icmp_check — multi-target ping

Free tier. Pings a list of targets every cycle and reports reachability, packet loss and round-trip-time statistics per target. This is the agent's equivalent of the classic PRTG ping sensor, multi-target in a single probe instance.

Quick start

# probes.d/20-ping.yaml
- name: ping-core
  type: icmp_check
  params:
    targets: ["10.0.0.1", "core-switch.lan", "8.8.8.8"]
    interval: 60

Parameters

Parameter	Must set	Default	Description
targets	Yes	-	Hostnames or addresses to ping. Example: 10.0.0.1, gw.example.com
count	No	4	Echo requests per target per cycle
timeout	No	5	Budget in seconds for the whole round on one target
interval	No	60	Seconds between cycles
packet_size	No	56	ICMP payload size in bytes
privileged	No	-	Raw ICMP sockets (true) or datagram sockets (false); default true on Windows and as root on Linux

Targets are pinged in parallel (bounded), so a large list does not stretch the cycle by the sum of timeouts.

Metrics

One series per metric per target (the target tag discriminates; the resolved ip rides along).

Metric	Unit	Description
senhub.icmp.up	bool	1 when at least one reply came back this cycle
senhub.icmp.packet_loss	%	Lost echo requests over the cycle
senhub.icmp.packets.sent / .received	count	Requests/replies this cycle
senhub.icmp.rtt.min / .avg / .max / .stddev	ms	Round-trip statistics (emitted only when at least one reply arrived)

An unreachable target is a measurement (up = 0, loss 100%), not a probe failure: the probe stays healthy and keeps reporting.

Privileges

• Linux: running as root (the current agent requirement), the probe defaults to privileged raw sockets — unprivileged ICMP datagram sockets are gated by net.ipv4.ping_group_range, which stock Ubuntu/Debian servers ship disabled, and root does not bypass that sysctl. As a non-root process the default is unprivileged: if pings fail with a permission error, widen the range to include the agent's group or set privileged: true (requires CAP_NET_RAW). The permission error now carries this hint in the log.
• Windows: raw sockets only — the probe defaults to privileged: true and the agent service runs elevated.
• macOS: unprivileged mode works out of the box.

Metric reference

Every metric this probe can emit. Metric is the OpenTelemetry name the OTLP, Prometheus and Zabbix outputs derive theirs from. Name is what a Nagios check and the API metrics= filter match. PRTG channel is the label PRTG shows, placeholders filled from the series' tags.

Metric	Name	PRTG channel	Unit	Description
senhub.icmp.up	senhub.icmp.up	Ping {target} Reachability	#	1 when the target answered at least one echo request in the last cycle, 0 otherwise
senhub.icmp.packet_loss	senhub.icmp.packet_loss	Ping {target} Packet Loss	%	Percentage of echo requests without a reply in the last cycle
senhub.icmp.packets.sent	senhub.icmp.packets.sent	Ping {target} Packets Sent	#	Echo requests sent in the last cycle
senhub.icmp.packets.received	senhub.icmp.packets.received	Ping {target} Packets Received	#	Echo replies received in the last cycle
senhub.icmp.rtt.min	senhub.icmp.rtt.min	Ping {target} RTT Min	ms	Minimum round-trip time over the last cycle
senhub.icmp.rtt.avg	senhub.icmp.rtt.avg	Ping {target} RTT Avg	ms	Average round-trip time over the last cycle
senhub.icmp.rtt.max	senhub.icmp.rtt.max	Ping {target} RTT Max	ms	Maximum round-trip time over the last cycle
senhub.icmp.rtt.stddev	senhub.icmp.rtt.stddev	Ping {target} RTT StdDev	ms	Round-trip time standard deviation (jitter proxy) over the last cycle