add validation interval, cleanup 0 byte log files

cmd/kops/validate_cluster.go

@@ -66,11 +66,13 @@ type ValidateClusterOptions struct {
 	output      string
 	wait        time.Duration
 	count       int
+	interval    time.Duration
 	kubeconfig  string
 }
 
 func (o *ValidateClusterOptions) InitDefaults() {
 	o.output = OutputTable
+	o.interval = 10 * time.Second
 }
 
 func NewCmdValidateCluster(f *util.Factory, out io.Writer) *cobra.Command {
@@ -105,6 +107,7 @@ func NewCmdValidateCluster(f *util.Factory, out io.Writer) *cobra.Command {
 	})
 	cmd.Flags().DurationVar(&options.wait, "wait", options.wait, "Amount of time to wait for the cluster to become ready")
 	cmd.Flags().IntVar(&options.count, "count", options.count, "Number of consecutive successful validations required")
+	cmd.Flags().DurationVar(&options.interval, "interval", options.interval, "Time in duration to wait between validation attempts")
 	cmd.Flags().StringVar(&options.kubeconfig, "kubeconfig", "", "Path to the kubeconfig file")
 
 	return cmd
@@ -164,7 +167,6 @@ func RunValidateCluster(ctx context.Context, f *util.Factory, out io.Writer, opt
 	}
 
 	timeout := time.Now().Add(options.wait)
-	pollInterval := 10 * time.Second
 
 	validator, err := validation.NewClusterValidator(cluster, cloud, list, config.Host, k8sClient)
 	if err != nil {
@@ -182,7 +184,7 @@ func RunValidateCluster(ctx context.Context, f *util.Factory, out io.Writer, opt
 			consecutive = 0
 			if options.wait > 0 {
 				klog.Warningf("(will retry): unexpected error during validation: %v", err)
-				time.Sleep(pollInterval)
+				time.Sleep(options.interval)
 				continue
 			} else {
 				return nil, fmt.Errorf("unexpected error during validation: %v", err)
@@ -219,7 +221,7 @@ func RunValidateCluster(ctx context.Context, f *util.Factory, out io.Writer, opt
 			if consecutive < options.count {
 				klog.Infof("(will retry): cluster passed validation %d consecutive times", consecutive)
 				if options.wait > 0 {
-					time.Sleep(pollInterval)
+					time.Sleep(options.interval)
 					continue
 				} else {
 					return nil, fmt.Errorf("cluster passed validation %d consecutive times", consecutive)
@@ -231,7 +233,7 @@ func RunValidateCluster(ctx context.Context, f *util.Factory, out io.Writer, opt
 			if options.wait > 0 {
 				klog.Warningf("(will retry): cluster not yet healthy")
 				consecutive = 0
-				time.Sleep(pollInterval)
+				time.Sleep(options.interval)
 				continue
 			} else {
 				return nil, fmt.Errorf("cluster not yet healthy")

docs/cli/kops_validate_cluster.md

@@ -31,6 +31,7 @@ kops validate cluster [CLUSTER] [flags]
 ```
       --count int           Number of consecutive successful validations required
   -h, --help                help for cluster
+      --interval duration   Time in duration to wait between validation attempts (default 10s)
       --kubeconfig string   Path to the kubeconfig file
   -o, --output string       Output format. One of json|yaml|table. (default "table")
       --wait duration       Amount of time to wait for the cluster to become ready

tests/e2e/kubetest2-kops/deployer/deployer.go

@@ -65,7 +65,9 @@ type deployer struct {
 	ControlPlaneIGOverrides []string `flag:"control-plane-instance-group-overrides" desc:"overrides for the control plane instance groups"`
 	NodeIGOverrides         []string `flag:"node-instance-group-overrides" desc:"overrides for the node instance groups"`
 
-	ValidationWait time.Duration `flag:"validation-wait" desc:"time to wait for newly created cluster to pass validation"`
+	ValidationWait     time.Duration `flag:"validation-wait" desc:"time to wait for newly created cluster to pass validation"`
+	ValidationCount    int           `flag:"validation-count" desc:"how many times should a validation pass"`
+	ValidationInterval time.Duration `flag:"validation-interval" desc:"time in duration to wait between validation attempts"`
 
 	TemplatePath string `flag:"template-path" desc:"The path to the manifest template used for cluster creation"`
 
@@ -109,13 +111,15 @@ func (d *deployer) Provider() string {
 
 // New implements deployer.New for kops
 func New(opts types.Options) (types.Deployer, *pflag.FlagSet) {
-	// create a deployer object and set fields that are not flag controlled
+	// create a deployer object and set fields that are not flag controlled, and default values
 	d := &deployer{
 		commonOptions: opts,
 		BuildOptions: &builder.BuildOptions{
 			BuildKubernetes: false,
 		},
 		boskosHeartbeatClose: make(chan struct{}),
+		ValidationCount:      10,
+		ValidationInterval:   10 * time.Second,
 	}
 
 	dir, err := defaultArtifactsDir()

tests/e2e/kubetest2-kops/deployer/dumplogs.go

@@ -163,7 +163,10 @@ func (d *deployer) dumpClusterInfo() error {
 		"leases",
 		"persistentvolumeclaims",
 		"poddisruptionbudgets",
+		"podmonitors",
+		"statefulsets",
 		"serviceaccounts",
+		"servicemonitors",
 		"rolebindings",
 		"roles",
 	}
@@ -199,6 +202,11 @@ func (d *deployer) dumpClusterInfo() error {
 			}
 		}
 	}
+	// cleanup zero byte files
+	cmd = exec.Command("find", d.ArtifactsDir, "-size", "0", "-print", "-delete", "-o")
+	if err := cmd.Run(); err != nil {
+		dumpErr = errors.Join(dumpErr, err)
+	}
 	return dumpErr
 }
 

tests/e2e/kubetest2-kops/deployer/up.go

@@ -21,6 +21,7 @@ import (
 	"fmt"
 	osexec "os/exec"
 	"path"
+	"strconv"
 	"strings"
 	"time"
 
@@ -290,7 +291,8 @@ func (d *deployer) IsUp() (bool, error) {
 	args := []string{
 		d.KopsBinaryPath, "validate", "cluster",
 		"--name", d.ClusterName,
-		"--count", "10",
+		"--count", strconv.Itoa(d.ValidationCount),
+		"--interval", d.ValidationInterval.String(),
 		"--wait", wait.String(),
 	}
 	klog.Info(strings.Join(args, " "))