Fix Node Architecture Mismatch

Signed-off-by: Davanum Srinivas <[email protected]>

Author: dims

tilt/Tiltfile

@@ -72,21 +72,31 @@ helm_resource(
 namespace_create('gpu-operator')
 namespace_create('nvsentinel')
 
+# Detect architecture early for KWOK node creation
+arch = str(local('uname -m')).strip()
+# Normalize architecture names for Kubernetes
+if arch in ['arm64', 'aarch64']:
+    k8s_arch = 'arm64'
+elif arch in ['x86_64', 'amd64']:
+    k8s_arch = 'amd64'
+else:
+    k8s_arch = 'amd64'  # Default to amd64 for unknown architectures
+
 skip_kwok_nodes = os.getenv('SKIP_KWOK_NODES_IN_TILT', '0') == '1'
 
 if not skip_kwok_nodes:
     # Create regular GPU nodes (all NUM_GPU_NODES are regular nodes)
     kwok_node_template = str(read_file('./kwok-node-template.yaml'))
     for i in range(num_gpu_nodes):
-        node_yaml = kwok_node_template.replace('PLACEHOLDER', str(i))
+        node_yaml = kwok_node_template.replace('PLACEHOLDER', str(i)).replace('amd64', k8s_arch)
         k8s_yaml(blob(node_yaml))
 
     # Create separate Kata test nodes (in addition to regular nodes)
     # These are named differently (kwok-kata-test-node-*) to avoid conflicts
     # Set NUM_KATA_TEST_NODES=0 to disable kata testing
     kwok_kata_test_node_template = str(read_file('./kwok-kata-test-node-template.yaml'))
     for i in range(num_kata_test_nodes):
-        node_yaml = kwok_kata_test_node_template.replace('PLACEHOLDER', str(i))
+        node_yaml = kwok_kata_test_node_template.replace('PLACEHOLDER', str(i)).replace('amd64', k8s_arch)
         k8s_yaml(blob(node_yaml))
 else:
     print("Skipping KWOK node creation in Tilt (SKIP_KWOK_NODES_IN_TILT=1)")
@@ -115,8 +125,8 @@ values_files = [
 if use_postgresql:
     values_files.append('../distros/kubernetes/nvsentinel/values-tilt-postgresql.yaml')
 
-arch = str(local('uname -m')).strip()
-if arch in ['arm64', 'aarch64']:
+# Add ARM64-specific values if on ARM64 (arch already detected earlier)
+if k8s_arch == 'arm64':
     values_files.append('../distros/kubernetes/nvsentinel/values-tilt-arm64.yaml')
 
 yaml = helm(

tilt/create-kwok-nodes.sh

@@ -22,15 +22,31 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 NODE_TEMPLATE="$SCRIPT_DIR/kwok-node-template.yaml"
 KATA_NODE_TEMPLATE="$SCRIPT_DIR/kwok-kata-test-node-template.yaml"
 
+# Detect architecture and normalize for Kubernetes
+ARCH=$(uname -m)
+case "$ARCH" in
+    arm64|aarch64)
+        K8S_ARCH="arm64"
+        ;;
+    x86_64|amd64)
+        K8S_ARCH="amd64"
+        ;;
+    *)
+        K8S_ARCH="amd64"  # Default to amd64
+        ;;
+esac
+
+echo "Detected architecture: $ARCH (using k8s arch: $K8S_ARCH)"
+
 echo "Creating $NUM_GPU_NODES regular KWOK GPU nodes..."
 for i in $(seq 0 $((NUM_GPU_NODES - 1))); do
-    sed "s/PLACEHOLDER/$i/g" "$NODE_TEMPLATE" | kubectl apply -f - >/dev/null 2>&1 || true
+    sed "s/PLACEHOLDER/$i/g; s/amd64/$K8S_ARCH/g" "$NODE_TEMPLATE" | kubectl apply -f - >/dev/null 2>&1 || true
 done
 echo "Created $NUM_GPU_NODES regular KWOK GPU nodes"
 
 echo "Creating $NUM_KATA_TEST_NODES KWOK Kata test nodes..."
 for i in $(seq 0 $((NUM_KATA_TEST_NODES - 1))); do
-    sed "s/PLACEHOLDER/$i/g" "$KATA_NODE_TEMPLATE" | kubectl apply -f - >/dev/null 2>&1 || true
+    sed "s/PLACEHOLDER/$i/g; s/amd64/$K8S_ARCH/g" "$KATA_NODE_TEMPLATE" | kubectl apply -f - >/dev/null 2>&1 || true
 done
 echo "Created $NUM_KATA_TEST_NODES KWOK Kata test nodes"