ci: harden CI against Docker Hub registry flakes (retries + auth) (#40700)

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

.github/workflows/check-python-deps.yml

@@ -38,6 +38,19 @@ jobs:
         if: steps.check.outputs.python
         uses: ./.github/actions/setup-backend/
 
+      # Authenticate the Docker daemon so the python:slim pull in
+      # uv-pip-compile.sh uses our (much higher) authenticated rate limit
+      # instead of the shared-runner anonymous one. Best-effort: on fork PRs the
+      # secrets are unavailable, so this no-ops and the pull falls back to
+      # anonymous (covered by the retry loop in the script).
+      - name: Login to Docker Hub
+        if: steps.check.outputs.python
+        continue-on-error: true
+        uses: docker/login-action@650006c6eb7dba73a995cc03b0b2d7f5ca915bee # v4.2.0
+        with:
+          username: ${{ secrets.DOCKERHUB_USER }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+
       - name: Run uv
         if: steps.check.outputs.python
         run: ./scripts/uv-pip-compile.sh

.github/workflows/superset-python-integrationtest.yml

@@ -27,6 +27,11 @@ jobs:
     services:
       mysql:
         image: mysql:8.0
+        # Authenticated pulls use our higher Docker Hub rate limit. Empty on
+        # fork PRs (secrets unavailable) -> runner falls back to anonymous.
+        credentials:
+          username: ${{ secrets.DOCKERHUB_USER }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
         env:
           MYSQL_ROOT_PASSWORD: root
         ports:
@@ -38,6 +43,9 @@ jobs:
           --health-retries=5
       redis:
         image: redis:7-alpine
+        credentials:
+          username: ${{ secrets.DOCKERHUB_USER }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
         options: --entrypoint redis-server
         ports:
           - 16379:6379
@@ -121,6 +129,9 @@ jobs:
     services:
       postgres:
         image: postgres:17-alpine
+        credentials:
+          username: ${{ secrets.DOCKERHUB_USER }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
         env:
           POSTGRES_USER: superset
           POSTGRES_PASSWORD: superset
@@ -130,6 +141,9 @@ jobs:
           - 15432:5432
       redis:
         image: redis:7-alpine
+        credentials:
+          username: ${{ secrets.DOCKERHUB_USER }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
         ports:
           - 16379:6379
     steps:
@@ -186,6 +200,9 @@ jobs:
     services:
       redis:
         image: redis:7-alpine
+        credentials:
+          username: ${{ secrets.DOCKERHUB_USER }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
         ports:
           - 16379:6379
     steps:

scripts/uv-pip-compile.sh

@@ -31,11 +31,32 @@ if [ -z "$RUNNING_IN_DOCKER" ]; then
 
   echo "Running in Docker (Python ${PYTHON_VERSION} on Linux)..."
 
+  IMAGE="python:${PYTHON_VERSION}-slim"
+
+  # Pre-pull the image with a few retries to absorb transient Docker Hub
+  # registry failures ("context deadline exceeded" / anonymous rate-limit blips
+  # on shared CI runners). Without this a flaky pull fails the whole
+  # check-python-deps job on an infrastructure hiccup rather than a real
+  # dependency drift. The pull is in the `until` condition so `set -e` does not
+  # abort on an individual failed attempt.
+  attempt=1
+  max_attempts=4
+  until docker pull "$IMAGE"; do
+    if [ "$attempt" -ge "$max_attempts" ]; then
+      echo "docker pull $IMAGE failed after ${max_attempts} attempts" >&2
+      exit 1
+    fi
+    delay=$((attempt * 10))
+    echo "docker pull $IMAGE failed (attempt ${attempt}/${max_attempts}); retrying in ${delay}s..." >&2
+    sleep "$delay"
+    attempt=$((attempt + 1))
+  done
+
   docker run --rm \
     -v "$(pwd)":/app \
     -w /app \
     -e RUNNING_IN_DOCKER=1 \
-    python:${PYTHON_VERSION}-slim \
+    "$IMAGE" \
     bash -c "pip install uv && ./scripts/uv-pip-compile.sh $*"
 
   exit $?