Karmada (Kubernetes Armada) is a Kubernetes management system that enables you to run your cloud-native applications across multiple Kubernetes clusters and clouds, with no changes to your applications. By speaking Kubernetes-native APIs and providing advanced scheduling capabilities, Karmada enables truly open, multi-cloud Kubernetes.
Karmada aims to provide turnkey automation for multi-cluster application management in multi-cloud and hybrid cloud scenarios, with key features such as centralized multi-cloud management, high availability, failure recovery, and traffic scheduling.
Karmada is an incubation project of the Cloud Native Computing Foundation (CNCF).
-
K8s Native API Compatible
- Zero change upgrade, from single-cluster to multi-cluster
- Seamless integration of existing K8s tool chain
-
Out of the Box
- Built-in policy sets for scenarios, including: Active-active, Remote DR, Geo Redundant, etc.
- Cross-cluster applications auto-scaling, failover and load-balancing on multi-cluster.
-
Avoid Vendor Lock-in
- Integration with mainstream cloud providers
- Automatic allocation, migration across clusters
- Not tied to proprietary vendor orchestration
-
Centralized Management
- Location agnostic cluster management
- Support clusters in Public cloud, on-prem or edge
-
Fruitful Multi-Cluster Scheduling Policies
- Cluster Affinity, Multi Cluster Splitting/Rebalancing
- Multi-Dimension HA: Region/AZ/Cluster/Provider
-
Open and Neutral
- Jointly initiated by Internet, finance, manufacturing, teleco, cloud providers, etc.
- Target for open governance with CNCF
Notice: this project is developed in continuation of Kubernetes Federation v1 and v2. Some basic concepts are inherited from these two versions.
The Karmada Control Plane consists of the following components:
- Karmada API Server
- Karmada Controller Manager
- Karmada Scheduler
ETCD stores the Karmada API objects, the API Server is the REST endpoint all other components talk to, and the Karmada Controller Manager performs operations based on the API objects you create through the API server.
The Karmada Controller Manager runs the various controllers, the controllers watch Karmada objects and then talk to the underlying clusters' API servers to create regular Kubernetes resources.
- Cluster Controller: attach Kubernetes clusters to Karmada for managing the lifecycle of the clusters by creating cluster objects.
- Policy Controller: the controller watches PropagationPolicy objects. When the PropagationPolicy object is added, it selects a group of resources matching the resourceSelector and creates ResourceBinding with each single resource object.
- Binding Controller: the controller watches ResourceBinding object and create Work object corresponding to each cluster with a single resource manifest.
- Execution Controller: the controller watches Work objects. When Work objects are created, it will distribute the resources to member clusters.
Resource template: Karmada uses Kubernetes Native API definition for federated resource template, to make it easy to integrate with existing tools that already adopt on Kubernetes
Propagation Policy: Karmada offers a standalone Propagation(placement) Policy API to define multi-cluster scheduling and spreading requirements.
- Support 1:n mapping of Policy: workload, users don't need to indicate scheduling constraints every time creating federated applications.
- With default policies, users can just interact with K8s API
Override Policy: Karmada provides standalone Override Policy API for specializing cluster relevant configuration automation. E.g.:
- Override image prefix according to member cluster region
- Override StorageClass according to cloud provider
The following diagram shows how Karmada resources are involved when propagating resources to member clusters.
This guide will cover:
- Install
karmadacontrol plane components in a Kubernetes cluster which is known ashost cluster. - Join a member cluster to
karmadacontrol plane. - Propagate an application by using
karmada.
git clone https://github.com/karmada-io/karmadacd karmadarun the following script:
hack/local-up-karmada.shThis script will do the following tasks for you:
- Start a Kubernetes cluster to run the Karmada control plane, aka. the
host cluster. - Build Karmada control plane components based on a current codebase.
- Deploy Karmada control plane components on the
host cluster. - Create member clusters and join Karmada.
If everything goes well, at the end of the script output, you will see similar messages as follows:
Local Karmada is running.
To start using your Karmada environment, run:
export KUBECONFIG="$HOME/.kube/karmada.config"
Please use 'kubectl config use-context karmada-host/karmada-apiserver' to switch the host and control plane cluster.
To manage your member clusters, run:
export KUBECONFIG="$HOME/.kube/members.config"
Please use 'kubectl config use-context member1/member2/member3' to switch to the different member cluster.There are two contexts in Karmada:
- karmada-apiserver
kubectl config use-context karmada-apiserver - karmada-host
kubectl config use-context karmada-host
The karmada-apiserver is the main kubeconfig to be used when interacting with the Karmada control plane, while karmada-host is only used for debugging Karmada installation with the host cluster. You can check all clusters at any time by running: kubectl config view. To switch cluster contexts, run kubectl config use-context [CONTEXT_NAME]
In the following steps, we are going to propagate a deployment by Karmada.
First, create a deployment named nginx:
kubectl create -f samples/nginx/deployment.yamlThen, we need to create a policy to propagate the deployment to our member cluster.
kubectl create -f samples/nginx/propagationpolicy.yamlYou can check deployment status from Karmada, don't need to access member cluster:
$ kubectl get deployment
NAME READY UP-TO-DATE AVAILABLE AGE
nginx 2/2 2 2 20s| Kubernetes 1.16 | Kubernetes 1.17 | Kubernetes 1.18 | Kubernetes 1.19 | Kubernetes 1.20 | Kubernetes 1.21 | Kubernetes 1.22 | Kubernetes 1.23 | Kubernetes 1.24 | Kubernetes 1.25 | Kubernetes 1.26 | Kubernetes 1.27 | Kubernetes 1.28 | Kubernetes 1.29 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Karmada v1.7 | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Karmada v1.8 | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Karmada v1.9 | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Karmada HEAD (master) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
Key:
✓Karmada and the Kubernetes version are exactly compatible.+Karmada has features or API objects that may not be present in the Kubernetes version.-The Kubernetes version has features or API objects that Karmada can't use.
Regular Community Meeting:
- Tuesday at 14:30 UTC+8 (Chinese)(biweekly). Convert to your timezone.
- Tuesday at 08:00 Pacific Time (English)(biweekly). Convert to your timezone.
Resources:
If you have questions, feel free to reach out to us in the following ways:
| Link | |
|---|---|
| KubeCon(EU 2021) | Beyond federation: automating multi-cloud workloads with K8s native APIs |
| KubeCon(EU 2022) | Sailing Multi Cloud Traffic Management With Karmada |
| KubeDay(Israel 2023) | Simplifying Multi-cluster Kubernetes Management with Karmada |
| KubeCon(China 2023) | Multi-Cloud Multi-Cluster HPA Helps Trip.com Group Deal with Business Downturn and Rapid Recovery |
| KubeCon(China 2023) | Break Through Cluster Boundaries to Autoscale Workloads Across Them on a Large Scale |
| KubeCon(China 2023) | Cross-Cluster Traffic Orchestration with eBPF |
| KubeCon(China 2023) | Non-Intrusively Enable OpenKruise and Argo Workflow in a Multi-Cluster Federation |
For blogs, please refer to website.
If you're interested in being a contributor and want to get involved in developing the Karmada code, please see CONTRIBUTING for details on submitting patches and the contribution workflow.
Karmada is under the Apache 2.0 license. See the LICENSE file for details.