Skip to main content
Mangrove shows how each app and the box as a whole are doing. It doesn’t have alerting rules; for alerts on an app, run a monitor like Uptime Kuma against it.

​
Per deployment

On a deployment’s Overview tab, each service shows:
  • its latest health check: healthy or not, and the response time;
  • its CPU and memory limits, replica count, image, ports and internal address.
Health checks run every 30 seconds by default and are kept for 7 days. See Health checks and self-healing. The History tab shows every run, and the deployment’s header shows a sparkline of its recent deploy outcomes.

​
The box

Server health in the sidebar shows the whole machine: CPU, memory, swap, disk, load average and uptime, refreshed every few seconds. Admin → Resource budget shows Mangrove’s own view:
  • memory allocated to deployments (their limits) against the budget ceiling;
  • memory actually used by containers;
  • running containers and disk usage;
  • a history of all of these, sampled every 5 minutes and kept for 30 days.

​
Warnings

  • For owners, the Mangrove mark in the sidebar turns red, with a link to Server health, when the box’s memory or disk use goes over 90%. It checks every 5 minutes.
  • A deploy that would exceed the memory budget is refused with an error rather than allowed to starve the box.

​
Getting metrics out

There’s no Prometheus endpoint. The same numbers are available from the API (GET /api/admin/system-health, /api/admin/resource-budget, /api/admin/resource-history?hours=24) if you want to scrape them. See HTTP API.