Monitoring splits into three problems that are easy to confuse: is it up, is it healthy, and what happened.
Updated 2026-10-03 · 13 options compared
The short answer
Uptime Kuma to find out a service is down before your users do. Prometheus + Grafana when you need to know *why* it went down. Netdata if you want per-second system detail with no configuration at all. Start with the first and add the second when a real incident makes you want history.
Monitoring splits into three problems that are easy to confuse: is it up, is it healthy, and what happened. The first is a two-minute install and pays for itself the first time a service dies at 3am. The second and third are real systems with real operational costs, and most people adopt them too early — collecting gigabytes of metrics nobody looks at. This list is ordered by the discipline of adding monitoring when a question actually demands it.
| # | Tool | License | Setup | Best for |
|---|---|---|---|---|
| 1 | Uptime Kuma | MIT | Beginner | Knowing a service is down before anyone tells you. |
| 2 | Gatus | Apache-2.0 | Beginner | Configuration-as-code health checks with a status page. |
| 3 | Netdata | GPL-3.0 | Beginner | Instant, zero-configuration insight into a machine's current state. |
| 4 | Grafana | AGPL-3.0 | Intermediate | Dashboards and visualisation over any metric source. |
| 5 | Prometheus | Apache-2.0 | Advanced | Collecting and querying metrics across many services. |
| 6 | Glances | LGPL-3.0 | Beginner | A quick per-machine overview from a terminal. |
| 7 | Grafana Loki | AGPL-3.0 | Advanced | Centralised logs without the cost of full-text indexing. |
| 8 | Dozzle | MIT | Beginner | Reading container logs quickly from a browser. |
| 9 | SigNoz | Apache-2.0 | Advanced | Traces, metrics and logs together for application debugging. |
| 10 | ntopng | GPL-3.0 | Advanced | Understanding what is actually happening on your network. |
| 11 | VictoriaMetrics | Apache-2.0 | Intermediate | Longer metric retention on limited hardware. |
| 12 | InfluxDB | MIT | Intermediate | Sensor and IoT metrics with high write volume. |
| 13 | Alertmanager | Apache-2.0 | Intermediate | Routing, grouping and silencing alerts once you have many rules. |
The correct sequence is: uptime monitoring first, because it answers the only question that matters at 3am. System metrics second, because they explain why. Log aggregation third, because it is the heaviest thing to run and you will not read it until something genuinely subtle breaks. Full observability with traces last — it is a substantial system, and if you adopt it before you have a problem it solves, you will spend your weekend maintaining it and your evenings not looking at it. Almost everyone who runs all four wishes they had started with the smaller two.
Best for: Knowing a service is down before anyone tells you.
The single highest-value self-hosted service in this entire directory. It handles HTTP, TCP, ping, DNS and push monitors, alerts through dozens of channels including Telegram and email, and publishes a public status page you can share. It is light enough to run on the same box it monitors, which is a slight limitation and also the reason people actually install it.
Replaces: UptimeRobot, Pingdom, Better Stack
Best for: Configuration-as-code health checks with a status page.
The declarative alternative: checks defined in a YAML file, with conditions on status codes, response bodies and response time. For anyone who already keeps configuration in version control, this is a far better fit than a click-through interface, and it produces a clean status page as a side effect.
Replaces: UptimeRobot, Pingdom, Statuspage
Best for: Instant, zero-configuration insight into a machine's current state.
Worth installing on every machine you own, because it requires no configuration and immediately shows per-second CPU, memory, disk, network and per-container detail. Its strength is instant diagnosis — you open it during a problem and the cause is usually visible within seconds. Long-term retention and alerting are the weaker parts.
Replaces: Datadog, Nagios, Zabbix
Best for: Dashboards and visualisation over any metric source.
Where metrics become readable: one place to build dashboards over Prometheus, Loki, InfluxDB, PostgreSQL and more. It is not itself a data store, and that is precisely why it pairs with almost everything else on this list. For most people it is the single most useful long-term addition to a homelab.
Replaces: Datadog, New Relic, Grafana Cloud
Best for: Collecting and querying metrics across many services.
The de facto standard for metrics collection, with a query language that has become the vocabulary of the whole discipline and an ecosystem of exporters for essentially everything. Adopting it is adopting a concept rather than a product. Budget for the storage, and configure retention deliberately.
Replaces: Datadog, New Relic, AWS CloudWatch
Best for: A quick per-machine overview from a terminal.
A terminal-based system monitor with a web mode, ideal for looking at one machine quickly without permanently running a metrics stack. Lower commitment than Netdata and easier to read than raw `top` output.
Replaces: htop, top, Netdata
Best for: Centralised logs without the cost of full-text indexing.
Log aggregation that indexes labels rather than the full text of every line, which makes it dramatically cheaper to run than a full-text log index. The trade is that queries are slower unless you have chosen your labels well. It is the pragmatic answer to "I want logs in Grafana next to my metrics".
Replaces: Elasticsearch, Splunk, Datadog Logs
Best for: Reading container logs quickly from a browser.
Live Docker container logs in a browser, with essentially no setup. It will not replace a log aggregation system, but it covers the most common day-to-day need — seeing what a misbehaving container is saying — better and faster than almost anything else.
Replaces: Datadog Logs, Papertrail, Docker Desktop
Best for: Traces, metrics and logs together for application debugging.
The strongest open-source observability platform here, with traces, metrics and logs in one interface built on OpenTelemetry. Adopt it when you have an actual distributed-systems problem, because instrumenting with OpenTelemetry from the start is what keeps you from being locked to a vendor later.
Replaces: Datadog, New Relic, Grafana Cloud
Best for: Understanding what is actually happening on your network.
Network-layer visibility that nothing else here provides: per-host and per-protocol traffic, flows, and history. Worth running temporarily when you need to answer "what on my network is using all the bandwidth", even if you do not keep it on permanently.
Replaces: Darktrace, PRTG, SolarWinds NPM
Best for: Longer metric retention on limited hardware.
A drop-in replacement for Prometheus storage that is markedly more efficient in both disk space and memory for the same data, and accepts the same remote-write protocol. If your Prometheus instance is straining its VPS, this is the least disruptive fix available.
Replaces: Prometheus, InfluxDB, Thanos
Best for: Sensor and IoT metrics with high write volume.
A time-series database with a gentler query language than PromQL and strong support for the IoT and sensor workloads that dominate home labs. The right choice when your metrics come from devices and scripts rather than from exporters.
Replaces: Prometheus, TimescaleDB, VictoriaMetrics
Best for: Routing, grouping and silencing alerts once you have many rules.
Prometheus detects conditions; Alertmanager decides who hears about them, how often, and how duplicates are collapsed. Without it, an incident produces two hundred identical notifications and you stop reading them. Essential once you go beyond a handful of alert rules.
Replaces: PagerDuty, Opsgenie, VictorOps
Cheap and easy, with free tiers usually capped at five-minute intervals — so you often learn about outages from users, not from it.
Superb and priced per host, per GB, per custom metric. That model makes sense on an enterprise budget and not on a home lab.
The same story: excellent product, costs that scale faster than your usage does.