Description
OPNsense 26.7.0 removed the is_reserved enrichment from api/kea/leases4/search and api/kea/leases6/search rows (upstream src/opnsense/mvc/app/controllers/OPNsense/Kea/Api/LeasesController.php: present at tag 26.1, gone at 26.7/26.7.3/master, replaced by a stats{active,inactive,total} aggregate and a new delLeaseAction). Our opnsense/kea.go:54 decodes is_reserved and kea.go:197 feeds opnsense_kea_dhcp4/6_leases_reserved_total / dynamic_total, so on a 26.7 box reserved reads 0 and dynamic == total, silently. Verdict shape per the canary triage taxonomy: chase. Replacement source: api/kea/dhcpv4/searchReservation + api/kea/dhcpv6/searchReservation, matched client-side (hw_address/duid), tolerant across the 26.1+26.7 support window. Found by upstream API-surface research 2026-08-30; not yet proven against a live 26.7 box.
Acceptance Criteria
- #1 Bug proven or disproven against a 26.7 leases4/leases6 payload (live box or upstream-derived fixture)
- #2 Reserved/dynamic lease metrics are correct on both 26.1-shape (is_reserved present) and 26.7-shape (reservation search) payloads, resolved by payload shape, not version sniffing
- #3 New reservation endpoints registered per AGENTS.md steps (endpoints map, schema registry, contract manifest, canary exemptions as needed)
- #4 Tests cover both generations; just schemas and just docs clean
Definition of Done
- #1 just check
- #2 just gen (if any generated artifact changed) and the diff committed
Implementation Plan
Wave 1 plan: prove the 26.7 chase from upstream-produced shapes with a failing regression test; implement shape-based reservation matching while preserving the 26.1 is_reserved path; return exact root-owned endpoint/schema/contract wiring; run focused Kea tests and identify required generated artifacts.
Implementation Notes
Wave 1 investigation disproved the task premise against released upstream source. OPNsense core tags 26.7 and 26.7.3 move reservation enrichment into src/opnsense/scripts/kea/get_kea_leases.py, but the script still emits is_reserved on every record: [] for dynamic leases and a non-empty identity array for reserved leases, matched within subnet-id. The exporter already accepts the array form via flexBool and existing v4/v6 tests pass. No synthetic missing-is_reserved fixture or reservation endpoint was added because inspected supported source cannot produce that shape. Live-box payload remains unobserved; source-derived proof is the accepted AC1 route.
Wave 1 result: released OPNsense 26.7 and 26.7.3 source disproved the task premise; get_kea_leases.py still emits is_reserved for every row, and the exporter already accepts the emitted array shape. Existing v4/v6 tests and the integrated just check passed. A local tracker-only commit records this, but it is not on origin/main because the remote advanced during the run. Resume: reconcile the local tracker commit onto current origin/main, land it as documentation-only, verify CI at that exact SHA, and then finalize the disproven task; use a live 26.7 payload only if source-derived proof is no longer accepted.
Closeout reconciliation: source-finding commit 4916dd73e8d195f244a45986275fcc3aa41ba420 is on main. The task remains Parked only because a live 26.7 lease payload was not observed. Resume: decide whether the released 26.7/26.7.3 payload-producing source is sufficient to close the disproved premise; if live proof is required, capture a 26.7 lease response and compare is_reserved rows before changing code. Do not add a synthetic missing-is_reserved fixture or reservation endpoint without a source- or live-produced shape.
Final Summary
Closed as DISPROVED on released-source evidence, decision by Rob 2026-09-02. OPNsense 26.7 and 26.7.3 still emit is_reserved on every lease row from src/opnsense/scripts/kea/get_kea_leases.py ([] for dynamic, a non-empty identity array for reserved), and the exporter’s existing flexBool already accepts that array form. The premise in the description - that 26.7 removed the field - was wrong, so the reserved/dynamic split was never broken. AC1 is disproved from the payload-producing source, which is the standard AGENTS.md sets; AC2 and AC4 are satisfied by the existing v4/v6 tests covering both generations. AC3 is VOID: no reservation endpoint is needed, and adding searchReservation would have modelled a shape upstream does not produce, which is recurring defect class 1. Source finding recorded in 4916dd73. No live 26.7 capture was taken and none is required. IMPORTANT for OPN-0015: it was filed on this same disproved premise, so its scope must be re-derived from source before it starts.