How Uptime Checks Work
Understand how MonitoringDog probes your endpoints, determines pass/fail, and manages check state.
When a check is created it starts in PENDING status, waiting for its first probe. Configuration is distributed to worker nodes, which run probes on your configured interval from each selected location simultaneously.
Check Interval
| Setting | Minimum | Maximum | Default |
|---|---|---|---|
| Interval | 15 seconds | 30 minutes | 30 seconds |
Check Types
The check type determines what a passing probe means:
| Type | Passes when |
|---|---|
status | Any 2xx HTTP response is received |
expected_status_code | A specific HTTP status code is received |
ping | ICMP ping succeeds |
keyword | Response body contains the specified string |
no_keyword | Response body does not contain the specified string |
tcp_port | The TCP port is reachable |
udp_port | The UDP port is reachable |
Timing Windows
Two timing windows control the state machine and prevent noise from transient failures.
Confirm Period
Range: 0–5 minutes | Default: 0
After a probe fails, the system waits for the confirm period to elapse before treating it as a real failure. If the probe recovers within this window, no incident is opened. This prevents a single flaky probe from creating an alert.
Recover Period
Range: 0–60 minutes | Default: 3 minutes
After a probe starts succeeding again, the check enters VALIDATING status. It must remain healthy for the full recover period before flipping back to UP. This prevents premature resolution of an unstable service.
Check States
| State | Description |
|---|---|
PENDING | Check created, waiting for the first probe result |
UP | Probes are passing |
DOWN | A confirmed failure — an incident is open |
VALIDATING | Probes are passing again, recovering period in progress |
PAUSED | Manually stopped — no probes are running |
MAINTENANCE | Within a configured maintenance window; probes run but failures are suppressed |
Maintenance Windows
Maintenance windows are defined by day-of-week, time range, and timezone. During a maintenance window the check stays in MAINTENANCE state — probes continue to run, but failures do not open incidents or trigger notifications.