update

Signed-off-by: Junpu Fan <junpu@amazon.com>

Author: junpuf

.github/scripts/cleanup_old_image.sh

@@ -1,33 +1,41 @@
 #!/usr/bin/env bash
-set -u  # only fail on unset vars
+set -euo pipefail
 
+# configurable cutoff
 CUTOFF_HOURS=${CUTOFF_HOURS:-24}
-# always compute cutoff in UTC for comparison
-CUTOFF_TS=$(date -u -d "${CUTOFF_HOURS} hours ago" +%s 2>/dev/null || \
-            date -u -v-"${CUTOFF_HOURS}"H +%s)
+CUTOFF_TS=$(date -u -d "${CUTOFF_HOURS} hours ago" +%s 2>/dev/null || date -u -v-"${CUTOFF_HOURS}"H +%s)
 
 echo "=== Docker disk usage before cleanup ==="
-docker system df -v || echo "(warning: docker system df failed)"
+docker system df -v || true
 echo
 
-echo "=== Checking images older than ${CUTOFF_HOURS}h (UTC-based) ==="
-
 deleted=0
 kept=0
 
-while IFS= read -r line; do
-  id=$(awk '{print $1}' <<<"$line")
-  name=$(awk '{print $2}' <<<"$line")
-  created_at=$(awk '{$1=$2=""; print substr($0,3)}' <<<"$line")
+echo "=== Checking images older than ${CUTOFF_HOURS}h (UTC) ==="
+
+docker images --format '{{json .}}' | while read -r json; do
+  id=$(jq -r '.ID' <<<"$json")
+  repo=$(jq -r '.Repository' <<<"$json")
+  tag=$(jq -r '.Tag' <<<"$json")
+  created_at=$(jq -r '.CreatedAt' <<<"$json")
 
+  # Skip empty or invalid
   [ -z "$id" ] && continue
-  [ "$name" = "<none>:<none>" ] && continue
 
-  # parse docker UTC timestamp safely
-  created_ts=$(date -u -d "$created_at" +%s 2>/dev/null || echo 0)
+  # Normalize name
+  name="${repo}:${tag}"
+
+  # Convert CreatedAt → epoch (cross-platform)
+  if date --version >/dev/null 2>&1; then
+    created_ts=$(date -u -d "$created_at" +%s)
+  else
+    created_ts=$(date -u -j -f "%Y-%m-%d %H:%M:%S %z %Z" "$created_at" +%s)
+  fi
 
+  # Compare
   if [ "$created_ts" -lt "$CUTOFF_TS" ]; then
-    echo "🗑️  Removing old image: $name (created $created_at UTC)"
+    echo "🗑️  Removing old image: $name (created $created_at)"
     if docker rmi -f "$id" >/dev/null 2>&1; then
       deleted=$((deleted+1))
     else
@@ -36,7 +44,7 @@ while IFS= read -r line; do
   else
     kept=$((kept+1))
   fi
-done < <(docker images --format '{{.ID}} {{.Repository}}:{{.Tag}} {{.CreatedAt}}' 2>/dev/null || true)
+done
 
 echo
 echo "=== Cleanup summary ==="
@@ -45,8 +53,8 @@ echo "Images deleted: $deleted"
 echo
 
 echo "=== Docker disk usage after cleanup ==="
-docker system df -v || echo "(warning: docker system df failed)"
+docker system df -v || true
 
 echo
-echo "=== Host disk space (for /var/lib/docker) ==="
+echo "=== Disk space for /var/lib/docker ==="
 df -h /var/lib/docker 2>/dev/null || df -h /

.github/workflows/pr-vllm-rayserve.yml

@@ -118,12 +118,15 @@ jobs:
             set -eux
             uv pip install --system -r requirements/common.txt -r requirements/dev.txt --torch-backend=auto
             uv pip install --system pytest pytest-asyncio
+            uv pip install --system -e tests/vllm_test_utils
+            uv pip install --system hf_transfer
           '
 
       - name: Run vLLM Tests
         run: |
           docker exec ${CONTAINER_ID} sh -c '
           set -eux
+          HF_HUB_ENABLE_HF_TRANSFER=1
           nvidia-smi          
           pytest -s -v tests/test_logger.py
           # Entrypoints Integration Test (LLM) # 30min