Skip to content
StatusOwlDocs

Ping monitor

Check host reachability over ICMP echo, with round-trip latency.

Last updated

The ping monitor sends ICMP echo requests to a host and records whether any came back and how long they took. It is the cheapest, most generic reachability check there is — good for "is this box alive", useless for "is my service working".

What each check records

  • Status — up if at least one reply arrived within the timeout, down otherwise.
  • Latency — the average round-trip time across the echo requests in that check.

Internally the checking node shells out to the system ping binary rather than opening a raw socket, so the agent does not need elevated privileges. It parses the summary line for the average RTT and the received count.

Configuration

OptionNotes
HostHostname or IP. DNS is resolved on every check.
IntervalHow often to check. Your plan sets the floor.

Packet count and per-reply timeout are fixed at 3 packets and 2 seconds respectively. They are not exposed as per-monitor settings.

When to use ping

  • A router, switch, gateway or VM with no L7 service to check.
  • A basic liveness signal alongside a service-level monitor.
  • A host you deliberately do not expose over HTTP.

For anything customer-facing, pair ping with an HTTP or TCP monitor. A reply to a ping means the kernel is alive. It says nothing about whether your application is.

When not to use ping

  • Managed services and load balancers often drop ICMP. Cloud load balancers, CDN edges and many managed endpoints do not answer. Use HTTP.
  • ICMP is heavily deprioritised on the public internet. Spiky latency and bursty loss on a ping graph frequently reflect upstream policing rather than anything happening to your host.
  • Some networks block it outright. If the path from the checking node to your host drops ICMP, the monitor fails permanently regardless of health.

Example

example
Type:     Ping
Host:     edge-router-iad-1.internal
Interval: 60s

A down result means every echo in that check was dropped or arrived late — usually a hard network failure.

Ping plus HTTP tells you more than either alone

Run both against a production web host. HTTP failing while ping succeeds means the host is up and the service is wedged. Both failing means the host or the network is gone. That is two different incidents, and knowing which one you have before you open the laptop is worth the second monitor.

Limitations

  • ICMP only. No TCP, no UDP, no application-level signal.
  • Latency is an average across the packets in a check, so outliers are smoothed away. For percentile latency, monitor the service itself.
  • Packet-loss percentage is not surfaced as its own metric; a check is up if any reply arrives.

See also