Runway DLQ answers FAILED for a merge whose push already landed

Aperta
  • Qualcosa non va?
#820 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 2 giorni

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
55/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
go
Ambito
distributed-systems

Direzione di ricerca

Start with runway/controller/dlq/dlq.go and the merger’s existing branch-containment check used for idempotence; the issue does not name that code’s file. Add a unit test for a successful push followed by failed answer publishing until dead-lettering. Done means the DLQ answers MERGED when the request’s changes are already on the target branch and never reports that landed request as FAILED.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

bug

Problem

Runway can tell SubmitQueue that a merge failed after the merge has landed on the target branch. landsignal applies the FAILED, and the batch and its requests are reported failed for a change that is on main.

Cause

Runway's merge DLQ (runway/controller/dlq/dlq.go) answers FAILED unconditionally; Runway is stateless and never checks whether the push happened. The merge controller pushes first and publishes its answer afterwards:

runway merge:  push lands on the target branch
runway merge:  publishing SUCCEEDED keeps failing; retries re-run the merge (idempotent: no new commits)
               → retries exhaust → runway-merge DLQ
runway DLQ:    publish FAILED under a distinct message ID, so it is not deduped
landsignal:    batch landing → failed

The git merger's redelivery idempotence covers the merge controller's retries. Nothing covers the DLQ's answer. No orchestrator-side fix (#819) can repair this.

Direction

Check the branch before answering: answer MERGED when the request's changes are already contained in the target ref, FAILED otherwise. The merger already computes this for idempotence. Note the residual hazard recorded in #819: once the DLQ has honestly answered FAILED, a later re-send of the same request must not merge.

Found by model checking spec/submitqueue/landoutcome/LandOutcome.tla.

Acceptance criteria

  • Runway's DLQ never answers FAILED for a request whose changes are on the target branch.
  • Unit test: the push succeeds, the answer publish fails until dead-lettered, and the DLQ answers MERGED.
Lingua principale
Go
Stelle
247
Fork
16
Merge medio
2g 1h
PR unite (30g)
125

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di uber/submitqueue

Tutte le issue di uber/submitqueue

Issue simili

Altre issue su Go

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.