| name | dhcp-fault-skill |
| description | Identify DHCP server-side faults — service down, missing subnet declaration, spoofed gateway or DNS pushed by the server. Use when one or more hosts cannot obtain a lease, or hosts received DHCP-supplied values that look wrong. |
DHCP Server-Side Faults
Direct-evidence requirement
The agent must show direct evidence on the DHCP server before submitting any DHCP fault. Hosts having no IP is downstream — it does NOT prove the DHCP server is the cause.
Rule out upstream cascade first
A broken path to the DHCP server looks identical to a server-side fault from the affected host's perspective. Before submitting any DHCP fault, confirm the upstream isn't the real cause: the routing / control plane between the affected host and the DHCP server must be healthy, no ACL on the relay path may be dropping lease traffic, and the server itself must be reachable from the host's access gateway or relay. If any of those is broken, the upstream wins.
A host having no IP, an expired lease, or a short valid_lft is NOT direct evidence of a DHCP fault by itself — only a dead daemon or a wrong server configuration is.
Submit on match
When direct evidence on the DHCP server is sufficient AND the upstream cascade has been ruled out, submit. Do not chain extra probes once the family is confirmed.
Disambiguation
- DHCP-spoofed vs host-local misconfig. If the host's wrong gateway/DNS matches what the DHCP server is pushing, this is a DHCP fault and faulty_devices includes both. If the DHCP server's config is correct but the host's
/etc/resolv.conf or default route differs from what was pushed, that's host-ip-skill territory.
- dhcp_service_down vs no-route-to-DHCP. A host without IP plus a DHCP server whose process is alive = upstream cascade, not
dhcp_service_down. The cascade guard above must pass first.
systemctl inactive is not enough. systemctl is-active may report inactive for services that are actually running under a different init pattern. Combine with pgrep (or equivalent process check) to confirm the daemon is actually absent before naming a service-down fault.dhcp_missing_subnet vs dhcp_spoofed_subnet. Both produce the same direct evidence (a subnet block missing from dhcpd.conf for an affected host's subnet). The label follows the scenario's category: dhcp_missing_subnet under misconfiguration scenarios, dhcp_spoofed_subnet under network-attack scenarios. Check list_avail_problems() to see which name is valid for the current scenario before submitting.
