applog-listen.ts self-heal: aro_join(app) before retry on origin_aro_not_member (option A)
Follow-up hardening from MARS-431. applog-listen.ts:242-250 on-fail path (hub 400 origin_aro_not_member) re-registers + retries once but never aro_joins, so it loops/misses paging until membership is externally restored. Scoped by pm-llmmsgsrv-cc (their boundary is hub-side only; this is app-side producer code, venus/scripts/applog-listen.ts, EVO-28, shared by all 4 app instances): implement option A (aro_join(app) before retry) — do NOT drop origin_aro/DM-to-PM (option B), that downgrades off the god-view rail. Patch once, then applog-listen-restart-all.sh restarts all 4 instances. Give pm-venus-cc/pm-pluto-cc/pm-ayudarg-cc a heads-up before/at patch time since it touches their instances. Priority LOW / defense-in-depth per pm-llmmsgsrv-cc — MSG-267 (hub-side prune-exemption) already mitigates the trigger; this is belt-and-suspenders for any future prune, not urgent.
Questions
Activity
-
Landed venus 39aeb07/v0.95.13 (coder-venus-cc pushed), all 4 applog-listen instances restarted clean by coder-mars-cc. Live-validated: aro_join-before-retry works — ayudarg's sender was ALSO independently pruned (2nd blind app), self-healed in one cycle on catch-up (07:58:07), vs the old infinite 30s-loop. Fix restored ayudarg alerting as a side effect. Venus-only heartbeat bug found during verify routed directly to coder-venus-cc, out of Mars scope.