robusta-dev/krr
Krr
KRR is an open-source tool for analyzing Kubernetes resource requests and recommendations. It evaluates how pods are currently configured and provides suggestions for optimizing CPU and memory allocations based on actual usage patterns.
The project focuses on helping teams right-size their Kubernetes workloads by identifying over-provisioned and under-provisioned resources. It scans clusters and generates reports that highlight where adjustments can reduce costs or improve performance without compromising reliability.
KRR is distributed as a Python command-line tool that can be run directly against a Kubernetes cluster. Its documentation covers installation, configuration, and interpretation of the generated recommendations.
Features
- Pod Resource Request Scaling - Analyzes historical pod resource usage to suggest and optimize CPU and memory requests and limits.
- Kubernetes Resource Optimization - Analyzes pod usage data from Prometheus to suggest optimal CPU and memory requests and limits for containers.
- Cloud Cost Analysis Agents - Uses an artificial intelligence agent to investigate clusters for cost-saving opportunities beyond static rule-based analysis.
- Cluster Cost Optimizers - Uses an AI agent to investigate Kubernetes clusters and find cost-saving opportunities beyond static rule-based analysis.
- Automatic Recommendation Applications - Automatically updates cluster resource settings based on calculated recommendations to eliminate manual intervention.
- Automated Infrastructure Tuning - Automatically applies resource recommendation changes to clusters to eliminate manual intervention and improve stability.
- Cloud Infrastructure Cost Optimization - Identifies overspending in clusters using AI agents and rule-based analysis to reduce infrastructure waste.
- Infrastructure Reconciliation Engines - Automatically applies calculated resource recommendations directly to the infrastructure to eliminate manual configuration steps.
- Kubernetes Autoscaling Optimizers - Reduces cloud spend by automatically adjusting resource allocations based on historical workload usage.
- Resource Recommendation Engines - Analyzes Prometheus metrics to suggest optimal CPU and memory requests and limits for Kubernetes pods.
- Label-Based Selection - Implements filtering mechanisms using container labels to define the scope of resource analysis.
- Resource Cost Management - Uses an AI agent to investigate clusters and identify overspending and cost-saving opportunities beyond fixed rule-based recommendations.
- Resource Calculation Strategies - Calculates optimal CPU and memory settings by applying predefined rule-based logic to retrieved metric data.
- Kubernetes Metrics Analysis - Connects to time-series databases and cloud Prometheus endpoints to retrieve and filter historical workload performance data.
- Prometheus Metric Analyzers - Retrieves time-series data from Prometheus to identify over-provisioned cluster resources.
- Metric Data Source Integrations - Implements a plugin-based integration layer to retrieve historical usage metrics from diverse time-series databases and cloud providers.
- Workload Aggregations - Aggregates multiple related jobs into single groups based on labels to provide consolidated resource recommendations.
- Resource Recommendation Strategies - Provides a configuration interface for defining custom rules and logic to calculate optimal resource suggestions.
- Workload Aggregation Strategies - Groups individual pods or namespaces into logical units using regular expression patterns and label selectors for consolidated analysis.
- Plugin-Based Architectures - Employs a plugin-based architecture to standardize historical usage retrieval from diverse time-series databases and cloud providers.
- External Report Sinks - Sends recommendation data to external destinations such as chat applications, cloud storage, or web interfaces.
- Recommendation Sinks - Decouples recommendation generation from delivery by routing output data to various external destinations via a standardized export layer.
- Resource Recommendation Reporting - Exports calculated cost savings and resource suggestions to external sinks like chat apps or cloud storage.
