One command: wget -O /tmp/netdata-kickstart.sh https://get.netdata.cloud/kickstart.sh && sh /tmp/netdata-kickstart.sh. Netdata auto-discovers all running containers immediately. Open http://localhost:19999 and your dashboard is ready. Full installation guide

Yes, perfectly. Run docker compose up and all your services appear in the dashboard automatically. Web server, database, cache, workers - everything monitored with zero configuration. Netdata sees all containers on the host regardless of how they’re orchestrated.

docker stats shows current values only - no history, no trends, no alerting. Netdata stores metrics for days/weeks/months, provides historical analysis, includes 400+ pre-built alerts, and adds ML anomaly detection. Plus logs, processes, and network connections all in one view.

No. One Netdata agent per host monitors all containers on that host. It reads cgroup metrics directly from the kernel - no sidecars, no per-container agents, no overhead inside your containers.

Netdata collects metrics every second using Linux cgroups (the same technology Docker uses for resource isolation). You see exactly what’s happening now, not 30-60 second averages. This catches transient spikes that crash apps but disappear from minute-based monitoring.

Everything: CPU usage and throttling, memory (used, cache, swap, pressure), disk I/O (reads, writes, queue depth), network (bandwidth, packets, errors per interface), container state, restarts, and health check status. All at per-second granularity.

No. Netdata charges per node (host), not per container. Monitor 10 or 1000 containers on a host - same price. This is 90% cheaper than tools that charge per container or per custom metric. See pricing

Netdata trains 18 ML models per metric locally on each host. These models learn what ’normal’ looks like for each container. When behavior deviates, it flags genuine anomalies - not every minor fluctuation. 99% fewer false positives than static thresholds. Works from day one, no training period required.

Yes. Netdata shows processes inside containers with CPU, memory, disk I/O per process. Also network connections per container. All from your browser - no docker exec or SSH needed. With historical data too.

Container logs that Docker writes to systemd-journal (journald logging driver) appear automatically, with no log pipeline, and container logs from any other logging setup are collected through Netdata’s OpenTelemetry logs pipeline. Search, filter by container or severity, and correlate with metrics. Click any spike in a chart to see logs from that exact moment. No Elasticsearch needed.

Yes. Netdata monitors containers regardless of orchestrator - Docker Swarm, Docker Compose, standalone Docker, or Kubernetes. Install on each host and all containers are monitored automatically.

Netdata replaces all three with one agent. Per-second collection (vs 10-60s scraping), auto-generated dashboards (vs manual Grafana building), 400+ alerts included (vs custom PromQL), and ML anomaly detection. Plus: cAdvisor has known issues preventing container removal (GitHub #771) - Netdata reads cgroups directly, no such problems.

Yes. Netdata provides Prometheus-compatible endpoints and a Grafana datasource plugin. Run them side-by-side during migration. Most teams eventually drop the old stack because Netdata’s zero-config approach saves significant time.

Supported but not recommended for full visibility. Container-mode Netdata can’t access host-level metrics like eBPF or full process trees. For complete Docker monitoring, install Netdata directly on the host. If container deployment is required, use privileged mode with host PID/network namespaces.

Perfectly. New containers appear in dashboards within seconds of starting. Stopped containers clean up automatically. Ephemeral workloads work great - this is exactly what Netdata’s auto-discovery was designed for.

Yes, both cgroups v1 and v2 are fully supported with automatic detection. Works with any Linux distribution and Docker version. CPU, memory, disk I/O, and network metrics all work identically on both cgroup versions.

Yes, pre-configured. Netdata includes alerts for restart storms, OOM kills, CPU throttling, health check failures, and more. 400+ alert templates ready out of the box. Customize thresholds if needed, but defaults work well.

Install Netdata on each host. Connect them to Netdata Cloud for unified dashboards across all hosts. Or set up a Netdata Parent for centralized storage. Either way, each host’s containers are auto-discovered and monitored. Multi-host setup

By default, all metrics, logs and traces stay on your infrastructure. Netdata Cloud (optional) provides unified dashboards across hosts but stores only metadata - the actual data stays local. For fully air-gapped environments, local dashboards work for metrics without any cloud connection, and Netdata Cloud On-Premises provides log and trace exploration inside your network.

Yes. Netdata ingests OpenTelemetry metrics, logs and traces over OTLP (gRPC or HTTP), natively scrapes Prometheus endpoints, and exports via Remote Write. Containerized apps instrumented with OpenTelemetry SDKs, or an OpenTelemetry Collector container, can send spans to the Netdata Agent on the host; you explore them in the Traces tab in Netdata Cloud, next to the container metrics and logs. See OpenTelemetry and integrations.

Netdata’s distributed architecture eliminates cardinality bottlenecks. Each agent stores its own metrics locally - no central aggregation, no memory explosions, no query degradation as container count grows. Resources scale linearly per metric, not exponentially. Add containers freely without worrying about cost multiplication - this is fundamentally different from centralized systems like Prometheus.

The open-source Netdata Agent is free forever on any number of nodes, with community support. Netdata Cloud’s free Community plan adds multi-node dashboards for personal, non-commercial use, up to 5 active connected nodes. For production use, the Business plan adds 10 AI credits per month, SSO, and enterprise support - priced per node, not per container. See pricing.

See the Docker operations guides: vendor-neutral runbooks for disk exhaustion, OOM cascades, daemon hangs, log explosions, and container network isolation. Each covers detection, diagnosis, and prevention, and shows the per-second signals Netdata surfaces to catch them before they take the host down.