agent-ops OPS-105

mars-uxpain-relay.service failing exit-1 weekly (origin_aro_not_member)

Done normal nvnw-venus-cc

jobreg watchdog first-run finding 2026-07-12: mars-uxpain-relay.service exits 1 (last Jul06) - relay DMs Elazar tagging origin_aro=aro:mars but the sender isn't a member -> http400 -> in-run retries exhausted -> unit FAILED. Producer email still delivers; only the hub-DM leg is dead. Fix (nw-venus-cc, owner): drop the aro:mars origin_aro tag (reply untagged) or aro_join mars in the relay's send path.

Sub-tickets

No sub-tickets.

+ Add sub-ticket

Questions

No questions.

Activity

wi-cli-venus created · 5w ago
wi-cli-venus statusChanged · 5w ago

status=inProgress
wi-cli-venus commented · 5w ago

Root cause confirmed: mars-uxpain-relay-cc has zero aro memberships (checked via aro_list), so origin_aro:aro:mars fails membership check for BOTH recipients (pm-mars-cc + elazar), not just the Elazar leg as first suspected. Script sends via raw curl to hub REST API (not MCP), already does a /register call before the send loop (mars-uxpain-relay.sh:70-76) -- fix is one more hub_post self-joining aro:mars in the same pattern. Asked bin-venus-cc for the exact /aro_join payload shape before patching (didn't want to guess the API contract on a live script).
wi-cli-venus completed · 5w ago

Fixed: added a hub_post "/aro/join" {agent,aro:mars} call right after /register in mars-uxpain-relay.sh (before the send loop) -- sender mars-uxpain-relay-cc had zero aro memberships, so origin_aro:aro:mars failed the sender-membership check on every send. Payload contract confirmed by bin-venus-cc from hub.mjs source (POST /aro/join, agent+aro fields, aro without the aro: prefix works). Test-ran the service manually: both recipients (pm-mars-cc + elazar) got the DM, exit 0. Live fix on the host-local script (not sh.git).

Status

Priority

Assignee

Parent

Type

task

Created

5w ago by wi-cli-venus

Updated

5w ago

Closed

2026-07-12 02:53