google/cadvisor
Cadvisor
cAdvisor is a container resource monitoring agent and performance analyzer that collects and exports CPU, memory, network, and disk usage statistics from running containers. It functions as a telemetry tool for discovering containers across various runtimes and serves as a Prometheus-compatible metrics exporter.
The agent distinguishes itself by analyzing Linux control groups to provide visibility into resource consumption and limits. It utilizes kernel perf events and NUMA statistics for low-level hardware performance tracking and diagnostics, and it can identify out-of-memory kill events through kernel log monitoring.
The system provides broad observability capabilities, including host hardware analysis for CPU topology and memory capacity, as well as real-time container event streaming. It includes a built-in web interface for metrics visualization and supports programmatic data retrieval via a REST API.
Installation is supported as a containerized image, a Kubernetes DaemonSet for cluster-wide node observability, or as a standalone static binary.
Features
- Container Monitoring - Functions as a daemon that collects and exports CPU, memory, network, and disk usage statistics from containers.
- Control-Group-Based Resource Management - Analyzes the Linux control groups virtual filesystem to track CPU, memory, and filesystem statistics.
- Data Exposure - Exposes gathered resource data through a JSON REST API for integration with external monitoring tools.
- System Resource APIs - Provides an interface that exposes internal container and machine resource states as structured data.
- Statistics APIs - Provides a remote API to share raw and processed container statistics with external monitoring clients.
- Container Metric Collectors - Provides a client-facing API for retrieving resource usage statistics and health metrics from containers.
- Container Runtime Discovery - Detects active containers by scanning filesystem paths across multiple different container runtime implementations.
- Runtime Discovery - Automatically detects and monitors containers managed by various runtimes and orchestrators.
- Telemetry Tools - Discovers and monitors containers across various runtimes to track hardware and software performance.
- Cgroup Resource Analyzers - Analyzes Linux control groups to provide deep visibility into container resource consumption and limits.
- Metrics Exporters - Pushes collected telemetry and resource metrics to external backends for long-term analysis.
- Kubernetes Monitors - Provides comprehensive resource tracking and health monitoring for nodes and containers within Kubernetes clusters.
- Performance Metrics APIs - Implements a programmatic interface allowing external tools to retrieve raw and processed performance statistics.
- Telemetry Exporters - Forwards real-time container resource statistics to external storage and analysis platforms.
- Containerized Deployments - Supports execution as a container image with host volume mappings to observe system states.
- DaemonSet Deployment Controllers - Utilizes the DaemonSet deployment pattern to ensure local data collection on every cluster node.
- Kubernetes Deployments - Supports deployment across Kubernetes clusters using a DaemonSet for node-level observability.
- Standalone Binaries - Can be executed as a standalone static binary without external dependencies.
- Out-of-Memory Analysis - Monitors kernel logs and cgroup events to identify and report out-of-memory kill events.
- Recursive Data Collection - Provides the ability to aggregate resource usage for a target container and all its sub-containers recursively.
- Kernel Event Tracers - Integrates with kernel perf events to gather low-level hardware events and CPU performance counters.
- Hardware Monitoring Tools - Monitors low-level hardware events, including NUMA statistics and hugepages via core and uncore monitoring.
- Hardware Performance Monitoring - Measures low-level kernel perf events to provide aggregated or per-CPU hardware performance data.
- Host Hardware Analysis Tools - Provides an API to retrieve detailed machine metadata including CPU topology and memory capacity.
- Metrics Visualizers - Ships a built-in web interface for real-time visualization of resource usage and machine specifications.
- Prometheus-Based Metric Exporters - Exposes container and host resource data as a Prometheus-compatible metrics exporter.
- Traffic Monitoring Tools - Collects detailed network statistics including TCP and UDP metrics across various device types.
- Container Infrastructure - Tool for analyzing resource usage of running containers.
- Data Collection Agents - Analyze resource usage of running containers.
- Monitoring and Logging - Analysis of resource usage and performance for containers.
- Monitoring and Status - Analyzes resource usage and performance of containers.
- Monitoring Systems - Container resource usage and performance analyzer.
- Observability and Monitoring - Analyzes resource usage of running containers.
