Uptime Monitoring
Request method
GET or POST. POST lets you check an endpoint that only accepts writes (for example, a webhook receiver's own health check).
TCP port checks
Not everything speaks HTTP — a database, a mail server, a game server. Enter tcp://host:port as the URL instead of an http(s):// one, and Dumza checks whether that port accepts a connection instead of making an HTTP request. Method, expected status, keyword and response-time threshold don't apply to a TCP check and are hidden from the form; alerting, incidents and everything else work exactly the same.
Expected status code
Any endpoint that returns something other than your expected code counts as down. This isn't limited to 200 — plenty of APIs correctly return 201, 204, or even 401/503 for a healthy check (an endpoint that's SUPPOSED to require auth, for instance). Set whatever status your endpoint returns when it's actually healthy.
Keyword check
Optionally require (or forbid) a specific piece of text somewhere in the response body — useful for endpoints that return 200 even when something inside the page is broken (a CMS serving a generic error page with a 200 status, for example). Choose "contains" to require the text, or "does not contain" to fail when it appears (handy for catching a maintenance-mode banner your app shows while still returning 200).
Response time threshold
Set a maximum acceptable response time and a check that responds correctly but too slowly is marked "degraded" rather than "down" — it's visible on your dashboard and counted in your uptime stats, but it does NOT open an incident or send an alert. Degraded is a signal to watch, not to page anyone over.
Check interval
How often Dumza checks the endpoint. The free plan checks every 5 minutes; paid plans can go as low as every 1 minute. A shorter interval means faster detection of an outage, at the cost of more requests hitting your server.
Alert timing: notifyAfter and re-alerts
"Alert after N consecutive failures" controls how many bad checks in a row open an incident (minimum 2, to avoid false alarms from a single blip). Separately, "re-alert every N minutes" (optional, minimum 30) repeats the notification for as long as the monitor stays down — useful if your team's first alert gets missed.
Monitor status meanings
up — last check passed. down — an incident is open. degraded — responding, but slower than your threshold. pending — no check has run yet (just created, or paused). auth_error — a credentialed monitor's login token was rejected; see Authenticated Monitoring.
Pausing a monitor
A paused monitor stops being checked entirely and any open incident on it closes automatically — it won't keep alerting you for something you've deliberately taken offline (planned maintenance, decommissioning). Resume it any time to pick monitoring back up.