nginx startup stuck in loop: curl returns 000 resolving 'api_server' on Alpine (wget works)

[Opeple0627]

Description

After starting Onyx with Docker Compose, nginx remains stuck in its startup health-check loop indefinitely, making localhost:3000 unreachable:

Waiting for API server to boot up; this may take a minute or two...
API server responded with 000, retrying in 5 seconds...
API server responded with 000, retrying in 5 seconds...
(repeats forever)

Root Cause

The nginx startup script (run-nginx.sh) uses curl to health-check the API server:

status_code=$(curl -o /dev/null -s -w "%{http_code}\n" "http://${ONYX_BACKEND_API_HOST}:8080/health")

On Alpine Linux, curl fails to resolve api_server via Docker's internal DNS (127.0.0.11), always returning 000 (connection failure). However, wget resolves the same hostname correctly:

# Inside nginx container:
curl -s http://api_server:8080/health   # → 000 (fails)
wget -qO- http://api_server:8080/health # → {"success":true} (works)

The issue appears to be a DNS resolution incompatibility between Alpine's musl-based curl and Docker's embedded DNS resolver (127.0.0.11) with search localdomain in /etc/resolv.conf.

Steps to Reproduce

docker compose --profile s3-filestore up -d
docker logs onyx-nginx-1 --follow
# Observe "API server responded with 000" repeating indefinitely

Workaround

Manually add the API server IP to /etc/hosts inside the nginx container:

# Get api_server IP
API_IP=$(docker inspect onyx-api_server-1 --format '{{range .NetworkSettings.Networks}}{{.IPAddress}}{{end}}')
docker exec onyx-nginx-1 sh -c "echo '$API_IP api_server' >> /etc/hosts"

Suggested Fix

Replace curl with wget in the health-check loop in run-nginx.sh, or use the IP address directly, or switch to a curl invocation that bypasses DNS (e.g. --resolve).

Environment

• OS: Ubuntu 24.04.4 (VMware)
• nginx image: nginx:1.25.5-alpine
• Docker version: latest