OpenShift OVN-Kubernetes Benchmark MCP Server

/tools/ovnk_benchmark_prometheus_ovnk_multus.py This is the main collector class MultusPODsUsageCollector that: Inherits from the uploaded modules: Uses PrometheusBaseQuery for Prometheus queries and OpenShiftAuth for Kubernetes authentication Reads metrics.yaml first: Checks for existing PromQL queries in metrics.yaml, falls back to default queries if not found Supports both scenarios: collect_instant_usage() - for instant metrics collect_duration_usage() - for time-series metrics over a duration collect_usage() - unified method that calls appropriate method based on duration parameter Gets pod-node mapping: Uses Kubernetes API via OpenShiftAuth to map pod names to node names Formats output: Converts memory units (bytes/KB/MB/GB) appropriately and provides top 10 pod usage summary Timezone support: All timestamps in UTC Error handling: Comprehensive error handling with detailed error messages Key features: Uses the default PromQL queries you provided for Multus pods Concurrent execution of multiple Prometheus queries Proper memory unit conversion and formatting Top 10 CPU usage ranking Complete JSON summary with pod names, node names, min/avg/max usage 2. analysis/ovnk_benchmark_performance_analysis_ovnk_multus.py This is the performance analysis module that: Analyzes pod usage data: Takes JSON input from the collector and performs comprehensive analysis No file I/O: Pure in-memory processing, returns JSON data Comprehensive metrics: Individual pod analysis with efficiency scores Aggregate cluster statistics Usage level categorization (low/medium/high/very_high) Performance variability analysis Resource utilization balance Smart recommendations: Generates actionable recommendations based on usage patterns Cluster insights: Node distribution, performance health, bottleneck detection Trend analysis: For duration-based collections, analyzes performance trends over time Key analysis features: CPU and memory efficiency scoring Usage variance and stability analysis Resource balance detection (CPU-heavy vs memory-heavy workloads) Node distribution balance analysis Performance health assessment Global and pod-specific recommendations