This is the multi-page printable view of this section. Click here to print.

Return to the regular view of this page.


Kubernetes is a portable, extensible, open source platform for managing containerized workloads and services, that facilitates both declarative configuration and automation. It has a large, rapidly growing ecosystem. Kubernetes services, support, and tools are widely available.

This page is an overview of Kubernetes.

Kubernetes is a portable, extensible, open source platform for managing containerized workloads and services, that facilitates both declarative configuration and automation. It has a large, rapidly growing ecosystem. Kubernetes services, support, and tools are widely available.

The name Kubernetes originates from Greek, meaning helmsman or pilot. K8s as an abbreviation results from counting the eight letters between the "K" and the "s". Google open-sourced the Kubernetes project in 2014. Kubernetes combines over 15 years of Google's experience running production workloads at scale with best-of-breed ideas and practices from the community.

Going back in time

Let's take a look at why Kubernetes is so useful by going back in time.

Deployment evolution

Traditional deployment era: Early on, organizations ran applications on physical servers. There was no way to define resource boundaries for applications in a physical server, and this caused resource allocation issues. For example, if multiple applications run on a physical server, there can be instances where one application would take up most of the resources, and as a result, the other applications would underperform. A solution for this would be to run each application on a different physical server. But this did not scale as resources were underutilized, and it was expensive for organizations to maintain many physical servers.

Virtualized deployment era: As a solution, virtualization was introduced. It allows you to run multiple Virtual Machines (VMs) on a single physical server's CPU. Virtualization allows applications to be isolated between VMs and provides a level of security as the information of one application cannot be freely accessed by another application.

Virtualization allows better utilization of resources in a physical server and allows better scalability because an application can be added or updated easily, reduces hardware costs, and much more. With virtualization you can present a set of physical resources as a cluster of disposable virtual machines.

Each VM is a full machine running all the components, including its own operating system, on top of the virtualized hardware.

Container deployment era: Containers are similar to VMs, but they have relaxed isolation properties to share the Operating System (OS) among the applications. Therefore, containers are considered lightweight. Similar to a VM, a container has its own filesystem, share of CPU, memory, process space, and more. As they are decoupled from the underlying infrastructure, they are portable across clouds and OS distributions.

Containers have become popular because they provide extra benefits, such as:

  • Agile application creation and deployment: increased ease and efficiency of container image creation compared to VM image use.
  • Continuous development, integration, and deployment: provides for reliable and frequent container image build and deployment with quick and efficient rollbacks (due to image immutability).
  • Dev and Ops separation of concerns: create application container images at build/release time rather than deployment time, thereby decoupling applications from infrastructure.
  • Observability: not only surfaces OS-level information and metrics, but also application health and other signals.
  • Environmental consistency across development, testing, and production: runs the same on a laptop as it does in the cloud.
  • Cloud and OS distribution portability: runs on Ubuntu, RHEL, CoreOS, on-premises, on major public clouds, and anywhere else.
  • Application-centric management: raises the level of abstraction from running an OS on virtual hardware to running an application on an OS using logical resources.
  • Loosely coupled, distributed, elastic, liberated micro-services: applications are broken into smaller, independent pieces and can be deployed and managed dynamically – not a monolithic stack running on one big single-purpose machine.
  • Resource isolation: predictable application performance.
  • Resource utilization: high efficiency and density.

Why you need Kubernetes and what it can do

Containers are a good way to bundle and run your applications. In a production environment, you need to manage the containers that run the applications and ensure that there is no downtime. For example, if a container goes down, another container needs to start. Wouldn't it be easier if this behavior was handled by a system?

That's how Kubernetes comes to the rescue! Kubernetes provides you with a framework to run distributed systems resiliently. It takes care of scaling and failover for your application, provides deployment patterns, and more. For example: Kubernetes can easily manage a canary deployment for your system.

Kubernetes provides you with:

  • Service discovery and load balancing Kubernetes can expose a container using the DNS name or using their own IP address. If traffic to a container is high, Kubernetes is able to load balance and distribute the network traffic so that the deployment is stable.
  • Storage orchestration Kubernetes allows you to automatically mount a storage system of your choice, such as local storages, public cloud providers, and more.
  • Automated rollouts and rollbacks You can describe the desired state for your deployed containers using Kubernetes, and it can change the actual state to the desired state at a controlled rate. For example, you can automate Kubernetes to create new containers for your deployment, remove existing containers and adopt all their resources to the new container.
  • Automatic bin packing You provide Kubernetes with a cluster of nodes that it can use to run containerized tasks. You tell Kubernetes how much CPU and memory (RAM) each container needs. Kubernetes can fit containers onto your nodes to make the best use of your resources.
  • Self-healing Kubernetes restarts containers that fail, replaces containers, kills containers that don't respond to your user-defined health check, and doesn't advertise them to clients until they are ready to serve.
  • Secret and configuration management Kubernetes lets you store and manage sensitive information, such as passwords, OAuth tokens, and SSH keys. You can deploy and update secrets and application configuration without rebuilding your container images, and without exposing secrets in your stack configuration.
  • Batch execution In addition to services, Kubernetes can manage your batch and CI workloads, replacing containers that fail, if desired.
  • Horizontal scaling Scale your application up and down with a simple command, with a UI, or automatically based on CPU usage.
  • IPv4/IPv6 dual-stack Allocation of IPv4 and IPv6 addresses to Pods and Services
  • Designed for extensibility Add features to your Kubernetes cluster without changing upstream source code.

What Kubernetes is not

Kubernetes is not a traditional, all-inclusive PaaS (Platform as a Service) system. Since Kubernetes operates at the container level rather than at the hardware level, it provides some generally applicable features common to PaaS offerings, such as deployment, scaling, load balancing, and lets users integrate their logging, monitoring, and alerting solutions. However, Kubernetes is not monolithic, and these default solutions are optional and pluggable. Kubernetes provides the building blocks for building developer platforms, but preserves user choice and flexibility where it is important.


  • Does not limit the types of applications supported. Kubernetes aims to support an extremely diverse variety of workloads, including stateless, stateful, and data-processing workloads. If an application can run in a container, it should run great on Kubernetes.
  • Does not deploy source code and does not build your application. Continuous Integration, Delivery, and Deployment (CI/CD) workflows are determined by organization cultures and preferences as well as technical requirements.
  • Does not provide application-level services, such as middleware (for example, message buses), data-processing frameworks (for example, Spark), databases (for example, MySQL), caches, nor cluster storage systems (for example, Ceph) as built-in services. Such components can run on Kubernetes, and/or can be accessed by applications running on Kubernetes through portable mechanisms, such as the Open Service Broker.
  • Does not dictate logging, monitoring, or alerting solutions. It provides some integrations as proof of concept, and mechanisms to collect and export metrics.
  • Does not provide nor mandate a configuration language/system (for example, Jsonnet). It provides a declarative API that may be targeted by arbitrary forms of declarative specifications.
  • Does not provide nor adopt any comprehensive machine configuration, maintenance, management, or self-healing systems.
  • Additionally, Kubernetes is not a mere orchestration system. In fact, it eliminates the need for orchestration. The technical definition of orchestration is execution of a defined workflow: first do A, then B, then C. In contrast, Kubernetes comprises a set of independent, composable control processes that continuously drive the current state towards the provided desired state. It shouldn't matter how you get from A to C. Centralized control is also not required. This results in a system that is easier to use and more powerful, robust, resilient, and extensible.

What's next

1 - Objects In Kubernetes

Kubernetes objects are persistent entities in the Kubernetes system. Kubernetes uses these entities to represent the state of your cluster. Learn about the Kubernetes object model and how to work with these objects.

This page explains how Kubernetes objects are represented in the Kubernetes API, and how you can express them in .yaml format.

Understanding Kubernetes objects

Kubernetes objects are persistent entities in the Kubernetes system. Kubernetes uses these entities to represent the state of your cluster. Specifically, they can describe:

  • What containerized applications are running (and on which nodes)
  • The resources available to those applications
  • The policies around how those applications behave, such as restart policies, upgrades, and fault-tolerance

A Kubernetes object is a "record of intent"--once you create the object, the Kubernetes system will constantly work to ensure that the object exists. By creating an object, you're effectively telling the Kubernetes system what you want your cluster's workload to look like; this is your cluster's desired state.

To work with Kubernetes objects—whether to create, modify, or delete them—you'll need to use the Kubernetes API. When you use the kubectl command-line interface, for example, the CLI makes the necessary Kubernetes API calls for you. You can also use the Kubernetes API directly in your own programs using one of the Client Libraries.

Object spec and status

Almost every Kubernetes object includes two nested object fields that govern the object's configuration: the object spec and the object status. For objects that have a spec, you have to set this when you create the object, providing a description of the characteristics you want the resource to have: its desired state.

The status describes the current state of the object, supplied and updated by the Kubernetes system and its components. The Kubernetes control plane continually and actively manages every object's actual state to match the desired state you supplied.

For example: in Kubernetes, a Deployment is an object that can represent an application running on your cluster. When you create the Deployment, you might set the Deployment spec to specify that you want three replicas of the application to be running. The Kubernetes system reads the Deployment spec and starts three instances of your desired application--updating the status to match your spec. If any of those instances should fail (a status change), the Kubernetes system responds to the difference between spec and status by making a correction--in this case, starting a replacement instance.

For more information on the object spec, status, and metadata, see the Kubernetes API Conventions.

Describing a Kubernetes object

When you create an object in Kubernetes, you must provide the object spec that describes its desired state, as well as some basic information about the object (such as a name). When you use the Kubernetes API to create the object (either directly or via kubectl), that API request must include that information as JSON in the request body. Most often, you provide the information to kubectl in a file known as a manifest. By convention, manifests are YAML (you could also use JSON format). Tools such as kubectl convert the information from a manifest into JSON or another supported serialization format when making the API request over HTTP.

Here's an example manifest that shows the required fields and object spec for a Kubernetes Deployment:

apiVersion: apps/v1
kind: Deployment
  name: nginx-deployment
      app: nginx
  replicas: 2 # tells deployment to run 2 pods matching the template
        app: nginx
      - name: nginx
        image: nginx:1.14.2
        - containerPort: 80

One way to create a Deployment using a manifest file like the one above is to use the kubectl apply command in the kubectl command-line interface, passing the .yaml file as an argument. Here's an example:

kubectl apply -f

The output is similar to this:

deployment.apps/nginx-deployment created

Required fields

In the manifest (YAML or JSON file) for the Kubernetes object you want to create, you'll need to set values for the following fields:

  • apiVersion - Which version of the Kubernetes API you're using to create this object
  • kind - What kind of object you want to create
  • metadata - Data that helps uniquely identify the object, including a name string, UID, and optional namespace
  • spec - What state you desire for the object

The precise format of the object spec is different for every Kubernetes object, and contains nested fields specific to that object. The Kubernetes API Reference can help you find the spec format for all of the objects you can create using Kubernetes.

For example, see the spec field for the Pod API reference. For each Pod, the .spec field specifies the pod and its desired state (such as the container image name for each container within that pod). Another example of an object specification is the spec field for the StatefulSet API. For StatefulSet, the .spec field specifies the StatefulSet and its desired state. Within the .spec of a StatefulSet is a template for Pod objects. That template describes Pods that the StatefulSet controller will create in order to satisfy the StatefulSet specification. Different kinds of objects can also have different .status; again, the API reference pages detail the structure of that .status field, and its content for each different type of object.

Server side field validation

Starting with Kubernetes v1.25, the API server offers server side field validation that detects unrecognized or duplicate fields in an object. It provides all the functionality of kubectl --validate on the server side.

The kubectl tool uses the --validate flag to set the level of field validation. It accepts the values ignore, warn, and strict while also accepting the values true (equivalent to strict) and false (equivalent to ignore). The default validation setting for kubectl is --validate=true.

Strict field validation, errors on validation failure
Field validation is performed, but errors are exposed as warnings rather than failing the request
No server side field validation is performed

When kubectl cannot connect to an API server that supports field validation it will fall back to using client-side validation. Kubernetes 1.27 and later versions always offer field validation; older Kubernetes releases might not. If your cluster is older than v1.27, check the documentation for your version of Kubernetes.

What's next

If you're new to Kubernetes, read more about the following:

Kubernetes Object Management explains how to use kubectl to manage objects. You might need to install kubectl if you don't already have it available.

To learn about the Kubernetes API in general, visit:

To learn about objects in Kubernetes in more depth, read other pages in this section:

1.1 - Kubernetes Object Management

The kubectl command-line tool supports several different ways to create and manage Kubernetes objects. This document provides an overview of the different approaches. Read the Kubectl book for details of managing objects by Kubectl.

Management techniques

Management techniqueOperates onRecommended environmentSupported writersLearning curve
Imperative commandsLive objectsDevelopment projects1+Lowest
Imperative object configurationIndividual filesProduction projects1Moderate
Declarative object configurationDirectories of filesProduction projects1+Highest

Imperative commands

When using imperative commands, a user operates directly on live objects in a cluster. The user provides operations to the kubectl command as arguments or flags.

This is the recommended way to get started or to run a one-off task in a cluster. Because this technique operates directly on live objects, it provides no history of previous configurations.


Run an instance of the nginx container by creating a Deployment object:

kubectl create deployment nginx --image nginx


Advantages compared to object configuration:

  • Commands are expressed as a single action word.
  • Commands require only a single step to make changes to the cluster.

Disadvantages compared to object configuration:

  • Commands do not integrate with change review processes.
  • Commands do not provide an audit trail associated with changes.
  • Commands do not provide a source of records except for what is live.
  • Commands do not provide a template for creating new objects.

Imperative object configuration

In imperative object configuration, the kubectl command specifies the operation (create, replace, etc.), optional flags and at least one file name. The file specified must contain a full definition of the object in YAML or JSON format.

See the API reference for more details on object definitions.


Create the objects defined in a configuration file:

kubectl create -f nginx.yaml

Delete the objects defined in two configuration files:

kubectl delete -f nginx.yaml -f redis.yaml

Update the objects defined in a configuration file by overwriting the live configuration:

kubectl replace -f nginx.yaml


Advantages compared to imperative commands:

  • Object configuration can be stored in a source control system such as Git.
  • Object configuration can integrate with processes such as reviewing changes before push and audit trails.
  • Object configuration provides a template for creating new objects.

Disadvantages compared to imperative commands:

  • Object configuration requires basic understanding of the object schema.
  • Object configuration requires the additional step of writing a YAML file.

Advantages compared to declarative object configuration:

  • Imperative object configuration behavior is simpler and easier to understand.
  • As of Kubernetes version 1.5, imperative object configuration is more mature.

Disadvantages compared to declarative object configuration:

  • Imperative object configuration works best on files, not directories.
  • Updates to live objects must be reflected in configuration files, or they will be lost during the next replacement.

Declarative object configuration

When using declarative object configuration, a user operates on object configuration files stored locally, however the user does not define the operations to be taken on the files. Create, update, and delete operations are automatically detected per-object by kubectl. This enables working on directories, where different operations might be needed for different objects.


Process all object configuration files in the configs directory, and create or patch the live objects. You can first diff to see what changes are going to be made, and then apply:

kubectl diff -f configs/
kubectl apply -f configs/

Recursively process directories:

kubectl diff -R -f configs/
kubectl apply -R -f configs/


Advantages compared to imperative object configuration:

  • Changes made directly to live objects are retained, even if they are not merged back into the configuration files.
  • Declarative object configuration has better support for operating on directories and automatically detecting operation types (create, patch, delete) per-object.

Disadvantages compared to imperative object configuration:

  • Declarative object configuration is harder to debug and understand results when they are unexpected.
  • Partial updates using diffs create complex merge and patch operations.

What's next

1.2 - Object Names and IDs

Each object in your cluster has a Name that is unique for that type of resource. Every Kubernetes object also has a UID that is unique across your whole cluster.

For example, you can only have one Pod named myapp-1234 within the same namespace, but you can have one Pod and one Deployment that are each named myapp-1234.

For non-unique user-provided attributes, Kubernetes provides labels and annotations.


A client-provided string that refers to an object in a resource URL, such as /api/v1/pods/some-name.

Only one object of a given kind can have a given name at a time. However, if you delete the object, you can make a new object with the same name.

Names must be unique across all API versions of the same resource. API resources are distinguished by their API group, resource type, namespace (for namespaced resources), and name. In other words, API version is irrelevant in this context.

Below are four types of commonly used name constraints for resources.

DNS Subdomain Names

Most resource types require a name that can be used as a DNS subdomain name as defined in RFC 1123. This means the name must:

  • contain no more than 253 characters
  • contain only lowercase alphanumeric characters, '-' or '.'
  • start with an alphanumeric character
  • end with an alphanumeric character

RFC 1123 Label Names

Some resource types require their names to follow the DNS label standard as defined in RFC 1123. This means the name must:

  • contain at most 63 characters
  • contain only lowercase alphanumeric characters or '-'
  • start with an alphanumeric character
  • end with an alphanumeric character

RFC 1035 Label Names

Some resource types require their names to follow the DNS label standard as defined in RFC 1035. This means the name must:

  • contain at most 63 characters
  • contain only lowercase alphanumeric characters or '-'
  • start with an alphabetic character
  • end with an alphanumeric character

Path Segment Names

Some resource types require their names to be able to be safely encoded as a path segment. In other words, the name may not be "." or ".." and the name may not contain "/" or "%".

Here's an example manifest for a Pod named nginx-demo.

apiVersion: v1
kind: Pod
  name: nginx-demo
  - name: nginx
    image: nginx:1.14.2
    - containerPort: 80


A Kubernetes systems-generated string to uniquely identify objects.

Every object created over the whole lifetime of a Kubernetes cluster has a distinct UID. It is intended to distinguish between historical occurrences of similar entities.

Kubernetes UIDs are universally unique identifiers (also known as UUIDs). UUIDs are standardized as ISO/IEC 9834-8 and as ITU-T X.667.

What's next

1.3 - Labels and Selectors

Labels are key/value pairs that are attached to objects such as Pods. Labels are intended to be used to specify identifying attributes of objects that are meaningful and relevant to users, but do not directly imply semantics to the core system. Labels can be used to organize and to select subsets of objects. Labels can be attached to objects at creation time and subsequently added and modified at any time. Each object can have a set of key/value labels defined. Each Key must be unique for a given object.

"metadata": {
  "labels": {
    "key1" : "value1",
    "key2" : "value2"

Labels allow for efficient queries and watches and are ideal for use in UIs and CLIs. Non-identifying information should be recorded using annotations.


Labels enable users to map their own organizational structures onto system objects in a loosely coupled fashion, without requiring clients to store these mappings.

Service deployments and batch processing pipelines are often multi-dimensional entities (e.g., multiple partitions or deployments, multiple release tracks, multiple tiers, multiple micro-services per tier). Management often requires cross-cutting operations, which breaks encapsulation of strictly hierarchical representations, especially rigid hierarchies determined by the infrastructure rather than by users.

Example labels:

  • "release" : "stable", "release" : "canary"
  • "environment" : "dev", "environment" : "qa", "environment" : "production"
  • "tier" : "frontend", "tier" : "backend", "tier" : "cache"
  • "partition" : "customerA", "partition" : "customerB"
  • "track" : "daily", "track" : "weekly"

These are examples of commonly used labels; you are free to develop your own conventions. Keep in mind that label Key must be unique for a given object.

Syntax and character set

Labels are key/value pairs. Valid label keys have two segments: an optional prefix and name, separated by a slash (/). The name segment is required and must be 63 characters or less, beginning and ending with an alphanumeric character ([a-z0-9A-Z]) with dashes (-), underscores (_), dots (.), and alphanumerics between. The prefix is optional. If specified, the prefix must be a DNS subdomain: a series of DNS labels separated by dots (.), not longer than 253 characters in total, followed by a slash (/).

If the prefix is omitted, the label Key is presumed to be private to the user. Automated system components (e.g. kube-scheduler, kube-controller-manager, kube-apiserver, kubectl, or other third-party automation) which add labels to end-user objects must specify a prefix.

The and prefixes are reserved for Kubernetes core components.

Valid label value:

  • must be 63 characters or less (can be empty),
  • unless empty, must begin and end with an alphanumeric character ([a-z0-9A-Z]),
  • could contain dashes (-), underscores (_), dots (.), and alphanumerics between.

For example, here's a manifest for a Pod that has two labels environment: production and app: nginx:

apiVersion: v1
kind: Pod
  name: label-demo
    environment: production
    app: nginx
  - name: nginx
    image: nginx:1.14.2
    - containerPort: 80

Label selectors

Unlike names and UIDs, labels do not provide uniqueness. In general, we expect many objects to carry the same label(s).

Via a label selector, the client/user can identify a set of objects. The label selector is the core grouping primitive in Kubernetes.

The API currently supports two types of selectors: equality-based and set-based. A label selector can be made of multiple requirements which are comma-separated. In the case of multiple requirements, all must be satisfied so the comma separator acts as a logical AND (&&) operator.

The semantics of empty or non-specified selectors are dependent on the context, and API types that use selectors should document the validity and meaning of them.

Equality-based requirement

Equality- or inequality-based requirements allow filtering by label keys and values. Matching objects must satisfy all of the specified label constraints, though they may have additional labels as well. Three kinds of operators are admitted =,==,!=. The first two represent equality (and are synonyms), while the latter represents inequality. For example:

environment = production
tier != frontend

The former selects all resources with key equal to environment and value equal to production. The latter selects all resources with key equal to tier and value distinct from frontend, and all resources with no labels with the tier key. One could filter for resources in production excluding frontend using the comma operator: environment=production,tier!=frontend

One usage scenario for equality-based label requirement is for Pods to specify node selection criteria. For example, the sample Pod below selects nodes with the label "accelerator=nvidia-tesla-p100".

apiVersion: v1
kind: Pod
  name: cuda-test
    - name: cuda-test
      image: ""
    accelerator: nvidia-tesla-p100

Set-based requirement

Set-based label requirements allow filtering keys according to a set of values. Three kinds of operators are supported: in,notin and exists (only the key identifier). For example:

environment in (production, qa)
tier notin (frontend, backend)
  • The first example selects all resources with key equal to environment and value equal to production or qa.
  • The second example selects all resources with key equal to tier and values other than frontend and backend, and all resources with no labels with the tier key.
  • The third example selects all resources including a label with key partition; no values are checked.
  • The fourth example selects all resources without a label with key partition; no values are checked.

Similarly the comma separator acts as an AND operator. So filtering resources with a partition key (no matter the value) and with environment different than qa can be achieved using partition,environment notin (qa). The set-based label selector is a general form of equality since environment=production is equivalent to environment in (production); similarly for != and notin.

Set-based requirements can be mixed with equality-based requirements. For example: partition in (customerA, customerB),environment!=qa.


LIST and WATCH filtering

LIST and WATCH operations may specify label selectors to filter the sets of objects returned using a query parameter. Both requirements are permitted (presented here as they would appear in a URL query string):

  • equality-based requirements: ?labelSelector=environment%3Dproduction,tier%3Dfrontend
  • set-based requirements: ?labelSelector=environment+in+%28production%2Cqa%29%2Ctier+in+%28frontend%29

Both label selector styles can be used to list or watch resources via a REST client. For example, targeting apiserver with kubectl and using equality-based one may write:

kubectl get pods -l environment=production,tier=frontend

or using set-based requirements:

kubectl get pods -l 'environment in (production),tier in (frontend)'

As already mentioned set-based requirements are more expressive. For instance, they can implement the OR operator on values:

kubectl get pods -l 'environment in (production, qa)'

or restricting negative matching via notin operator:

kubectl get pods -l 'environment,environment notin (frontend)'

Set references in API objects

Some Kubernetes objects, such as services and replicationcontrollers, also use label selectors to specify sets of other resources, such as pods.

Service and ReplicationController

The set of pods that a service targets is defined with a label selector. Similarly, the population of pods that a replicationcontroller should manage is also defined with a label selector.

Label selectors for both objects are defined in json or yaml files using maps, and only equality-based requirement selectors are supported:

"selector": {
    "component" : "redis",


  component: redis

This selector (respectively in json or yaml format) is equivalent to component=redis or component in (redis).

Resources that support set-based requirements

Newer resources, such as Job, Deployment, ReplicaSet, and DaemonSet, support set-based requirements as well.

    component: redis
    - { key: tier, operator: In, values: [cache] }
    - { key: environment, operator: NotIn, values: [dev] }

matchLabels is a map of {key,value} pairs. A single {key,value} in the matchLabels map is equivalent to an element of matchExpressions, whose key field is "key", the operator is "In", and the values array contains only "value". matchExpressions is a list of pod selector requirements. Valid operators include In, NotIn, Exists, and DoesNotExist. The values set must be non-empty in the case of In and NotIn. All of the requirements, from both matchLabels and matchExpressions are ANDed together -- they must all be satisfied in order to match.

Selecting sets of nodes

One use case for selecting over labels is to constrain the set of nodes onto which a pod can schedule. See the documentation on node selection for more information.

Using labels effectively

You can apply a single label to any resources, but this is not always the best practice. There are many scenarios where multiple labels should be used to distinguish resource sets from one another.

For instance, different applications would use different values for the app label, but a multi-tier application, such as the guestbook example, would additionally need to distinguish each tier. The frontend could carry the following labels:

  app: guestbook
  tier: frontend

while the Redis master and replica would have different tier labels, and perhaps even an additional role label:

  app: guestbook
  tier: backend
  role: master


  app: guestbook
  tier: backend
  role: replica

The labels allow for slicing and dicing the resources along any dimension specified by a label:

kubectl apply -f examples/guestbook/all-in-one/guestbook-all-in-one.yaml
kubectl get pods -Lapp -Ltier -Lrole
NAME                           READY  STATUS    RESTARTS   AGE   APP         TIER       ROLE
guestbook-fe-4nlpb             1/1    Running   0          1m    guestbook   frontend   <none>
guestbook-fe-ght6d             1/1    Running   0          1m    guestbook   frontend   <none>
guestbook-fe-jpy62             1/1    Running   0          1m    guestbook   frontend   <none>
guestbook-redis-master-5pg3b   1/1    Running   0          1m    guestbook   backend    master
guestbook-redis-replica-2q2yf  1/1    Running   0          1m    guestbook   backend    replica
guestbook-redis-replica-qgazl  1/1    Running   0          1m    guestbook   backend    replica
my-nginx-divi2                 1/1    Running   0          29m   nginx       <none>     <none>
my-nginx-o0ef1                 1/1    Running   0          29m   nginx       <none>     <none>
kubectl get pods -lapp=guestbook,role=replica
NAME                           READY  STATUS   RESTARTS  AGE
guestbook-redis-replica-2q2yf  1/1    Running  0         3m
guestbook-redis-replica-qgazl  1/1    Running  0         3m

Updating labels

Sometimes you may want to relabel existing pods and other resources before creating new resources. This can be done with kubectl label. For example, if you want to label all your NGINX Pods as frontend tier, run:

kubectl label pods -l app=nginx tier=fe
pod/my-nginx-2035384211-j5fhi labeled
pod/my-nginx-2035384211-u2c7e labeled
pod/my-nginx-2035384211-u3t6x labeled

This first filters all pods with the label "app=nginx", and then labels them with the "tier=fe". To see the pods you labeled, run:

kubectl get pods -l app=nginx -L tier
NAME                        READY     STATUS    RESTARTS   AGE       TIER
my-nginx-2035384211-j5fhi   1/1       Running   0          23m       fe
my-nginx-2035384211-u2c7e   1/1       Running   0          23m       fe
my-nginx-2035384211-u3t6x   1/1       Running   0          23m       fe

This outputs all "app=nginx" pods, with an additional label column of pods' tier (specified with -L or --label-columns).

For more information, please see kubectl label.

What's next

1.4 - Namespaces

In Kubernetes, namespaces provide a mechanism for isolating groups of resources within a single cluster. Names of resources need to be unique within a namespace, but not across namespaces. Namespace-based scoping is applicable only for namespaced objects (e.g. Deployments, Services, etc.) and not for cluster-wide objects (e.g. StorageClass, Nodes, PersistentVolumes, etc.).

When to Use Multiple Namespaces

Namespaces are intended for use in environments with many users spread across multiple teams, or projects. For clusters with a few to tens of users, you should not need to create or think about namespaces at all. Start using namespaces when you need the features they provide.

Namespaces provide a scope for names. Names of resources need to be unique within a namespace, but not across namespaces. Namespaces cannot be nested inside one another and each Kubernetes resource can only be in one namespace.

Namespaces are a way to divide cluster resources between multiple users (via resource quota).

It is not necessary to use multiple namespaces to separate slightly different resources, such as different versions of the same software: use labels to distinguish resources within the same namespace.

Initial namespaces

Kubernetes starts with four initial namespaces:

Kubernetes includes this namespace so that you can start using your new cluster without first creating a namespace.
This namespace holds Lease objects associated with each node. Node leases allow the kubelet to send heartbeats so that the control plane can detect node failure.
This namespace is readable by all clients (including those not authenticated). This namespace is mostly reserved for cluster usage, in case that some resources should be visible and readable publicly throughout the whole cluster. The public aspect of this namespace is only a convention, not a requirement.
The namespace for objects created by the Kubernetes system.

Working with Namespaces

Creation and deletion of namespaces are described in the Admin Guide documentation for namespaces.

Viewing namespaces

You can list the current namespaces in a cluster using:

kubectl get namespace
NAME              STATUS   AGE
default           Active   1d
kube-node-lease   Active   1d
kube-public       Active   1d
kube-system       Active   1d

Setting the namespace for a request

To set the namespace for a current request, use the --namespace flag.

For example:

kubectl run nginx --image=nginx --namespace=<insert-namespace-name-here>
kubectl get pods --namespace=<insert-namespace-name-here>

Setting the namespace preference

You can permanently save the namespace for all subsequent kubectl commands in that context.

kubectl config set-context --current --namespace=<insert-namespace-name-here>
# Validate it
kubectl config view --minify | grep namespace:

Namespaces and DNS

When you create a Service, it creates a corresponding DNS entry. This entry is of the form <service-name>.<namespace-name>.svc.cluster.local, which means that if a container only uses <service-name>, it will resolve to the service which is local to a namespace. This is useful for using the same configuration across multiple namespaces such as Development, Staging and Production. If you want to reach across namespaces, you need to use the fully qualified domain name (FQDN).

As a result, all namespace names must be valid RFC 1123 DNS labels.

Not all objects are in a namespace

Most Kubernetes resources (e.g. pods, services, replication controllers, and others) are in some namespaces. However namespace resources are not themselves in a namespace. And low-level resources, such as nodes and persistentVolumes, are not in any namespace.

To see which Kubernetes resources are and aren't in a namespace:

# In a namespace
kubectl api-resources --namespaced=true

# Not in a namespace
kubectl api-resources --namespaced=false

Automatic labelling

FEATURE STATE: Kubernetes 1.22 [stable]

The Kubernetes control plane sets an immutable label on all namespaces. The value of the label is the namespace name.

What's next

1.5 - Annotations

You can use Kubernetes annotations to attach arbitrary non-identifying metadata to objects. Clients such as tools and libraries can retrieve this metadata.

Attaching metadata to objects

You can use either labels or annotations to attach metadata to Kubernetes objects. Labels can be used to select objects and to find collections of objects that satisfy certain conditions. In contrast, annotations are not used to identify and select objects. The metadata in an annotation can be small or large, structured or unstructured, and can include characters not permitted by labels. It is possible to use labels as well as annotations in the metadata of the same object.

Annotations, like labels, are key/value maps:

"metadata": {
  "annotations": {
    "key1" : "value1",
    "key2" : "value2"

Here are some examples of information that could be recorded in annotations:

  • Fields managed by a declarative configuration layer. Attaching these fields as annotations distinguishes them from default values set by clients or servers, and from auto-generated fields and fields set by auto-sizing or auto-scaling systems.

  • Build, release, or image information like timestamps, release IDs, git branch, PR numbers, image hashes, and registry address.

  • Pointers to logging, monitoring, analytics, or audit repositories.

  • Client library or tool information that can be used for debugging purposes: for example, name, version, and build information.

  • User or tool/system provenance information, such as URLs of related objects from other ecosystem components.

  • Lightweight rollout tool metadata: for example, config or checkpoints.

  • Phone or pager numbers of persons responsible, or directory entries that specify where that information can be found, such as a team web site.

  • Directives from the end-user to the implementations to modify behavior or engage non-standard features.

Instead of using annotations, you could store this type of information in an external database or directory, but that would make it much harder to produce shared client libraries and tools for deployment, management, introspection, and the like.

Syntax and character set

Annotations are key/value pairs. Valid annotation keys have two segments: an optional prefix and name, separated by a slash (/). The name segment is required and must be 63 characters or less, beginning and ending with an alphanumeric character ([a-z0-9A-Z]) with dashes (-), underscores (_), dots (.), and alphanumerics between. The prefix is optional. If specified, the prefix must be a DNS subdomain: a series of DNS labels separated by dots (.), not longer than 253 characters in total, followed by a slash (/).

If the prefix is omitted, the annotation Key is presumed to be private to the user. Automated system components (e.g. kube-scheduler, kube-controller-manager, kube-apiserver, kubectl, or other third-party automation) which add annotations to end-user objects must specify a prefix.

The and prefixes are reserved for Kubernetes core components.

For example, here's a manifest for a Pod that has the annotation imageregistry: :

apiVersion: v1
kind: Pod
  name: annotations-demo
    imageregistry: ""
  - name: nginx
    image: nginx:1.14.2
    - containerPort: 80

What's next

1.6 - Field Selectors

Field selectors let you select Kubernetes objects based on the value of one or more resource fields. Here are some examples of field selector queries:

  • metadata.namespace!=default
  • status.phase=Pending

This kubectl command selects all Pods for which the value of the status.phase field is Running:

kubectl get pods --field-selector status.phase=Running

Supported fields

Supported field selectors vary by Kubernetes resource type. All resource types support the and metadata.namespace fields. Using unsupported field selectors produces an error. For example:

kubectl get ingress --field-selector
Error from server (BadRequest): Unable to find "ingresses" that match label selector "", field selector "": "" is not a known field selector: only "", "metadata.namespace"

List of supported fields


Supported operators

You can use the =, ==, and != operators with field selectors (= and == mean the same thing). This kubectl command, for example, selects all Kubernetes Services that aren't in the default namespace:

kubectl get services  --all-namespaces --field-selector metadata.namespace!=default

Chained selectors

As with label and other selectors, field selectors can be chained together as a comma-separated list. This kubectl command selects all Pods for which the status.phase does not equal Running and the spec.restartPolicy field equals Always:

kubectl get pods --field-selector=status.phase!=Running,spec.restartPolicy=Always

Multiple resource types

You can use field selectors across multiple resource types. This kubectl command selects all Statefulsets and Services that are not in the default namespace:

kubectl get statefulsets,services --all-namespaces --field-selector metadata.namespace!=default

1.7 - Finalizers

Finalizers are namespaced keys that tell Kubernetes to wait until specific conditions are met before it fully deletes resources marked for deletion. Finalizers alert controllers to clean up resources the deleted object owned.

When you tell Kubernetes to delete an object that has finalizers specified for it, the Kubernetes API marks the object for deletion by populating .metadata.deletionTimestamp, and returns a 202 status code (HTTP "Accepted"). The target object remains in a terminating state while the control plane, or other components, take the actions defined by the finalizers. After these actions are complete, the controller removes the relevant finalizers from the target object. When the metadata.finalizers field is empty, Kubernetes considers the deletion complete and deletes the object.

You can use finalizers to control garbage collection of resources. For example, you can define a finalizer to clean up related resources or infrastructure before the controller deletes the target resource.

You can use finalizers to control garbage collection of objects by alerting controllers to perform specific cleanup tasks before deleting the target resource.

Finalizers don't usually specify the code to execute. Instead, they are typically lists of keys on a specific resource similar to annotations. Kubernetes specifies some finalizers automatically, but you can also specify your own.

How finalizers work

When you create a resource using a manifest file, you can specify finalizers in the metadata.finalizers field. When you attempt to delete the resource, the API server handling the delete request notices the values in the finalizers field and does the following:

  • Modifies the object to add a metadata.deletionTimestamp field with the time you started the deletion.
  • Prevents the object from being removed until all items are removed from its metadata.finalizers field
  • Returns a 202 status code (HTTP "Accepted")

The controller managing that finalizer notices the update to the object setting the metadata.deletionTimestamp, indicating deletion of the object has been requested. The controller then attempts to satisfy the requirements of the finalizers specified for that resource. Each time a finalizer condition is satisfied, the controller removes that key from the resource's finalizers field. When the finalizers field is emptied, an object with a deletionTimestamp field set is automatically deleted. You can also use finalizers to prevent deletion of unmanaged resources.

A common example of a finalizer is, which prevents accidental deletion of PersistentVolume objects. When a PersistentVolume object is in use by a Pod, Kubernetes adds the pv-protection finalizer. If you try to delete the PersistentVolume, it enters a Terminating status, but the controller can't delete it because the finalizer exists. When the Pod stops using the PersistentVolume, Kubernetes clears the pv-protection finalizer, and the controller deletes the volume.

Owner references, labels, and finalizers

Like labels, owner references describe the relationships between objects in Kubernetes, but are used for a different purpose. When a controller manages objects like Pods, it uses labels to track changes to groups of related objects. For example, when a Job creates one or more Pods, the Job controller applies labels to those pods and tracks changes to any Pods in the cluster with the same label.

The Job controller also adds owner references to those Pods, pointing at the Job that created the Pods. If you delete the Job while these Pods are running, Kubernetes uses the owner references (not labels) to determine which Pods in the cluster need cleanup.

Kubernetes also processes finalizers when it identifies owner references on a resource targeted for deletion.

In some situations, finalizers can block the deletion of dependent objects, which can cause the targeted owner object to remain for longer than expected without being fully deleted. In these situations, you should check finalizers and owner references on the target owner and dependent objects to troubleshoot the cause.

What's next

1.8 - Owners and Dependents

In Kubernetes, some objects are owners of other objects. For example, a ReplicaSet is the owner of a set of Pods. These owned objects are dependents of their owner.

Ownership is different from the labels and selectors mechanism that some resources also use. For example, consider a Service that creates EndpointSlice objects. The Service uses labels to allow the control plane to determine which EndpointSlice objects are used for that Service. In addition to the labels, each EndpointSlice that is managed on behalf of a Service has an owner reference. Owner references help different parts of Kubernetes avoid interfering with objects they don’t control.

Owner references in object specifications

Dependent objects have a metadata.ownerReferences field that references their owner object. A valid owner reference consists of the object name and a UID within the same namespace as the dependent object. Kubernetes sets the value of this field automatically for objects that are dependents of other objects like ReplicaSets, DaemonSets, Deployments, Jobs and CronJobs, and ReplicationControllers. You can also configure these relationships manually by changing the value of this field. However, you usually don't need to and can allow Kubernetes to automatically manage the relationships.

Dependent objects also have an ownerReferences.blockOwnerDeletion field that takes a boolean value and controls whether specific dependents can block garbage collection from deleting their owner object. Kubernetes automatically sets this field to true if a controller (for example, the Deployment controller) sets the value of the metadata.ownerReferences field. You can also set the value of the blockOwnerDeletion field manually to control which dependents block garbage collection.

A Kubernetes admission controller controls user access to change this field for dependent resources, based on the delete permissions of the owner. This control prevents unauthorized users from delaying owner object deletion.

Ownership and finalizers

When you tell Kubernetes to delete a resource, the API server allows the managing controller to process any finalizer rules for the resource. Finalizers prevent accidental deletion of resources your cluster may still need to function correctly. For example, if you try to delete a PersistentVolume that is still in use by a Pod, the deletion does not happen immediately because the PersistentVolume has the finalizer on it. Instead, the volume remains in the Terminating status until Kubernetes clears the finalizer, which only happens after the PersistentVolume is no longer bound to a Pod.

Kubernetes also adds finalizers to an owner resource when you use either foreground or orphan cascading deletion. In foreground deletion, it adds the foreground finalizer so that the controller must delete dependent resources that also have ownerReferences.blockOwnerDeletion=true before it deletes the owner. If you specify an orphan deletion policy, Kubernetes adds the orphan finalizer so that the controller ignores dependent resources after it deletes the owner object.

What's next

1.9 - Recommended Labels

You can visualize and manage Kubernetes objects with more tools than kubectl and the dashboard. A common set of labels allows tools to work interoperably, describing objects in a common manner that all tools can understand.

In addition to supporting tooling, the recommended labels describe applications in a way that can be queried.

The metadata is organized around the concept of an application. Kubernetes is not a platform as a service (PaaS) and doesn't have or enforce a formal notion of an application. Instead, applications are informal and described with metadata. The definition of what an application contains is loose.

Shared labels and annotations share a common prefix: Labels without a prefix are private to users. The shared prefix ensures that shared labels do not interfere with custom user labels.


In order to take full advantage of using these labels, they should be applied on every resource object.

KeyDescriptionExampleType name of the applicationmysqlstring unique name identifying the instance of an applicationmysql-abcxyzstring current version of the application (e.g., a SemVer 1.0, revision hash, etc.)5.7.21string component within the architecturedatabasestring name of a higher level application this one is part ofwordpressstring tool being used to manage the operation of an applicationHelmstring

To illustrate these labels in action, consider the following StatefulSet object:

# This is an excerpt
apiVersion: apps/v1
kind: StatefulSet
  labels: mysql mysql-abcxyz "5.7.21" database wordpress Helm

Applications And Instances Of Applications

An application can be installed one or more times into a Kubernetes cluster and, in some cases, the same namespace. For example, WordPress can be installed more than once where different websites are different installations of WordPress.

The name of an application and the instance name are recorded separately. For example, WordPress has a of wordpress while it has an instance name, represented as with a value of wordpress-abcxyz. This enables the application and instance of the application to be identifiable. Every instance of an application must have a unique name.


To illustrate different ways to use these labels the following examples have varying complexity.

A Simple Stateless Service

Consider the case for a simple stateless service deployed using Deployment and Service objects. The following two snippets represent how the labels could be used in their simplest form.

The Deployment is used to oversee the pods running the application itself.

apiVersion: apps/v1
kind: Deployment
  labels: myservice myservice-abcxyz

The Service is used to expose the application.

apiVersion: v1
kind: Service
  labels: myservice myservice-abcxyz

Web Application With A Database

Consider a slightly more complicated application: a web application (WordPress) using a database (MySQL), installed using Helm. The following snippets illustrate the start of objects used to deploy this application.

The start to the following Deployment is used for WordPress:

apiVersion: apps/v1
kind: Deployment
  labels: wordpress wordpress-abcxyz "4.9.4" Helm server wordpress

The Service is used to expose WordPress:

apiVersion: v1
kind: Service
  labels: wordpress wordpress-abcxyz "4.9.4" Helm server wordpress

MySQL is exposed as a StatefulSet with metadata for both it and the larger application it belongs to:

apiVersion: apps/v1
kind: StatefulSet
  labels: mysql mysql-abcxyz "5.7.21" Helm database wordpress

The Service is used to expose MySQL as part of WordPress:

apiVersion: v1
kind: Service
  labels: mysql mysql-abcxyz "5.7.21" Helm database wordpress

With the MySQL StatefulSet and Service you'll notice information about both MySQL and WordPress, the broader application, are included.

2 - Kubernetes Components

A Kubernetes cluster consists of the components that are a part of the control plane and a set of machines called nodes.

When you deploy Kubernetes, you get a cluster.

A Kubernetes cluster consists of a set of worker machines, called nodes, that run containerized applications. Every cluster has at least one worker node.

The worker node(s) host the Pods that are the components of the application workload. The control plane manages the worker nodes and the Pods in the cluster. In production environments, the control plane usually runs across multiple computers and a cluster usually runs multiple nodes, providing fault-tolerance and high availability.

This document outlines the various components you need to have for a complete and working Kubernetes cluster.

Components of Kubernetes

The components of a Kubernetes cluster

Control Plane Components

The control plane's components make global decisions about the cluster (for example, scheduling), as well as detecting and responding to cluster events (for example, starting up a new pod when a Deployment's replicas field is unsatisfied).

Control plane components can be run on any machine in the cluster. However, for simplicity, setup scripts typically start all control plane components on the same machine, and do not run user containers on this machine. See Creating Highly Available clusters with kubeadm for an example control plane setup that runs across multiple machines.


The API server is a component of the Kubernetes control plane that exposes the Kubernetes API. The API server is the front end for the Kubernetes control plane.

The main implementation of a Kubernetes API server is kube-apiserver. kube-apiserver is designed to scale horizontally—that is, it scales by deploying more instances. You can run several instances of kube-apiserver and balance traffic between those instances.


Consistent and highly-available key value store used as Kubernetes' backing store for all cluster data.

If your Kubernetes cluster uses etcd as its backing store, make sure you have a back up plan for the data.

You can find in-depth information about etcd in the official documentation.


Control plane component that watches for newly created Pods with no assigned node, and selects a node for them to run on.

Factors taken into account for scheduling decisions include: individual and collective resource requirements, hardware/software/policy constraints, affinity and anti-affinity specifications, data locality, inter-workload interference, and deadlines.


Control plane component that runs controller processes.

Logically, each controller is a separate process, but to reduce complexity, they are all compiled into a single binary and run in a single process.

There are many different types of controllers. Some examples of them are:

  • Node controller: Responsible for noticing and responding when nodes go down.
  • Job controller: Watches for Job objects that represent one-off tasks, then creates Pods to run those tasks to completion.
  • EndpointSlice controller: Populates EndpointSlice objects (to provide a link between Services and Pods).
  • ServiceAccount controller: Create default ServiceAccounts for new namespaces.

The above is not an exhaustive list.


A Kubernetes control plane component that embeds cloud-specific control logic. The cloud controller manager lets you link your cluster into your cloud provider's API, and separates out the components that interact with that cloud platform from components that only interact with your cluster.

The cloud-controller-manager only runs controllers that are specific to your cloud provider. If you are running Kubernetes on your own premises, or in a learning environment inside your own PC, the cluster does not have a cloud controller manager.

As with the kube-controller-manager, the cloud-controller-manager combines several logically independent control loops into a single binary that you run as a single process. You can scale horizontally (run more than one copy) to improve performance or to help tolerate failures.

The following controllers can have cloud provider dependencies:

  • Node controller: For checking the cloud provider to determine if a node has been deleted in the cloud after it stops responding
  • Route controller: For setting up routes in the underlying cloud infrastructure
  • Service controller: For creating, updating and deleting cloud provider load balancers

Node Components

Node components run on every node, maintaining running pods and providing the Kubernetes runtime environment.


An agent that runs on each node in the cluster. It makes sure that containers are running in a Pod.

The kubelet takes a set of PodSpecs that are provided through various mechanisms and ensures that the containers described in those PodSpecs are running and healthy. The kubelet doesn't manage containers which were not created by Kubernetes.


kube-proxy is a network proxy that runs on each node in your cluster, implementing part of the Kubernetes Service concept.

kube-proxy maintains network rules on nodes. These network rules allow network communication to your Pods from network sessions inside or outside of your cluster.

kube-proxy uses the operating system packet filtering layer if there is one and it's available. Otherwise, kube-proxy forwards the traffic itself.

Container runtime

A fundamental component that empowers Kubernetes to run containers effectively. It is responsible for managing the execution and lifecycle of containers within the Kubernetes environment.

Kubernetes supports container runtimes such as containerd, CRI-O, and any other implementation of the Kubernetes CRI (Container Runtime Interface).


Addons use Kubernetes resources (DaemonSet, Deployment, etc) to implement cluster features. Because these are providing cluster-level features, namespaced resources for addons belong within the kube-system namespace.

Selected addons are described below; for an extended list of available addons, please see Addons.


While the other addons are not strictly required, all Kubernetes clusters should have cluster DNS, as many examples rely on it.

Cluster DNS is a DNS server, in addition to the other DNS server(s) in your environment, which serves DNS records for Kubernetes services.

Containers started by Kubernetes automatically include this DNS server in their DNS searches.

Web UI (Dashboard)

Dashboard is a general purpose, web-based UI for Kubernetes clusters. It allows users to manage and troubleshoot applications running in the cluster, as well as the cluster itself.

Container Resource Monitoring

Container Resource Monitoring records generic time-series metrics about containers in a central database, and provides a UI for browsing that data.

Cluster-level Logging

A cluster-level logging mechanism is responsible for saving container logs to a central log store with search/browsing interface.

Network Plugins

Network plugins are software components that implement the container network interface (CNI) specification. They are responsible for allocating IP addresses to pods and enabling them to communicate with each other within the cluster.

What's next

Learn more about the following:

3 - The Kubernetes API

The Kubernetes API lets you query and manipulate the state of objects in Kubernetes. The core of Kubernetes' control plane is the API server and the HTTP API that it exposes. Users, the different parts of your cluster, and external components all communicate with one another through the API server.

The core of Kubernetes' control plane is the API server. The API server exposes an HTTP API that lets end users, different parts of your cluster, and external components communicate with one another.

The Kubernetes API lets you query and manipulate the state of API objects in Kubernetes (for example: Pods, Namespaces, ConfigMaps, and Events).

Most operations can be performed through the kubectl command-line interface or other command-line tools, such as kubeadm, which in turn use the API. However, you can also access the API directly using REST calls. Kubernetes provides a set of client libraries for those looking to write applications using the Kubernetes API.

Each Kubernetes cluster publishes the specification of the APIs that the cluster serves. There are two mechanisms that Kubernetes uses to publish these API specifications; both are useful to enable automatic interoperability. For example, the kubectl tool fetches and caches the API specification for enabling command-line completion and other features. The two supported mechanisms are as follows:

  • The Discovery API provides information about the Kubernetes APIs: API names, resources, versions, and supported operations. This is a Kubernetes specific term as it is a separate API from the Kubernetes OpenAPI. It is intended to be a brief summary of the available resources and it does not detail specific schema for the resources. For reference about resource schemas, please refer to the OpenAPI document.

  • The Kubernetes OpenAPI Document provides (full) OpenAPI v2.0 and 3.0 schemas for all Kubernetes API endpoints. The OpenAPI v3 is the preferred method for accessing OpenAPI as it provides a more comprehensive and accurate view of the API. It includes all the available API paths, as well as all resources consumed and produced for every operations on every endpoints. It also includes any extensibility components that a cluster supports. The data is a complete specification and is significantly larger than that from the Discovery API.

Discovery API

Kubernetes publishes a list of all group versions and resources supported via the Discovery API. This includes the following for each resource:

  • Name
  • Cluster or namespaced scope
  • Endpoint URL and supported verbs
  • Alternative names
  • Group, version, kind

The API is available both aggregated and unaggregated form. The aggregated discovery serves two endpoints while the unaggregated discovery serves a separate endpoint for each group version.

Aggregated discovery

FEATURE STATE: Kubernetes v1.30 [stable]

Kubernetes offers stable support for aggregated discovery, publishing all resources supported by a cluster through two endpoints (/api and /apis). Requesting this endpoint drastically reduces the number of requests sent to fetch the discovery data from the cluster. You can access the data by requesting the respective endpoints with an Accept header indicating the aggregated discovery resource: Accept: application/json;v=v2;;as=APIGroupDiscoveryList.

Without indicating the resource type using the Accept header, the default response for the /api and /apis endpoint is an unaggregated discovery document.

The discovery document for the built-in resources can be found in the Kubernetes GitHub repository. This Github document can be used as a reference of the base set of the available resources if a Kubernetes cluster is not available to query.

The endpoint also supports ETag and protobuf encoding.

Unaggregated discovery

Without discovery aggregation, discovery is published in levels, with the root endpoints publishing discovery information for downstream documents.

A list of all group versions supported by a cluster is published at the /api and /apis endpoints. Example:

  "kind": "APIGroupList",
  "apiVersion": "v1",
  "groups": [
      "name": "",
      "versions": [
          "groupVersion": "",
          "version": "v1"
      "preferredVersion": {
        "groupVersion": "",
        "version": "v1"
      "name": "apps",
      "versions": [
          "groupVersion": "apps/v1",
          "version": "v1"
      "preferredVersion": {
        "groupVersion": "apps/v1",
        "version": "v1"

Additional requests are needed to obtain the discovery document for each group version at /apis/<group>/<version> (for example: /apis/, which advertises the list of resources served under a particular group version. These endpoints are used by kubectl to fetch the list of resources supported by a cluster.

OpenAPI interface definition

For details about the OpenAPI specifications, see the OpenAPI documentation.

Kubernetes serves both OpenAPI v2.0 and OpenAPI v3.0. OpenAPI v3 is the preferred method of accessing the OpenAPI because it offers a more comprehensive (lossless) representation of Kubernetes resources. Due to limitations of OpenAPI version 2, certain fields are dropped from the published OpenAPI including but not limited to default, nullable, oneOf.

OpenAPI V2

The Kubernetes API server serves an aggregated OpenAPI v2 spec via the /openapi/v2 endpoint. You can request the response format using request headers as follows:

Valid request header values for OpenAPI v2 queries
HeaderPossible valuesNotes
Accept-Encodinggzipnot supplying this header is also acceptable
Acceptapplication/com.github.proto-openapi.spec.v2@v1.0+protobufmainly for intra-cluster use
*serves application/json

OpenAPI V3

FEATURE STATE: Kubernetes v1.27 [stable]

Kubernetes supports publishing a description of its APIs as OpenAPI v3.

A discovery endpoint /openapi/v3 is provided to see a list of all group/versions available. This endpoint only returns JSON. These group/versions are provided in the following format:

    "paths": {
        "api/v1": {
            "serverRelativeURL": "/openapi/v3/api/v1?hash=CC0E9BFD992D8C59AEC98A1E2336F899E8318D3CF4C68944C3DEC640AF5AB52D864AC50DAA8D145B3494F75FA3CFF939FCBDDA431DAD3CA79738B297795818CF"
        "apis/": {
            "serverRelativeURL": "/openapi/v3/apis/"

The relative URLs are pointing to immutable OpenAPI descriptions, in order to improve client-side caching. The proper HTTP caching headers are also set by the API server for that purpose (Expires to 1 year in the future, and Cache-Control to immutable). When an obsolete URL is used, the API server returns a redirect to the newest URL.

The Kubernetes API server publishes an OpenAPI v3 spec per Kubernetes group version at the /openapi/v3/apis/<group>/<version>?hash=<hash> endpoint.

Refer to the table below for accepted request headers.

Valid request header values for OpenAPI v3 queries
HeaderPossible valuesNotes
Accept-Encodinggzipnot supplying this header is also acceptable
Acceptapplication/com.github.proto-openapi.spec.v3@v1.0+protobufmainly for intra-cluster use
*serves application/json

A Golang implementation to fetch the OpenAPI V3 is provided in the package

Kubernetes 1.30 publishes OpenAPI v2.0 and v3.0; there are no plans to support 3.1 in the near future.

Protobuf serialization

Kubernetes implements an alternative Protobuf based serialization format that is primarily intended for intra-cluster communication. For more information about this format, see the Kubernetes Protobuf serialization design proposal and the Interface Definition Language (IDL) files for each schema located in the Go packages that define the API objects.


Kubernetes stores the serialized state of objects by writing them into etcd.

API groups and versioning

To make it easier to eliminate fields or restructure resource representations, Kubernetes supports multiple API versions, each at a different API path, such as /api/v1 or /apis/

Versioning is done at the API level rather than at the resource or field level to ensure that the API presents a clear, consistent view of system resources and behavior, and to enable controlling access to end-of-life and/or experimental APIs.

To make it easier to evolve and to extend its API, Kubernetes implements API groups that can be enabled or disabled.

API resources are distinguished by their API group, resource type, namespace (for namespaced resources), and name. The API server handles the conversion between API versions transparently: all the different versions are actually representations of the same persisted data. The API server may serve the same underlying data through multiple API versions.

For example, suppose there are two API versions, v1 and v1beta1, for the same resource. If you originally created an object using the v1beta1 version of its API, you can later read, update, or delete that object using either the v1beta1 or the v1 API version, until the v1beta1 version is deprecated and removed. At that point you can continue accessing and modifying the object using the v1 API.

API changes

Any system that is successful needs to grow and change as new use cases emerge or existing ones change. Therefore, Kubernetes has designed the Kubernetes API to continuously change and grow. The Kubernetes project aims to not break compatibility with existing clients, and to maintain that compatibility for a length of time so that other projects have an opportunity to adapt.

In general, new API resources and new resource fields can be added often and frequently. Elimination of resources or fields requires following the API deprecation policy.

Kubernetes makes a strong commitment to maintain compatibility for official Kubernetes APIs once they reach general availability (GA), typically at API version v1. Additionally, Kubernetes maintains compatibility with data persisted via beta API versions of official Kubernetes APIs, and ensures that data can be converted and accessed via GA API versions when the feature goes stable.

If you adopt a beta API version, you will need to transition to a subsequent beta or stable API version once the API graduates. The best time to do this is while the beta API is in its deprecation period, since objects are simultaneously accessible via both API versions. Once the beta API completes its deprecation period and is no longer served, the replacement API version must be used.

Refer to API versions reference for more details on the API version level definitions.

API Extension

The Kubernetes API can be extended in one of two ways:

  1. Custom resources let you declaratively define how the API server should provide your chosen resource API.
  2. You can also extend the Kubernetes API by implementing an aggregation layer.

What's next