Merge pull request #16231 from hakman/cluster-deletion-configurable

Make cluster deletion configurable

cmd/kops/delete_cluster.go

@@ -20,6 +20,7 @@ import (
 	"context"
 	"fmt"
 	"io"
+	"time"
 
 	"github.com/spf13/cobra"
 	"k8s.io/client-go/tools/clientcmd"
@@ -44,6 +45,14 @@ type DeleteClusterOptions struct {
 	External    bool
 	Unregister  bool
 	ClusterName string
+	wait        time.Duration
+	count       int
+	interval    time.Duration
+}
+
+func (o *DeleteClusterOptions) InitDefaults() {
+	o.count = 42
+	o.interval = 10 * time.Second
 }
 
 var (
@@ -64,6 +73,7 @@ var (
 
 func NewCmdDeleteCluster(f *util.Factory, out io.Writer) *cobra.Command {
 	options := &DeleteClusterOptions{}
+	options.InitDefaults()
 
 	cmd := &cobra.Command{
 		Use:               "cluster [CLUSTER]",
@@ -84,6 +94,10 @@ func NewCmdDeleteCluster(f *util.Factory, out io.Writer) *cobra.Command {
 	cmd.Flags().StringVar(&options.Region, "region", options.Region, "External cluster's cloud region")
 	cmd.RegisterFlagCompletionFunc("region", completeRegion)
 
+	cmd.Flags().DurationVar(&options.wait, "wait", options.wait, "Amount of time to wait for the cluster resources to de deleted")
+	cmd.Flags().IntVar(&options.count, "count", options.count, "Number of consecutive failures to make progress deleting the cluster resources")
+	cmd.Flags().DurationVar(&options.interval, "interval", options.interval, "Time in duration to wait between deletion attempts")
+
 	return cmd
 }
 
@@ -171,7 +185,7 @@ func RunDeleteCluster(ctx context.Context, f *util.Factory, out io.Writer, optio
 
 			fmt.Fprintf(out, "\n")
 
-			err = resourceops.DeleteResources(cloud, clusterResources)
+			err = resourceops.DeleteResources(cloud, clusterResources, options.count, options.interval, options.wait)
 			if err != nil {
 				return err
 			}

docs/cli/kops_delete_cluster.md

@@ -24,10 +24,13 @@ kops delete cluster [CLUSTER] [flags]
 ### Options
 
 ```
+      --count int           Number of consecutive failures to make progress deleting the cluster resources (default 42)
       --external            Delete an external cluster
   -h, --help                help for cluster
+      --interval duration   Time in duration to wait between deletion attempts (default 10s)
       --region string       External cluster's cloud region
       --unregister          Don't delete cloud resources, just unregister the cluster
+      --wait duration       Amount of time to wait for the cluster resources to de deleted
   -y, --yes                 Specify --yes to delete the cluster
 ```
 

pkg/resources/ops/delete.go

@@ -28,7 +28,7 @@ import (
 )
 
 // DeleteResources deletes the resources, as previously collected by ListResources
-func DeleteResources(cloud fi.Cloud, resourceMap map[string]*resources.Resource) error {
+func DeleteResources(cloud fi.Cloud, resourceMap map[string]*resources.Resource, count int, interval, wait time.Duration) error {
 	depMap := make(map[string][]string)
 
 	done := make(map[string]*resources.Resource)
@@ -52,9 +52,12 @@ func DeleteResources(cloud fi.Cloud, resourceMap map[string]*resources.Resource)
 		klog.V(2).Infof("\t%s\t%v", k, v)
 	}
 
+	timeout := time.Now().Add(wait)
 	iterationsWithNoProgress := 0
 	for {
-		// TODO: Some form of default ordering based on types?
+		if wait > 0 && time.Now().After(timeout) {
+			return fmt.Errorf("wait time exceeded during resources deletion")
+		}
 
 		failed := make(map[string]*resources.Resource)
 
@@ -167,10 +170,10 @@ func DeleteResources(cloud fi.Cloud, resourceMap map[string]*resources.Resource)
 		}
 
 		iterationsWithNoProgress++
-		if iterationsWithNoProgress > 42 {
+		if iterationsWithNoProgress > count && count != 0 {
 			return fmt.Errorf("not making progress deleting resources; giving up")
 		}
 
-		time.Sleep(10 * time.Second)
+		time.Sleep(interval)
 	}
 }