Self-hosted uptime monitors, from home lab to status page

Uptime Kuma is a monitor you set up entirely in the browser, and it’s what we’d install first if you run a home lab or a few small sites and want to hear about outages on the app you already use. One docker run command starts it, and data lives in a mounted folder with SQLite, or in MariaDB if you prefer.

Its strongest point is alerts. The README counts more than 90 notification services, the longest list of the four, so whatever your phone or team chat uses is almost certainly covered. Checks can repeat every 20 seconds.

The trade-offs show up over years rather than on day one. The move to the current major version removed backup and restore from JSON, so copying the data folder is now the only supported backup, and the docs describe no other way to export your monitors. If you want to script changes or connect other tools, the API behind the dashboard is internal, not officially supported, and may change between versions without notice. There is one login: support for several admin users is the subject of a pull request, not a documented feature.

Kener calls itself a status page system, and that is the job it does best. We’d recommend it when the page your users see matters more than how many places alerts can go, because it does the communication side more fully than the others: incidents carry a public timeline of updates, scheduled maintenance is announced ahead of time, and visitors can subscribe by email or follow an RSS feed. One gap for a customer-facing page: the docs don’t describe giving a status page its own domain, which Uptime Kuma and Checkmate both document.

It monitors more than the label suggests. Monitors cover HTTP APIs, ping, TCP, DNS, certificate expiry, SQL queries, heartbeats and Docker containers, and you can create them from the admin panel or through its REST API with API keys. Sign-in covers local accounts and OpenID Connect, which lets people log in through an identity provider such as Keycloak or Okta, with roles you can define yourself.

It is slower to notice an outage than the others can be. Checks run on cron expressions, the scheduling format Unix systems use, and the shortest step the docs show is one minute, where Uptime Kuma goes down to 20 seconds and Checkmate to 15.

Gatus is for people who would rather write a monitor than click one together. Every endpoint goes in a YAML file with conditions on the status code, response time, body, certificate expiry or even domain expiry. Those conditions check what a page or API returns as well as whether it answers, and the README suggests using Gatus for automated user acceptance tests. That file is the whole setup, which makes Gatus the easiest of the four to rebuild, review or move: it is already text you can keep in version control.

It is also light: the README calls its footprint “negligibly small”, and it runs as one container with no database server. It says it can handle any interval, though it advises longer intervals for slow checks, and its author uses 30 seconds for simple ones. Alerts can go to a long list of providers, from AWS SES and Gotify to Slack, PagerDuty and Zulip, plus custom alerts, and a read-only API, Prometheus metrics and badges let other tools use the results.

What you give up is the dashboard as a place to work. The docs describe no way to edit monitors in the browser, and no status page features beyond announcements shown at the top of the page. Sign-in is Basic authentication, the browser’s own username-and-password prompt, or OpenID Connect, and the docs describe no user roles. They also list no database query or Docker container checks; Kener documents both, and Checkmate documents Docker checks.

Checkmate is an uptime monitor that also watches the servers themselves: with its Capture agent installed on a machine, it tracks CPU, memory and disk next to your uptime checks. We’d recommend it to a small team that runs its own servers. It has three user roles, so colleagues can watch the dashboards while only admins change monitors.

Its checks reach further than most. They can run as often as every 15 seconds, and geo checks send the same request from other locations every 5 to 30 minutes, so you can tell a regional outage from a global one; the docs don’t say whose machines those locations are. The project says one deployment monitors more than 1,600 servers in production.

Moving to a new server is straightforward: export every monitor’s settings to a JSON file and import it into the new Checkmate install. A REST API, documented on an interactive Swagger page, can create monitors too.

Its weaker side is outward-facing. Alerts reach 13 services, fewer than Uptime Kuma or Gatus. Its docs also describe no Prometheus metrics or badges for monitors, and no incident notices or subscriptions on its status pages.