Add Eviction Timeout Field for Maintenance Time

README.md

@@ -46,6 +46,8 @@ Follow the instructions [here](https://sdk.operatorframework.io/docs/building-op
 
 To set maintenance on a node a `NodeMaintenance` custom resource should be created.
 The `NodeMaintenance` CR spec contains:
+
+- evictionTimeout: The timeout for pods eviction by drain/delete before giving up. Zero means infinite, and the default value is 30s.
 - nodeName: The name of the node which will be put into maintenance mode.
 - reason: The reason why the node will be under maintenance.
 
@@ -58,6 +60,7 @@ kind: NodeMaintenance
 metadata:
   name: nodemaintenance-sample
 spec:
+  evictionTimeout: "30s"
   nodeName: node02
   reason: "Test node maintenance"
 

api/v1beta1/nodemaintenance_types.go

@@ -45,10 +45,18 @@ type NodeMaintenanceSpec struct {
 	// INSERT ADDITIONAL SPEC FIELDS - desired state of cluster
 	// Important: Run "make" to regenerate code after modifying this file
 
-	// Node name to apply maintanance on/off
+	// EvictionTimeout is the timeout for pods eviction by drain/delete before giving up
+	// Zero means infinite
+	// Valid time units are "ms", "s", "m", "h".
+	// +kubebuilder:default:="30s"
+	// +kubebuilder:validation:Pattern="^(0|([0-9]+(\\.[0-9]+)?(ms|s|m|h)))$"
+	// +kubebuilder:validation:Type:=string
+	//+operator-sdk:csv:customresourcedefinitions:type=spec
+	EvictionTimeout metav1.Duration `json:"evictionTimeout,omitempty"`
+	// Node name to apply maintenance on/off
 	//+operator-sdk:csv:customresourcedefinitions:type=spec
 	NodeName string `json:"nodeName"`
-	// Reason for maintanance
+	// Reason for maintenance
 	//+operator-sdk:csv:customresourcedefinitions:type=spec
 	Reason string `json:"reason,omitempty"`
 }

bundle/manifests/node-maintenance-operator.clusterserviceversion.yaml

@@ -11,6 +11,7 @@ metadata:
             "name": "nodemaintenance-sample"
           },
           "spec": {
+            "evictionTimeout": "30s",
             "nodeName": "node02",
             "reason": "Test node maintenance"
           }
@@ -43,10 +44,15 @@ spec:
         name: nodemaintenances
         version: v1beta1
       specDescriptors:
-      - description: Node name to apply maintanance on/off
+      - description: EvictionTimeout is the timeout for pods eviction by drain/delete
+          before giving up Zero means infinite Valid time units are "ms", "s", "m",
+          "h".
+        displayName: Eviction Timeout
+        path: evictionTimeout
+      - description: Node name to apply maintenance on/off
         displayName: Node Name
         path: nodeName
-      - description: Reason for maintanance
+      - description: Reason for maintenance
         displayName: Reason
         path: reason
       statusDescriptors:

bundle/manifests/nodemaintenance.medik8s.io_nodemaintenances.yaml

@@ -43,11 +43,19 @@ spec:
           spec:
             description: NodeMaintenanceSpec defines the desired state of NodeMaintenance
             properties:
+              evictionTimeout:
+                default: 30s
+                description: |-
+                  EvictionTimeout is the timeout for pods eviction by drain/delete before giving up
+                  Zero means infinite
+                  Valid time units are "ms", "s", "m", "h".
+                pattern: ^(0|([0-9]+(\.[0-9]+)?(ms|s|m|h)))$
+                type: string
               nodeName:
-                description: Node name to apply maintanance on/off
+                description: Node name to apply maintenance on/off
                 type: string
               reason:
-                description: Reason for maintanance
+                description: Reason for maintenance
                 type: string
             required:
             - nodeName

config/crd/bases/nodemaintenance.medik8s.io_nodemaintenances.yaml

@@ -41,11 +41,19 @@ spec:
           spec:
             description: NodeMaintenanceSpec defines the desired state of NodeMaintenance
             properties:
+              evictionTimeout:
+                default: 30s
+                description: |-
+                  EvictionTimeout is the timeout for pods eviction by drain/delete before giving up
+                  Zero means infinite
+                  Valid time units are "ms", "s", "m", "h".
+                pattern: ^(0|([0-9]+(\.[0-9]+)?(ms|s|m|h)))$
+                type: string
               nodeName:
-                description: Node name to apply maintanance on/off
+                description: Node name to apply maintenance on/off
                 type: string
               reason:
-                description: Reason for maintanance
+                description: Reason for maintenance
                 type: string
             required:
             - nodeName

config/manifests/bases/node-maintenance-operator.clusterserviceversion.yaml

@@ -28,10 +28,15 @@ spec:
         name: nodemaintenances
         version: v1beta1
       specDescriptors:
-      - description: Node name to apply maintanance on/off
+      - description: EvictionTimeout is the timeout for pods eviction by drain/delete
+          before giving up Zero means infinite Valid time units are "ms", "s", "m",
+          "h".
+        displayName: Eviction Timeout
+        path: evictionTimeout
+      - description: Node name to apply maintenance on/off
         displayName: Node Name
         path: nodeName
-      - description: Reason for maintanance
+      - description: Reason for maintenance
         displayName: Reason
         path: reason
       statusDescriptors:

config/samples/nodemaintenance_v1beta1_nodemaintenance.yaml

@@ -3,5 +3,6 @@ kind: NodeMaintenance
 metadata:
   name: nodemaintenance-sample
 spec:
+  evictionTimeout: "30s"
   nodeName: node02
   reason: "Test node maintenance"

controllers/controllers_suite_test.go

@@ -20,6 +20,7 @@ import (
 	"context"
 	"path/filepath"
 	"testing"
+	"time"
 
 	"github.com/medik8s/common/pkg/lease"
 	. "github.com/onsi/ginkgo/v2"
@@ -95,7 +96,8 @@ var _ = BeforeSuite(func() {
 		logger:       ctrl.Log.WithName("unit test"),
 	}
 	ctx, cancel = context.WithCancel(ctrl.SetupSignalHandler())
-	drainer, err = createDrainer(ctx, cfg)
+	evictionTimeout := time.Duration(30)
+	drainer, err = createDrainer(ctx, evictionTimeout, cfg)
 	Expect(err).NotTo(HaveOccurred())
 	// in test pods are not evicted, so don't wait forever for them
 	drainer.SkipWaitForDeleteTimeoutSeconds = 0

controllers/nodemaintenance_controller.go

@@ -52,7 +52,6 @@ const (
 	//lease consts
 	LeaseHolderIdentity = "node-maintenance"
 	LeaseDuration       = 3600 * time.Second
-	DrainerTimeout      = 30 * time.Second
 )
 
 // NodeMaintenanceReconciler reconciles a NodeMaintenance object
@@ -118,7 +117,7 @@ func (r *NodeMaintenanceReconciler) Reconcile(ctx context.Context, req ctrl.Requ
 	}
 
 	// Add finalizer when object is created
-	drainer, err := createDrainer(ctx, r.MgrConfig)
+	drainer, err := createDrainer(ctx, nm.Spec.EvictionTimeout.Duration, r.MgrConfig)
 	if err != nil {
 		return emptyResult, err
 	}
@@ -229,7 +228,7 @@ func (r *NodeMaintenanceReconciler) Reconcile(ctx context.Context, req ctrl.Requ
 }
 
 // createDrainer creates a drain.Helper struct for external cordon and drain API
-func createDrainer(ctx context.Context, mgrConfig *rest.Config) (*drain.Helper, error) {
+func createDrainer(ctx context.Context, evictionTimeout time.Duration, mgrConfig *rest.Config) (*drain.Helper, error) {
 	drainer := &drain.Helper{}
 
 	//Continue even if there are pods not managed by a ReplicationController, ReplicaSet, Job, DaemonSet or StatefulSet.
@@ -254,9 +253,8 @@ func createDrainer(ctx context.Context, mgrConfig *rest.Config) (*drain.Helper,
 	//Period of time in seconds given to each pod to terminate gracefully. If negative, the default value specified in the pod will be used.
 	drainer.GracePeriodSeconds = -1
 
-	// TODO - add logical value or attach from the maintenance CR
 	//The length of time to wait before giving up, zero means infinite
-	drainer.Timeout = DrainerTimeout
+	drainer.Timeout = evictionTimeout
 
 	cs, err := kubernetes.NewForConfig(mgrConfig)
 	if err != nil {

controllers/taint.go

@@ -56,7 +56,7 @@ func AddOrRemoveTaint(clientset kubernetes.Interface, add bool, node *corev1.Nod
 			return err
 		}
 		taintStr = "remove"
-		log.Infof("Maintenance taints  will be removed from node %s", node.Name)
+		log.Infof("Maintenance taints will be removed from node %s", node.Name)
 		patch = fmt.Sprintf(`{ "op": "replace", "path": "/spec/taints", "value": %s }`, string(removeTaints))
 	}
 

test/e2e/e2e_suite_test.go

@@ -19,14 +19,15 @@ import (
 	"fmt"
 	"os"
 	"testing"
+	"time"
 
 	. "github.com/onsi/ginkgo/v2"
 	"github.com/onsi/ginkgo/v2/reporters"
 	. "github.com/onsi/gomega"
 
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/errors"
-	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 
 	"github.com/medik8s/node-maintenance-operator/api/v1beta1"
 )
@@ -41,8 +42,10 @@ var (
 	// The ns for test deployments
 	testNsName    string
 	testNamespace *corev1.Namespace
-	//namespace leases are created in
+	// namespace leases are created in
 	leaseNs = "medik8s-leases"
+	// default eviction timeout (30s)
+	evicitonTimeout = time.Second * 50
 )
 
 var _ = BeforeSuite(func() {
@@ -52,7 +55,7 @@ var _ = BeforeSuite(func() {
 	testNsName = os.Getenv("TEST_NAMESPACE")
 	Expect(testNsName).ToNot(BeEmpty(), "TEST_NAMESPACE env var not set, can't start e2e test")
 	testNamespace = &corev1.Namespace{
-		ObjectMeta: v1.ObjectMeta{
+		ObjectMeta: metav1.ObjectMeta{
 			Name: testNsName,
 		},
 	}
@@ -66,7 +69,7 @@ var _ = BeforeSuite(func() {
 	Expect(err).ToNot(HaveOccurred())
 
 	// wait until webhooks are up and running by trying to create a CR and ignoring unexpected errors
-	testCR := getNodeMaintenance("webhook-test", "some-not-existing-node-name")
+	testCR := getNodeMaintenance("webhook-test", "some-not-existing-node-name", evicitonTimeout)
 	_ = createCRIgnoreUnrelatedErrors(testCR)
 })