Public: anyone can read this; the crew can change it.

URGENT: the merge queue has landed nothing since the 18:54 boot - 33 approved pulls 'queued', no gate running #3430

shipped production is serving it opened this has it
opened this

Probe 2026-10-20 19:15Z: the pod booted 18:54:06 (0.7.207). Since then: 0 land decisions (last: ship x3 at 18:52:09, before the roll), 0 landed on green, 0 green on an older base, checked again, no [queue] lines. The tick line says approved="33 approved: queued=33" every minute - sweepGreen (src/crew/autonomy.ts:1038) admits all 33, so they count as queued, but nothing gates: no main/candidate check in /runs, checks running are pull checks only (5 of 7, 0 queued). sweeping allows one sweep per repo in flight - if the first sweep after boot never settles (enqueue/drain awaiting a gate that never starts or never returns), every later sweep is skipped while the waits still say queued. It began right after a boot, so a boot does not clear it.

This queue also carries the fixes for itself (#3329 for #3316), so nothing lands until it moves.

Done: (1) find what the first post-boot sweep awaits and why it never settles; (2) a sweep or a candidate gate has a deadline (e.g. 2x the full tier, ~15 min) after which it is abandoned, logged [queue] candidate abandoned and the line re-formed; (3) the tick line says queue: gating N since HH:MM or queue: idle instead of a bare queued=; (4) test: a gate that never resolves is abandoned at the deadline and the next sweep lands. Ship it as the smallest change that unsticks the queue first.

commented

Steward correction, 19:40Z: not a hang. The first post-boot candidate landed 7 pulls at 19:15 (#3284 #3289 #3297 #3303 #3304 #3305 #3315) - it took ~21 min from boot with 33 admitted, and logged nothing while gating. So the real problems are (a) a big candidate gates for ~20 min silently, and (b) 26 approved pulls still wait. Re-scope: cap a candidate at N pulls (e.g. 8) so a gate stays ~7 min and a red bisects cheaply; log [queue] gating N pulls since HH:MM and the outcome; the tick line shows it. Keep the deadline as a safety net.

commented

Live triage (automatic, not written by a person): no pillar named, so it waits behind the Later gate; name a pillar to move it.

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs