#604 ·
llmmsg-srv · child of #531Fleet Health false-green: /fleet_health reports host green while host-dark (hub.mjs:868-878 probe-age tripwires too lax + agents:[] reads as healthy)
- Ref
#604(#604)- Project
llmmsg-srv- Parent
- backlog #531 Transport reliability + observability program (whey/venus/lezama)
- Status
- done
- Priority
- normal
- Type
- bug
- Assigned
- hub-llmmsgsrv-cc
- Created by
- —
- Created
- 2026-06-02T00:38:22.215Z
- Updated
- 2026-06-02T00:43:36.954Z
- Closed
- 2026-06-02T00:43:36.954Z
Questions
No questions.
Event log
-
wi cli; parent=#531
-
Root cause hub.mjs:868-878: status starts green, bumped only by agent-silence worstFactor then probe-age (>14400s yellow / >86400s red). lezama agents:[] (pruned at STALE_TTL 600s)->worstFactor 0 AND probeAge 12767s/3.5h < 14400 -> stayed GREEN while host-dark since ~18:03. FIX: (1) env-tunable probe-age yellow>LLMMSG_FLEET_PROBE_YELLOW_S(3600) / red>LLMMSG_FLEET_PROBE_RED_S(10800); (2) agents.length===0 + probeAge>yellow forces >=yellow. Keep no-probe->red. Bump hub VERSION, no schema change. Full spec DM'd to hub-llmmsgsrv-cc. Surfaced by nw-whey-cc + support-llmmsgsrv-cc from Elazar.
-
Shipped hub v2.9.16 (commits deabd7c + e072f45). env-tunable FLEET_PROBE_YELLOW_S=3600/_RED_S=10800 + zero-agent-aging-probe guard. Verified live: lezama now RED (was false-green); pre-ship test 30m=green/90m=yellow/4h=red all pass. By hub-llmmsgsrv-cc.