usdrecord keeps waiting after its Arras render worker exits

Aperta
  • Qualcosa non va?
#309 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
4/5
Tempo stimato
3-5 giorni
Idoneità per principianti
53/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
cpp
Ambito
backend

Direzione di ricerca

Start by reading ArrasRenderer::statusHandler and the frame completion logic around isFrameComplete() in hdMoonray, then trace how usdrecord waits for convergence after the Arras session stops. The control probe and logs linked in the issue reproduce the hang by killing execComp during rendering. Done means usdrecord reports the failed render and stops waiting after Arras reports a stopped session, allowing a batch to continue.

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

Descrizione

During a benchmark run with the MoonRay Hydra delegate in OpenMoonRay v2026.29.1 and OpenUSD 24.08 on Rocky Linux 8.10.
The Moana Island worker exited with signal 9. Arras reported the failure, but usdrecord stayed alive without its render worker for about 2 hours 46 minutes, until the benchmark's own deadline stopped it. No Moana image was produced. I don't have evidence establishing why the worker died in this run.

I reproduced this behavior separately on OpenUSD 23.08, using McUsd under Xvfb/Mesa and deliberately killing the Arras worker (execComp). No OOM or GPU was needed:

Error: Arras session stopped, compExited: mcrt exited due to signal 9
RESULT: CLIENT STILL ALIVE 150s after its render worker died

Control probe and logs. The probe script starts usdrecord on McUsd, waits for the worker, kills it while the frame is still rendering, and observes the client.

In hdMoonray, ArrasRenderer::statusHandler logs stopped/stopping and sets the connection state to disconnected, but leaves the unfinished frame waiting for completion. Without a final image, isFrameComplete() remains false and usdrecord keeps waiting for convergence.

I would expect usdrecord to stop waiting and report the failed render when Arras reports that the worker has stopped. One failed render should not block the rest of a batch until an external timeout.

The reason the worker stopped and usdrecord's handling of an aborted recording are separate questions. This report is about the client continuing to wait after Arras has already reported the stopped session.

Benchmark issue and investigation.

Lingua principale
CMake
Stelle
4.7k
Fork
321
Merge medio
13m
PR unite (30g)
3

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 OpenMoonRay/openmoonray

Tutte le issue di OpenMoonRay/openmoonray

Issue simili

Altre issue su Backend & API Design

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.