Ping monitoring with real round-trip times

What ping answers, and what it does not

An echo request asks a host to prove it exists on the network and can answer. A reply means the machine is powered, has an address, has a route back to the internet, and its kernel is alive enough to respond. No reply means one of those is not true.

That is a smaller claim than "the service works", and keeping the two apart is the whole value. A host that replies while its web server returns nothing tells you to look at the application or the proxy. A host that has stopped replying tells you to look at the provider, the network, or the power.

Ping also gives you a latency series reflecting the network path and nothing else. HTTP response time bundles network, TLS, server, and application logic together. Round-trip time is the network alone. When both climb, it is the path; when only the HTTP number climbs, it is your code.

Round-trip time, measured properly

The probe runs a single echo through the system ping binary and parses the time field out of the output, the familiar "time=12.3 ms". That parsed value is what gets recorded, not the wall-clock duration of the check.

The distinction matters more than it looks. Timing the whole operation would fold in process fork and exec overhead, which is milliseconds of noise on a measurement often only a few milliseconds to begin with. Wall-clock time is a fallback, used only when the output cannot be parsed.

Because the latency is meaningful, ping monitors get a full response-time chart and hourly roll-ups alongside uptime history. A ping series is one of the more legible graphs in monitoring: a step change is a routing change, a widening fuzzy band is congestion, and gaps are packet loss you would otherwise have argued about without evidence.

IPv6 targets are supported, with the platform-appropriate command chosen automatically. Targets are shape-validated before anything is executed, and the resolved address is checked first and used for the echo, so a name cannot be rebound to somewhere private between the check and the send.

Good candidates for a ping monitor

• A VPS or dedicated server, as the coarse "is the box there" signal underneath your service checks.
• A router, firewall, or gateway with a public address, where reachability is the whole job.
• A NAS or appliance exposed to the internet, so you learn about a power cut at home.
• A remote office link, where rising round-trip time is the early sign of a saturated line.
• Any host that already has an HTTP or TCP monitor, as the companion check telling you which layer failed.

Setting up a ping monitor

• Create a monitor and choose the Ping type.
• Enter a bare hostname or IP address. No scheme, no port, no path; anything implausible is rejected outright.
• Set the timeout. The 15-second default is very forgiving for ICMP.
• Set the failure threshold. Two consecutive failures is the default, with one success to clear. Packet loss is a fact of the internet, so three is reasonable for a distant target.
• Pick an interval: five minutes on the free plan, down to one minute on Pro.
• Attach notification channels, or leave the monitor unrouted so every team channel is alerted.

A ping monitor and a TCP monitor on the same host, tagged together, is worth adopting as a default pairing. Between them, the alert tells you whether to call your hosting provider or open your own logs.

Alerts, incidents, and plans

A confirmed failure opens an incident automatically, records the cause and duration, and resolves itself when the host answers again. Both transitions go out over email, ntfy, Telegram, Slack, Discord, and generic webhooks, all available on the free plan as well as Pro, with exactly-once delivery per transition per channel. Flap suppression mutes a host bouncing repeatedly into paired notifications, with the final recovery always delivered.

Free covers ten monitors permanently, five-minute checks, one public status page, and 30 days of history. Pro is a flat price for the whole team with no per-seat or per-monitor fees: 50 monitors, one-minute checks, a year of history, unlimited status pages and members, and custom domains, after a 14-day trial that needs no card.

Questions people ask