Hello Vercel Team,
We need technical assistance with stuck messages in Vercel Queues for our production project moj-novi-dom.
Queue topic: mnd-listing-translation
Problem:
- Old translation messages are being redelivered repeatedly, some exceeding 120 delivery attempts.
- The old deployment uses the deprecated OpenAI model
gpt-4.1-nano, causing HTTP 404 errors. - We deployed a fix using
gpt-4.1-miniand added handling for permanent errors and excessive retries. - However, the old messages appear to remain pinned to the original deployment and continue executing the outdated code.
- We have observed retry counts exceeding 120, and some messages appear to start again at low delivery counts.
Old deployment ID: dpl_FWTzsygmgaW2yJAJhJNKHgbZS5cu
Database status (last checked):
- COMPLETED: 49
- FAILED: 32
- PAUSED: 63
- No PENDING or PROCESSING jobs.
Request:
Could a Vercel engineer investigate and help us selectively stop, acknowledge, or purge only the stuck messages associated with the old deployment?
Please do not delete the current production deployment, database, translation cache, or unrelated queues.
We would also appreciate clarification about message retention and why these retries continue despite the configured 24-hour retention period.
The repeated retries are causing unnecessary function executions and potentially additional charges.
Thank you for your assistance.