FastAPI Python 3.13 FUNCTION_INVOCATION_FAILED with no runtime logs

Hi,

I’m debugging a FastAPI Preview deployment that is successfully built and marked Ready, but the Python function fails during invocation before any runtime/application logs become visible.

Deployment ID: dpl_7A7TrCZhSWGd4ryf1eSntFNhsNQA Request ID: gru1::8mqgk-1790208339168-d6585638b66e Request: GET /health HTTP status: 500 X-Vercel-Error: FUNCTION_INVOCATION_FAILED Runtime: python3.13

The cloud build completed successfully and Vercel selected Python 3.13 from .python-version.

We reproduced the failure while vercel logs --follow --json was already active. The single GET /health returned HTTP 500 again, but the live log stream produced 0 runtime events and no traceback.

We also queried the existing deployment logs afterward and found no runtime traceback for the failed invocation.

Vercel’s preliminary support response indicated that this behavior is consistent with a failure during cold-start initialization/module evaluation, before the ASGI application and user-visible logging are initialized. However, the actual underlying exception is still unknown.

Could a Vercel team member please check the internal platform telemetry for the Request ID above and provide, if available:

the underlying Python exception or container exit reason; the internal traceback associated with this invocation; whether the failure occurs while importing/evaluating the FastAPI entrypoint, before the ASGI app starts; why no runtime log event is emitted for this failed invocation.

We have intentionally not redeployed or modified the source/environment because we want to identify the actual failure before making changes.

Thanks.

Update: local CPython 3.13 reproduction passes

I completed a controlled local reproduction using the same R6 application source and its exact pinned dependency contract.

Local environment

CPython 3.13.12 Linux x86_64 exact 52 pinned dependencies from the deployment requirements.txt 52/52 dependencies installed successfully pip check: PASS

Cold-start import test

python -X faulthandler -c "import app"

Result:

return code: 0 terminated by signal: no traceback: none stdout: 0 bytes stderr: 0 bytes

So the application entrypoint and its import-time dependency chain complete successfully in this controlled Linux + CPython 3.13 environment.

However, the same project in Vercel cloud still behaves as follows:

Deployment ID: dpl_7A7TrCZhSWGd4ryf1eSntFNhsNQA Request ID: gru1::8mqgk-1790208339168-d6585638b66e GET /health → HTTP 500 X-Vercel-Error: FUNCTION_INVOCATION_FAILED Cloud runtime: Python 3.13 Visible runtime log events: 0 No traceback is exposed in either historical or live runtime logs.

This means we are currently unable to reproduce the cold-start failure locally with the same source and pinned dependency set.

Could you please check the internal platform telemetry for the Request ID above and provide the underlying container/process error, exit reason, or traceback, especially anything occurring before user-visible runtime logging is initialized?

We are intentionally not redeploying or modifying source, environment variables, or Firebase until the underlying cloud-side failure is identified.

Thanks.