---
title: "Monitoring and availability"
description: "Nextvision Status — Monitoring and availability"
---

> Documentation Index
> Fetch the complete documentation index at: https://status.nextvisions.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Monitoring and availability

## Check types

HTTP/HTTPS checks validate status code, an optional body substring, response time and TLS validity/expiry. TCP checks attempt a connection. DNS checks resolve the configured hostname. ICMP checks IPv4 echo; use TCP/HTTP for IPv6. Heartbeats detect missing cron signals. System agents collect CPU, memory, root-filesystem usage and network byte counters. Prometheus scraping accepts only allowed metric names and labels; scalar gauge, counter and untyped samples are supported.

Public probes reject private, loopback, link-local, metadata and special-purpose destinations. Private probes must match a configured CIDR allowlist and still cannot reach metadata/link-local addresses. Address checks run for DNS answers, dial operations and redirects. Probe networking never uses environment proxy settings.

## State evaluation

Default interval is 60 seconds and timeout is 10 seconds. A probe declares failure after three consecutive failed checks and recovery after two successes. Older results do not overwrite newer probe states. Job ID plus scheduled time prevents duplicate samples.

A strict majority of all assigned probes must agree; offline probes remain in the denominator. Without a majority the status is Unknown. A component with both healthy and failed monitors shows Partial outage; all failed shows Major outage.

The UI uses text/icons alongside status colors. Gray uptime bars represent missing data.

## Availability calculation

Uptime = good seconds / (good seconds + bad seconds). Maintenance and Unknown are excluded from that ratio. Coverage = (good + bad + maintenance seconds) / expected seconds. Daily buckets use UTC. Skipped scheduler slots are recorded as Unknown when scheduling resumes, so downtime lowers data coverage. Compare last-updated times while the scheduler is unavailable.

Maintenance continues measurement and suppresses ordinary outage alerts for linked components. Manual incident content remains operator-controlled.

## Metrics bounds

Free allows three infrastructure monitors and 100 active metric series per organization. Metric and label allowlists prevent unbounded Prometheus cardinality. Numeric metrics are stored raw, then averaged hourly. Latency p50/p95 uses successful raw checks from the preceding 24 hours. Counters remain counters; do not interpret an hourly average as a rate.

## Heartbeat

```sh
curl -X POST https://your-status-host/api/v1/heartbeats/ONE_TIME_MONITOR_TOKEN
```

Treat the heartbeat token as a secret. Create a replacement monitor to replace a lost token.

Source: https://status.nextvisions.ai/docs/en/monitoring/index.mdx