Metrics
run and daemon serve Prometheus metrics at http://127.0.0.1:2112/metrics unless metrics.enabled is false. run --once serves none. A port that cannot be opened is a warning, never a reason to stop sweeping.
The endpoint listens on loopback. To scrape it from another machine, set metrics.listen: "0.0.0.0:2112" and restrict it with your firewall: it is unauthenticated.
curl -s http://127.0.0.1:2112/metrics | grep ^infrality_What is recorded
Section titled “What is recorded”| Metric | What it records |
|---|---|
infrality_scan_duration_seconds |
The last sweep, survey included |
infrality_scans_total{outcome} |
complete, partial (reached scan.timeout), cancelled |
infrality_hosts_discovered_total |
Devices filed — the same count as the board, router included |
infrality_hosts_by_discovery_total{method} |
icmp, arp, route, assumed; a host found both ways counts under each |
infrality_identification_total{source} |
Which protocol’s name won |
infrality_hosts_unnamed_total |
Devices nothing named — whether identification works on this network |
infrality_icmp_rtt_milliseconds |
Ping round trip, histogram |
infrality_icmp_no_answer_total |
Addresses that did not answer |
infrality_icmp_errors_total |
ICMP failures — a rising count usually means a missing CAP_NET_RAW |
infrality_active_workers |
Workers busy now |
infrality_tasks_processed_total{task} |
Work done, per task |
Plus the standard go_* and process_* series.
No metric carries an address. The page lists counts, never the hosts that answered.
Useful alerts
Section titled “Useful alerts”increase(infrality_scans_total{outcome="partial"}[1d]) > 0— sweeps are hitting their time limit; raise it.- A high
infrality_hosts_unnamed_totalagainstinfrality_hosts_discovered_total— the network gives the agent little to name hosts by; see naming hosts. infrality_icmp_errors_totalrising — the agent lost its capability.