeNa…Pod-to-Node binding an…PodSpec, Secrets, Conf…sandbox config, image …API clientAPI serveretcd storageControllermanagerSchedulerKubeletContainerruntimekube-proxy
Connections
BuildingsEntryServiceComputeStorageRuntimeToolingExternal

Terms, in plain English

Desired stateThe configuration a user declares, such as three replicas of a Pod template.

How to study this repo

  1. 1
    Trace an API write

    Follow client-go request construction into API server setup and the etcd storage factory.

  2. 2
    Learn reconciliation

    Read the Deployment controller and identify its watch, queue, sync function, and API writes.

  3. 3
    Cross the node boundary

    Trace an assigned Pod through kubelet into the CRI runtime manager, then find where status returns.