Enforce HTTP 200 requirement for MCP gateway /health endpoint with 120s curl retry

actions/setup/sh/start_mcp_gateway.sh

@@ -141,36 +141,30 @@ echo "Waiting for gateway to be ready..."
 HEALTH_CHECK_HOST="localhost"
 echo "Health endpoint: http://${HEALTH_CHECK_HOST}:${MCP_GATEWAY_PORT}/health"
 echo "(Note: MCP_GATEWAY_DOMAIN is '${MCP_GATEWAY_DOMAIN}' for container access)"
-MAX_ATTEMPTS=30
-ATTEMPT=0
-while [ $ATTEMPT -lt $MAX_ATTEMPTS ]; do
-  # First check if the gateway process is still running
-  if ! ps -p $GATEWAY_PID > /dev/null 2>&1; then
-    echo "ERROR: Gateway process (PID: $GATEWAY_PID) has exited unexpectedly!"
-    echo ""
-    echo "Gateway stdout output:"
-    cat /tmp/gh-aw/mcp-config/gateway-output.json 2>/dev/null || echo "No stdout output available"
-    echo ""
-    echo "Gateway stderr logs:"
-    cat /tmp/gh-aw/mcp-logs/stderr.log 2>/dev/null || echo "No stderr logs available"
-    exit 1
-  fi
-
-  # Check health endpoint using localhost (since we're running on the host)
-  HEALTH_RESPONSE=$(curl -f -s "http://${HEALTH_CHECK_HOST}:${MCP_GATEWAY_PORT}/health" 2>&1) && {
-    echo "Gateway is ready!"
-    echo "Health response: $HEALTH_RESPONSE"
-    break
-  }
-  ATTEMPT=$((ATTEMPT + 1))
-  if [ $ATTEMPT -lt $MAX_ATTEMPTS ]; then
-    echo "Attempt $ATTEMPT/$MAX_ATTEMPTS: Gateway not ready yet (curl response: $HEALTH_RESPONSE), waiting 1 second..."
-    sleep 1
-  fi
-done
+echo "Retrying up to 120 times with 1s delay (120s total timeout)"
+
+# Check health endpoint using localhost (since we're running on the host)
+# Per MCP Gateway Specification v1.3.0, /health must return HTTP 200 with JSON body containing specVersion and gatewayVersion
+# Use curl retry: 120 attempts with 1 second delay = 120s total
+RESPONSE=$(curl -s --retry 120 --retry-delay 1 --retry-connrefused --retry-all-errors -w "\n%{http_code}" "http://${HEALTH_CHECK_HOST}:${MCP_GATEWAY_PORT}/health" 2>&1)
+HTTP_CODE=$(echo "$RESPONSE" | tail -n 1)
+HEALTH_RESPONSE=$(echo "$RESPONSE" | head -n -1)
+
+# Always log the health response for debugging
+echo "Health endpoint HTTP code: $HTTP_CODE"
+if [ -n "$HEALTH_RESPONSE" ]; then
+  echo "Health response body: $HEALTH_RESPONSE"
+else
+  echo "Health response body: (empty)"
+fi
 
-if [ $ATTEMPT -eq $MAX_ATTEMPTS ]; then
-  echo "ERROR: Gateway failed to become ready after $MAX_ATTEMPTS attempts"
+if [ "$HTTP_CODE" = "200" ] && [ -n "$HEALTH_RESPONSE" ]; then
+  echo "Gateway is ready!"
+else
+  echo ""
+  echo "ERROR: Gateway failed to become ready"
+  echo "Last HTTP code: $HTTP_CODE"
+  echo "Last health response: ${HEALTH_RESPONSE:-(empty)}"
   echo ""
   echo "Checking if gateway process is still alive..."
   if ps -p $GATEWAY_PID > /dev/null 2>&1; then

actions/setup/sh/verify_mcp_gateway_health.sh

@@ -47,40 +47,39 @@ echo ''
 
 # Wait for gateway to be ready FIRST before checking config
 echo '=== Testing Gateway Health ==='
-max_retries=30
-retry_count=0
-gateway_ready=false
-health_response=""
-
-while [ $retry_count -lt $max_retries ]; do
-  # Capture both response body and HTTP code in a single curl call
-  response=$(curl -s -w "\n%{http_code}" "${gateway_url}/health")
-  http_code=$(echo "$response" | tail -n 1)
-  health_response=$(echo "$response" | head -n -1)
-
-  if echo "$http_code" | grep -q "200\|204"; then
-    echo "✓ MCP Gateway is ready!"
-    gateway_ready=true
-    break
-  fi
-  retry_count=$((retry_count + 1))
-  echo "Waiting for gateway... (attempt $retry_count/$max_retries)"
-  sleep 1
-done
 
-if [ "$gateway_ready" = false ]; then
-  echo "✗ Error: MCP Gateway failed to start after $max_retries attempts"
+# Capture both response body and HTTP code in a single curl call
+# Use curl retry: 120 attempts with 1 second delay = 120s total
+echo "Calling health endpoint: ${gateway_url}/health"
+echo "Retrying up to 120 times with 1s delay (120s total timeout)"
+response=$(curl -s --retry 120 --retry-delay 1 --retry-connrefused --retry-all-errors -w "\n%{http_code}" "${gateway_url}/health")
+http_code=$(echo "$response" | tail -n 1)
+health_response=$(echo "$response" | head -n -1)
+
+# Always log the health response for debugging
+echo "Health endpoint HTTP code: $http_code"
+if [ -n "$health_response" ]; then
+  echo "Health response body: $health_response"
+else
+  echo "Health response body: (empty)"
+fi
+
+if [ "$http_code" = "200" ]; then
+  echo "✓ MCP Gateway is ready!"
+else
+  echo ''
+  echo "✗ Error: MCP Gateway failed to start"
+  echo "Last HTTP code: $http_code"
+  echo "Last health response: ${health_response:-(empty)}"
   echo ''
   echo '=== Gateway Logs (Full) ==='
   cat "${logs_folder}/gateway.log" || echo 'No gateway logs found'
   exit 1
 fi
 
 # Parse and display version information from health response
+echo ''
 if [ -n "$health_response" ]; then
-  echo "Health response: $health_response"
-  echo ''
-
   # Extract version information using jq if available
   if command -v jq >/dev/null 2>&1; then
     spec_version=$(echo "$health_response" | jq -r '.specVersion // "unknown"')