PROBLEM DESCRIPTION

Docker build on Railway is stuck indefinitely at:

[internal] load metadata for docker.io/library/python:3.9-slim
[internal] load metadata for docker.io/library/node:20-alpine

After ~30 minutes, the build times out with:

Build Failed: build daemon returned an error < failed to run Build function: Canceled: context canceled >

ROOT CAUSE

Railway's Docker build environment has slow/blocked access to Docker Hub.

The build cannot fetch metadata for base images from Docker Hub within the timeout window.

This is NOT a code issue - it's an infrastructure/network issue with Railway's build servers.

WHAT WE ALREADY TRIED

Attempt 1: Blamed SHAP Library

What we did: Removed shap==0.43.0 from

backend-requirements.txt

Why it failed: The build was stuck BEFORE pip runs, so SHAP wasn't the issue.

Result: Didn't help - build still timed out at Docker metadata fetch.

Attempt 2: Added PyPI Mirror Optimization

What we did: Added Aliyun PyPI mirror to Dockerfile:

dockerfile

RUN mkdir -p ~/.pip && echo "[global]\nindex-url = https://mirrors.aliyun.com/pypi/simple/" > ~/.pip/pip.conf

Why it failed: The build never gets to pip install - it's stuck at base image pull.

Result: Didn't help - build still times out at Docker metadata fetch.

Attempt 3: Added BuildKit Optimization

What we did: Added BuildKit syntax to Dockerfile:

dockerfile

# syntax=docker/dockerfile:1.4

Why it failed: BuildKit helps with caching, but doesn't solve the base image pull timeout.

Result: Didn't help - build still times out at Docker metadata fetch.

Attempt 4: Verified .dockerignore

What we did: Checked

.dockerignore file (already optimized)

Result: File is good, but doesn't solve the base image pull issue.