Cortex Resource Manager

# Worker Management Configuration for Resource Manager MCP Server # Kubernetes configuration kubernetes: # Context to use (null for default) context: null # Namespace for worker operations (default: all namespaces) namespace: null # MCP Server endpoints for VM provisioning mcp_servers: # Talos MCP server endpoint (for Talos Linux VM provisioning) talos: endpoint: "stdio://talos-mcp-server" enabled: true # Proxmox MCP server endpoint (for Proxmox VM provisioning) proxmox: endpoint: "stdio://proxmox-mcp-server" enabled: true # Preferred provider for burst workers preferred_provider: "talos" # Worker size templates worker_sizes: small: cpu: 2 memory_gb: 4 disk_gb: 50 medium: cpu: 4 memory_gb: 8 disk_gb: 100 large: cpu: 8 memory_gb: 16 disk_gb: 200 xlarge: cpu: 16 memory_gb: 32 disk_gb: 400 # Burst worker configuration burst_workers: # Maximum number of burst workers that can be provisioned max_burst_workers: 10 # Default TTL in hours (if not specified) default_ttl_hours: 24 # Minimum TTL in hours min_ttl_hours: 1 # Maximum TTL in hours (1 week) max_ttl_hours: 168 # Label to identify burst workers worker_label: "worker-type=burst" # Annotation for TTL ttl_annotation: "worker-ttl" # Automatic cleanup check interval (seconds) cleanup_interval_seconds: 300 # Drain configuration drain: # Grace period for pod termination (seconds) grace_period_seconds: 300 # Ignore DaemonSets during drain ignore_daemonsets: true # Delete pods with emptyDir volumes delete_emptydir_data: true # Force deletion of standalone pods force: true # Timeout for drain operation (seconds) timeout_seconds: 600 # Safety configuration safety: # Require confirmation before destroying permanent workers (always true) protect_permanent_workers: true # Require drain before destroy (unless force=true) require_drain_before_destroy: true # List of protected worker name patterns (regex) protected_worker_patterns: - "^master-.*" - "^control-plane-.*" - "^permanent-.*" # Logging configuration logging: level: "INFO" format: "%(asctime)s - %(name)s - %(levelname)s - %(message)s" # Monitoring and metrics monitoring: # Enable metrics collection enabled: true # Metrics to track metrics: - "worker_count_by_type" - "worker_provision_time" - "worker_drain_time" - "worker_destroy_time" - "burst_worker_ttl_remaining"