Merge pull request #1360 from torredil/fargate-affinity

Add controller nodeAffinity to prefer EC2 over Fargate + Retrieve region/AZ from topology label in K8s

charts/aws-ebs-csi-driver/values.yaml

@@ -110,7 +110,16 @@ fullnameOverride:
 controller:
   # If arbitrary args like "--aws-sdk-debug-log=true" need to be passed, use this value
   additionalArgs: []
-  affinity: {}
+  affinity:
+    nodeAffinity:
+      preferredDuringSchedulingIgnoredDuringExecution:
+      - weight: 1
+        preference:
+          matchExpressions:
+          - key: eks.amazonaws.com/compute-type
+            operator: NotIn
+            values:
+            - fargate
   # The default filesystem type of the volume to provision when fstype is unspecified in the StorageClass.
   # If the default is not set and fstype is unset in the StorageClass, then no fstype will be set
   defaultFsType: ext4

deploy/kubernetes/base/controller.yaml

@@ -23,6 +23,16 @@ spec:
         kubernetes.io/os: linux
       serviceAccountName: ebs-csi-controller-sa
       priorityClassName: system-cluster-critical
+      affinity:
+        nodeAffinity:
+          preferredDuringSchedulingIgnoredDuringExecution:
+          - preference:
+              matchExpressions:
+              - key: eks.amazonaws.com/compute-type
+                operator: NotIn
+                values:
+                - fargate
+            weight: 1
       tolerations:
         - key: CriticalAddonsOnly
           operator: Exists

pkg/cloud/metadata_k8s.go

@@ -45,31 +45,33 @@ func KubernetesAPIInstanceInfo(clientset kubernetes.Interface) (*Metadata, error
 		return nil, fmt.Errorf("node providerID empty, cannot parse")
 	}
 
-	awsRegionRegex := "(snow)|(([a-z]{2}(-gov)?)-(central|(north|south)?(east|west)?)-[0-9])"
-	awsAvailabilityZoneRegex := "(snow)|(([a-z]{2}(-gov)?)-(central|(north|south)?(east|west)?)-[0-9][a-z])"
 	awsInstanceIDRegex := "s\\.i-[a-z0-9]+|i-[a-z0-9]+$"
 
-	re := regexp.MustCompile(awsRegionRegex)
-	region := re.FindString(providerID)
-	if region == "" {
-		return nil, fmt.Errorf("did not find aws region in node providerID string")
-	}
-
-	re = regexp.MustCompile(awsAvailabilityZoneRegex)
-	availabilityZone := re.FindString(providerID)
-	if availabilityZone == "" {
-		return nil, fmt.Errorf("did not find aws availability zone in node providerID string")
-	}
-
-	re = regexp.MustCompile(awsInstanceIDRegex)
+	re := regexp.MustCompile(awsInstanceIDRegex)
 	instanceID := re.FindString(providerID)
 	if instanceID == "" {
 		return nil, fmt.Errorf("did not find aws instance ID in node providerID string")
 	}
 
 	var instanceType string
-	if it, ok := node.GetLabels()[corev1.LabelInstanceTypeStable]; ok {
-		instanceType = it
+	if val, ok := node.GetLabels()[corev1.LabelInstanceTypeStable]; ok {
+		instanceType = val
+	} else {
+		return nil, fmt.Errorf("could not retrieve instance type from topology label")
+	}
+
+	var region string
+	if val, ok := node.GetLabels()[corev1.LabelTopologyRegion]; ok {
+		region = val
+	} else {
+		return nil, fmt.Errorf("could not retrieve region from topology label")
+	}
+
+	var availabilityZone string
+	if val, ok := node.GetLabels()[corev1.LabelTopologyZone]; ok {
+		availabilityZone = val
+	} else {
+		return nil, fmt.Errorf("could not retrieve AZ from topology label")
 	}
 
 	instanceInfo := Metadata{

pkg/cloud/metadata_test.go

@@ -130,6 +130,8 @@ func TestNewMetadataService(t *testing.T) {
 				ObjectMeta: metav1.ObjectMeta{
 					Labels: map[string]string{
 						"node.kubernetes.io/instance-type": stdInstanceType,
+						"topology.kubernetes.io/region":    stdRegion,
+						"topology.kubernetes.io/zone":      stdAvailabilityZone,
 					},
 					Name: nodeName,
 				},
@@ -179,38 +181,46 @@ func TestNewMetadataService(t *testing.T) {
 			nodeNameEnvVar: nodeName,
 		},
 		{
-			name:                 "failure: metadata not available, invalid region",
+			name:                 "failure: metadata not available, could not retrieve region",
 			ec2metadataAvailable: false,
-			expectedErr:          fmt.Errorf("did not find aws region in node providerID string"),
+			expectedErr:          fmt.Errorf("could not retrieve region from topology label"),
 			node: v1.Node{
 				TypeMeta: metav1.TypeMeta{
 					Kind:       "Node",
 					APIVersion: "v1",
 				},
 				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{
+						"node.kubernetes.io/instance-type": stdInstanceType,
+						"topology.kubernetes.io/zone":      stdAvailabilityZone,
+					},
 					Name: nodeName,
 				},
 				Spec: v1.NodeSpec{
-					ProviderID: "aws:///us-est-2b/i-abcdefgh123456789", // invalid region
+					ProviderID: "aws:///" + stdAvailabilityZone + "/" + stdInstanceID,
 				},
 				Status: v1.NodeStatus{},
 			},
 			nodeNameEnvVar: nodeName,
 		},
 		{
-			name:                 "failure: metadata not available, invalid az",
+			name:                 "failure: metadata not available, could not retrieve AZ",
 			ec2metadataAvailable: false,
-			expectedErr:          fmt.Errorf("did not find aws availability zone in node providerID string"),
+			expectedErr:          fmt.Errorf("could not retrieve AZ from topology label"),
 			node: v1.Node{
 				TypeMeta: metav1.TypeMeta{
 					Kind:       "Node",
 					APIVersion: "v1",
 				},
 				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{
+						"node.kubernetes.io/instance-type": stdInstanceType,
+						"topology.kubernetes.io/region":    stdRegion,
+					},
 					Name: nodeName,
 				},
 				Spec: v1.NodeSpec{
-					ProviderID: "aws:///us-west-21/i-abcdefgh123456789", // invalid AZ
+					ProviderID: "aws:///" + stdAvailabilityZone + "/" + stdInstanceID,
 				},
 				Status: v1.NodeStatus{},
 			},
@@ -226,6 +236,11 @@ func TestNewMetadataService(t *testing.T) {
 					APIVersion: "v1",
 				},
 				ObjectMeta: metav1.ObjectMeta{
+					Labels: map[string]string{
+						"node.kubernetes.io/instance-type": stdInstanceType,
+						"topology.kubernetes.io/region":    stdRegion,
+						"topology.kubernetes.io/zone":      stdAvailabilityZone,
+					},
 					Name: nodeName,
 				},
 				Spec: v1.NodeSpec{