Skip to content

release: v1.0.0-alpha.24 - #58

Merged
cativo23 merged 5 commits into
mainfrom
release/v1.0.0-alpha.24
Sep 29, 2026
Merged

cativo23 merged 5 commits into
mainfrom
release/v1.0.0-alpha.24

Conversation

@cativo23

Copy link
Copy Markdown
Owner

Fixed

  • BullMQ queue collision between local dev and production (shared Redis, no prefix). See CHANGELOG.

Deploy notes

Fourth same-day hotfix, following v1.0.0-alpha.20/21/22/23.

Local dev and production both point at the same Upstash Redis instance
(same REDIS_HOST/TOKEN, copied from local .env into the server's for
convenience during launch verification), and both construct the
"analysis-queue" BullMQ queue/worker with no prefix — so they collide
on the exact same Redis keys. A 3-day-old local `nuxt dev` process
(which runs the worker in-process, since DISABLE_WORKER is unset there)
was racing production's dedicated worker container for the same jobs.

Discovered live during plan 13-04's LAUNCH-01 verification: BullMQ's
own job history showed both test jobs completed successfully, but the
production worker container's logs never mentioned them at all — some
other consumer claimed and failed them before production's worker got
a chance.

Adds a `prefix` (BullMQ's built-in Redis key namespacing) derived from
NODE_ENV: "prod" in production, "dev" everywhere else. Both the Queue
(server/utils/queue.ts) and the Worker (server/plugins/worker.ts) must
use the same prefix to see each other's jobs, so it's exported as a
single shared constant rather than duplicated.
…ronment

fix(worker): namespace BullMQ queue per environment (prod vs dev)
@cativo23
cativo23 merged commit 75fbe2a into main Sep 29, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant