karmada-io/karmada
Karmada
Karmada is a Kubernetes multi-cluster orchestrator and multi-cloud cluster manager designed to deploy and manage cloud-native applications across multiple clusters and cloud providers. It serves as a centralized control plane that functions as a resource propagator and workload scheduler, coordinating resources across public clouds, on-premises data centers, and edge locations.
The project distinguishes itself through a policy-based engine that distributes applications using affinity, topology constraints, and resource quotas. It provides specific capabilities for multi-region disaster recovery, including automated application failover and geo-redundant deployments to maintain service availability. Additionally, it allows for cluster-aware resource overriding to specialize configuration parameters based on the target cluster region or provider.
The system covers a broad range of operational areas, including bidirectional state synchronization, aggregated API proxying, and multi-dimensional scheduling. It includes tools for cluster lifecycle management, global resource search, and cross-cluster traffic balancing.
Installation and management can be performed using an operator-based installation process and a dedicated command line interface for administration and control plane operations.
Features
- Multi-Cluster Orchestrators - Provides a centralized control plane for deploying and managing cloud-native applications across multiple Kubernetes clusters.
- Multi-Cluster Fleet Managers - Acts as a centralized control plane for coordinating and managing multiple Kubernetes clusters across public clouds, on-premises, and edge locations.
- Automated Application Failovers - Provides automated migration of replicas from faulty clusters to healthy ones to ensure continuous service availability.
- Bidirectional Resource Synchronization - Synchronizes resource manifests and status updates bidirectionally between the central control plane and member clusters.
- Cluster Administration - Allows accessing and querying resources from multiple registered Kubernetes clusters through a single unified API endpoint.
- Workload Scheduling - Distributes applications across clusters using affinity and topology constraints to optimize resource placement.
- Affinity-Based Placement - Uses affinity rules to provide hints about which specific clusters are most suitable for workload deployment.
- Topology-Aware Placement - Schedules workloads based on metadata such as region, provider, and zone to optimize for proximity.
- Cross-Cluster Resource Synchronization - Synchronizes manifests and status bidirectionally between a central control plane and remote clusters.
- Propagation Controllers - Distributes custom resource definitions and instances from a central control plane based on defined propagation policies.
- Hub-and-Spoke Control Planes - Uses a central management cluster to coordinate and distribute resource state to multiple registered member clusters.
- High Availability Deployments - Implements active-active and geo-redundant deployment strategies to ensure application uptime across multiple zones.
- Hybrid Cloud Kubernetes Management - Controls Kubernetes clusters residing across public clouds, on-premises data centers, and edge locations from one interface.
- Kubernetes Resource Propagation - Distributes native and custom resources from a central API to multiple member clusters.
- Multi-Cloud Kubernetes Coordination - Provides a centralized system for coordinating Kubernetes resources across public clouds, on-premises data centers, and edge locations.
- Unified API Access - Aggregates multiple registered clusters into a single endpoint for consistent interface access.
- Resource Management Policies - Defines how resources are distributed to member clusters using namespace or cluster-wide orchestration policies.
- Multi-Dimensional Scheduling - Allocates workloads across clusters by evaluating availability zones, cloud providers, and custom affinity rules.
- Resource Distribution Policies - Determines workload distribution by mapping resources to target clusters using affinity and topology constraints.
- Kubernetes API Aggregations - Collects and caches objects and events from multiple remote clusters into a unified Kubernetes API endpoint.
- Cross-Cluster Failover Orchestrators - Coordinates active failover and traffic switching between clusters to maintain availability during infrastructure outages.
- Cross-Cluster Resource Querying - Retrieves and views resources from connected member clusters via a unified API endpoint.
- Application Failover Automation - Automates application failover and geo-redundant deployments across different regions to ensure continuous service availability.
- Application Dependency Co-location - Deploys required dependencies to the same cluster as the main application to ensure functional consistency.
- Cluster Configuration Management - Provides tools to define cluster-specific behavior and configuration rules for resources across a multi-cluster environment.
- Resource Parameter Specialization - Adjusts resource parameters like image prefixes and storage classes based on the target cluster region.
- Programmatic Cluster Registration - Connects remote clusters to a central control plane using direct API push or agent-based pull mechanisms.
- Cluster Lifecycle Management - Attaches and tracks clusters to a central management system to coordinate their operational state.
- Cluster Management Connectivity - Establishes management connections to member clusters using either direct API server communication or agent-based delegation.
- Configuration Overrides - Implements capabilities to modify specific fields of propagated resources to meet local cluster requirements.
- Multi-Region Application Availability Controllers - Automates application failover and geo-redundant deployments to maintain service availability across regions.
- Resource Overriding - Modifies specific resource fields based on the target cluster region or provider requirements.
- Namespace Propagation Controls - Controls whether namespaces are automatically distributed to all member clusters or manually assigned via policies.
- Resource Distribution Rebalancing - Detects instance state changes and triggers rescheduling to maintain optimal resource distribution across clusters.
- Unified Cluster Operations - Executes management commands and resource searches across all member clusters from a single API context via CLI.
- Multi-Cluster Service Connectivity - Connects services and containers across different clusters using integrated import and export tools.
- Push-Pull State Distribution - Supports both direct API server communication (push) and agent-based delegation (pull) to synchronize cluster state.
- Kubernetes Resource Status Aggregators - Collects and aggregates the status of resources from all member clusters into a single unified view.
- Global Resource Searches - Caches objects and events from multiple clusters to provide unified retrieval and proxy access.
- Deployment Status Monitors - Tracks the synchronization and readiness state of distributed resources across member clusters from a central interface.
- Custom Resource APIs - Allows the definition and management of custom resource types and endpoints within the cluster orchestrator.
- Workload Orchestration and Scheduling - Multi-cluster Kubernetes orchestration.
