kubernetes-sigs/descheduler

Features

• Dynamic Workload Rebalancers - Dynamically rearranges pods within the cluster to optimize distribution and balance load.
• Node Resource Balancing - Evicts pods from overutilized nodes to underutilized ones based on CPU, memory, and pod count thresholds.
• Node Utilization Optimizers - Balances CPU and memory utilization by moving pods from overutilized to underutilized nodes.
• Utilization-Based Evictions - Removes pods from over-utilized or under-utilized nodes to balance the workload across the cluster.
• Scheduling-Fit Validations - Ensures pods can be successfully rescheduled before eviction to prevent them from becoming unschedulable.
• Cluster Node Management - Verifies that a pod can be rescheduled based on selectors, taints, and resource requests before removal.
• Resource-Based Pod Redistribution - Evicts pods from overutilized nodes to underutilized nodes based on CPU, memory, and pod capacity thresholds.
• Workload Compaction - Evicts pods from underutilized nodes to pack workloads more densely and facilitate cluster downscaling.
• Cluster Resource Rebalancing - Evicts pods from overutilized nodes to underutilized ones to balance CPU and memory.
• Lifecycle State Evictions - Removes pods based on age, status phases, exit codes, or condition transitions to clean up stuck workloads.
• Pod Health Cleanup - Removes aged, unstable, or failed pods that exceed restart thresholds.
• Duplicate Pod Removal - Evicts duplicate pods from the same node that belong to the same controller to improve workload distribution.
• Pod Lifecycle Management - Manages the removal of pods based on their lifecycle state, including age and status transitions.
• Anti-Affinity Enforcement - Removes pods that violate anti-affinity rules to ensure pods that should not coexist on the same node are relocated.
• Topology Spread Enforcement - Moves pods to satisfy topology spread constraints and anti-affinity rules.
• Kubernetes API Clients - Integrates with the Kubernetes API to query cluster state and issue pod deletion requests.
• Cluster Autoscalers - Compacts workloads to enable cluster autoscalers to shut down unused worker nodes.
• Constraint-Based Evictions - Moves pods that no longer satisfy node affinity or topology spread constraints to more appropriate nodes.
• Node Scheduling Optimization - Relocates pods that no longer satisfy node affinity or taint requirements.
• Pod Eviction Managers - Removes unhealthy, aged, or duplicate pods to maintain cluster freshness and stability.
• Pod Health Evictions - Identifies and removes pods that have failed too many times or exceeded a specific lifetime.
• Pod Lifecycle Cleanups - Provides automated eviction of pods based on their age, status transitions, and exit codes.
• Pod Termination - Implements the removal of pods that exceed restart thresholds or enter failed status phases.
• Resource Distribution Rebalancing - Detects resource utilization imbalances and triggers pod rescheduling to optimize allocation.
• Scheduling-Fit Validations - Checks if a pod can be successfully rescheduled to another node before evicting it to prevent unschedulable pods.
• Topology Spread Enforcement - Evicts the minimum number of pods required to satisfy topology spread constraints and maintain maximum skew.
• Workload Compaction - Evicts pods from underutilized nodes into fewer, denser nodes to facilitate cluster downscaling.
• Node Load Balancing - Balances CPU and memory utilization by evicting pods from overutilized nodes to underutilized ones.
• Pod - Removes pods that exceed restart thresholds or are stuck in failure states.
• Eviction Strategies - Implements a modular plugin system for evaluating cluster state and identifying pods eligible for removal.
• Node Utilization Optimization - Evicts pods from overutilized nodes to underutilized nodes or compacts them to trigger autoscaling.
• Unstable Pod Evictions - Evicts pods that exceed a restart threshold or remain in a failed state to encourage rescheduling.
• Kubernetes Workload Label Filters - Uses label selectors and namespaces to filter and protect specific workloads from being evicted.
• Dry-Run Previews - Provides a dry-run mode to preview which pods would be evicted without performing actual deletions.
• Autoscaler Helpers - Packs pods onto fewer nodes to trigger cluster autoscaler downscaling of unused infrastructure.
• Node Affinity Rules - Evicts pods that no longer satisfy defined node affinity rules to ensure optimal placement.
• Distributed Recurring Task Scheduling - Executes cluster rebalancing tasks on a recurring schedule across multiple nodes while preventing concurrent execution.
• Dry-Run Modes - Provides a simulation mode to identify pod eviction candidates without applying actual changes to the cluster.
• Duplicate Pod Elimination - Evicts duplicate pods from the same node that belong to the same controller to improve workload distribution.
• Eviction Filters - Excludes pods from eviction based on labels, namespaces, pod age, minimum replica counts, or protection policies.
• Leader Election - Implements distributed locking to ensure only one instance manages the eviction process in high-availability deployments.
• Pod and Namespace Label Filtering - Filters pods eligible for removal using criteria such as labels, namespaces, and protection policies.
• Eviction Target Filtering - Restricts eviction candidates using label selectors, namespaces, and priority thresholds to protect critical workloads.
• Pod Distribution Optimizations - Evicts duplicate pods from a single node to ensure workloads are better distributed across the cluster.
• Pod Eviction Protections - Prevents the eviction of system-critical pods and DaemonSets to maintain cluster stability.
• Pod Distribution Balancing - Evicts duplicate pods from the same node to ensure better workload distribution.
• Topology Balancing Controllers - Enforces topology spread constraints by evicting pods to reduce skew across failure domains.
• Taint-Based Pod Evictions - Evicts workloads from nodes with specific taints to facilitate automated node recovery and scaling.
• Taint Enforcement - Removes pods that do not possess the necessary tolerations for taints assigned to a node.
• Failed Pod Cleanup - Evicts pods in a failed status phase based on configurable exit codes, reasons, or minimum lifetimes.
• Pod Health Dashboards - Identifies and removes pods based on failure phases or state transitions to maintain health.
• Pod Lifecycle Cleanup - Removes pods based on their lifetime or status transitions to maintain cluster freshness.
• Unstable Pod Cleanup - Evicts pods that exceed a restart threshold or maximum lifetime to clear failing or stale workloads.
• Infrastructure and Operations - Evicts pods based on custom policies.
• Workload Orchestration and Scheduling - Descheduler for Kubernetes.

स्टार हिस्ट्री