feat(controller): use CRD status subresource

Signed-off-by: Jesse Suen <[email protected]>

manifests/crds/rollout-crd.yaml

@@ -36,6 +36,7 @@ spec:
       labelSelectorPath: .status.selector
       specReplicasPath: .spec.replicas
       statusReplicasPath: .status.HPAReplicas
+    status: {}
   validation:
     openAPIV3Schema:
       properties:
@@ -3059,8 +3060,6 @@ spec:
                   - name
                   - status
                   type: object
-                stableRS:
-                  type: string
               type: object
             collisionCount:
               format: int32

manifests/install.yaml

@@ -11057,11 +11057,13 @@ spec:
     name: Current
     type: integer
   - JSONPath: .status.updatedReplicas
-    description: Total number of non-terminated pods targeted by this rollout that have the desired template spec
+    description: Total number of non-terminated pods targeted by this rollout that
+      have the desired template spec
     name: Up-to-date
     type: integer
   - JSONPath: .status.availableReplicas
-    description: Total number of available pods (ready for at least minReadySeconds) targeted by this rollout
+    description: Total number of available pods (ready for at least minReadySeconds)
+      targeted by this rollout
     name: Available
     type: integer
   group: argoproj.io
@@ -11078,6 +11080,7 @@ spec:
       labelSelectorPath: .status.selector
       specReplicasPath: .spec.replicas
       statusReplicasPath: .status.HPAReplicas
+    status: {}
   validation:
     openAPIV3Schema:
       properties:
@@ -14101,8 +14104,6 @@ spec:
                   - name
                   - status
                   type: object
-                stableRS:
-                  type: string
               type: object
             collisionCount:
               format: int32

manifests/namespace-install.yaml

@@ -11057,11 +11057,13 @@ spec:
     name: Current
     type: integer
   - JSONPath: .status.updatedReplicas
-    description: Total number of non-terminated pods targeted by this rollout that have the desired template spec
+    description: Total number of non-terminated pods targeted by this rollout that
+      have the desired template spec
     name: Up-to-date
     type: integer
   - JSONPath: .status.availableReplicas
-    description: Total number of available pods (ready for at least minReadySeconds) targeted by this rollout
+    description: Total number of available pods (ready for at least minReadySeconds)
+      targeted by this rollout
     name: Available
     type: integer
   group: argoproj.io
@@ -11078,6 +11080,7 @@ spec:
       labelSelectorPath: .status.selector
       specReplicasPath: .spec.replicas
       statusReplicasPath: .status.HPAReplicas
+    status: {}
   validation:
     openAPIV3Schema:
       properties:
@@ -14101,8 +14104,6 @@ spec:
                   - name
                   - status
                   type: object
-                stableRS:
-                  type: string
               type: object
             collisionCount:
               format: int32

pkg/apis/rollouts/v1alpha1/types.go

@@ -10,14 +10,14 @@ import (
 )
 
 // +genclient
-// +genclient:noStatus
 // +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
 // +kubebuilder:resource:path=rollouts,shortName=ro
 // +kubebuilder:subresource:scale:specpath=.spec.replicas,statuspath=.status.HPAReplicas,selectorpath=.status.selector
 // +kubebuilder:printcolumn:name="Desired",type="integer",JSONPath=".spec.replicas",description="Number of desired pods"
 // +kubebuilder:printcolumn:name="Current",type="integer",JSONPath=".status.replicas",description="Total number of non-terminated pods targeted by this rollout"
 // +kubebuilder:printcolumn:name="Up-to-date",type="integer",JSONPath=".status.updatedReplicas",description="Total number of non-terminated pods targeted by this rollout that have the desired template spec"
 // +kubebuilder:printcolumn:name="Available",type="integer",JSONPath=".status.availableReplicas",description="Total number of available pods (ready for at least minReadySeconds) targeted by this rollout"
+// +kubebuilder:subresource:status
 
 // Rollout is a specification for a Rollout resource
 type Rollout struct {
@@ -599,9 +599,6 @@ type BlueGreenStatus struct {
 
 // CanaryStatus status fields that only pertain to the canary rollout
 type CanaryStatus struct {
-	// StableRS indicates the last replicaset that walked through all the canary steps or was the only replicaset
-	// +optional
-	StableRS string `json:"stableRS,omitempty"`
 	// CurrentStepAnalysisRun indicates the analysisRun for the current step index
 	// TODO(Deprecated): Remove in v0.10
 	CurrentStepAnalysisRun string `json:"currentStepAnalysisRun,omitempty"`

pkg/kubectl-argo-rollouts/cmd/abort/abort.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 
 	"github.com/spf13/cobra"
+	k8serrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	types "k8s.io/apimachinery/pkg/types"
 
@@ -58,5 +59,10 @@ func NewCmdAbort(o *options.ArgoRolloutsOptions) *cobra.Command {
 // AbortRollout aborts a rollout
 func AbortRollout(rolloutIf clientset.RolloutInterface, name string) (*v1alpha1.Rollout, error) {
 	ctx := context.TODO()
-	return rolloutIf.Patch(ctx, name, types.MergePatchType, []byte(abortPatch), metav1.PatchOptions{})
+	// attempt using status subresource, first
+	ro, err := rolloutIf.Patch(ctx, name, types.MergePatchType, []byte(abortPatch), metav1.PatchOptions{}, "status")
+	if err != nil && k8serrors.IsNotFound(err) {
+		ro, err = rolloutIf.Patch(ctx, name, types.MergePatchType, []byte(abortPatch), metav1.PatchOptions{})
+	}
+	return ro, err
 }

pkg/kubectl-argo-rollouts/cmd/get/get_test.go

@@ -118,6 +118,7 @@ func TestGetBlueGreenRollout(t *testing.T) {
 Name:            bluegreen-demo
 Namespace:       jesse-test
 Status:          ॥ Paused
+Message:         BlueGreenPause
 Strategy:        BlueGreen
 Images:          argoproj/rollouts-demo:blue (stable, active)
                  argoproj/rollouts-demo:green (preview)
@@ -165,6 +166,7 @@ func TestGetBlueGreenRolloutScaleDownDelay(t *testing.T) {
 Name:            bluegreen-demo
 Namespace:       jesse-test
 Status:          ॥ Paused
+Message:         BlueGreenPause
 Strategy:        BlueGreen
 Images:          argoproj/rollouts-demo:blue (stable, active)
                  argoproj/rollouts-demo:green
@@ -212,6 +214,7 @@ func TestGetBlueGreenRolloutScaleDownDelayPassed(t *testing.T) {
 Name:            bluegreen-demo
 Namespace:       jesse-test
 Status:          ॥ Paused
+Message:         BlueGreenPause
 Strategy:        BlueGreen
 Images:          argoproj/rollouts-demo:blue (stable, active)
                  argoproj/rollouts-demo:green

pkg/kubectl-argo-rollouts/cmd/promote/promote.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 
 	"github.com/spf13/cobra"
+	k8serrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	types "k8s.io/apimachinery/pkg/types"
 
@@ -29,20 +30,12 @@ If not on a pause step use '--skip-current-step' to progress to the next step in
 )
 
 const (
-	setCurrentStepIndex = `{
-	"status": {
-		"currentStepIndex": %d
-	}
-}`
+	setCurrentStepIndex                 = `{"status":{"currentStepIndex":%d}}`
+	unpausePatch                        = `{"spec":{"paused":false}}`
+	clearPauseConditionsPatch           = `{"status":{"pauseConditions":null}}`
+	unpauseAndClearPauseConditionsPatch = `{"spec":{"paused":false},"status":{"pauseConditions":null}}`
+	fullPromotionPatch                  = `{"status":{"stableRS":"%s"}}`
 
-	unpausePatch = `{
-	"spec": {
-		"paused": false
-	},
-	"status": {
-		"pauseConditions": null
-	}
-}`
 	useBothSkipFlagsError         = "Cannot use skip-current-step and skip-all-steps flags at the same time"
 	skipFlagsWithBlueGreenError   = "Cannot skip steps of a bluegreen rollout. Run without a flags"
 	skipFlagWithNoStepCanaryError = "Cannot skip steps of a rollout without steps"
@@ -53,6 +46,7 @@ func NewCmdPromote(o *options.ArgoRolloutsOptions) *cobra.Command {
 	var (
 		skipCurrentStep = false
 		skipAllSteps    = false
+		full            = false
 	)
 	var cmd = &cobra.Command{
 		Use:          "promote ROLLOUT_NAME",
@@ -69,7 +63,7 @@ func NewCmdPromote(o *options.ArgoRolloutsOptions) *cobra.Command {
 			}
 			name := args[0]
 			rolloutIf := o.RolloutsClientset().ArgoprojV1alpha1().Rollouts(o.Namespace())
-			ro, err := PromoteRollout(rolloutIf, name, skipCurrentStep, skipAllSteps)
+			ro, err := PromoteRollout(rolloutIf, name, skipCurrentStep, skipAllSteps, full)
 			if err != nil {
 				return err
 			}
@@ -79,11 +73,12 @@ func NewCmdPromote(o *options.ArgoRolloutsOptions) *cobra.Command {
 	}
 	cmd.Flags().BoolVarP(&skipCurrentStep, "skip-current-step", "c", false, "Skip current step")
 	cmd.Flags().BoolVarP(&skipAllSteps, "skip-all-steps", "a", false, "Skip remaining steps")
+	cmd.Flags().BoolVar(&full, "full", false, "Perform a full promotion (skip analysis, pausing)")
 	return cmd
 }
 
 // PromoteRollout promotes a rollout to the next step, or to end of all steps
-func PromoteRollout(rolloutIf clientset.RolloutInterface, name string, skipCurrentStep, skipAllSteps bool) (*v1alpha1.Rollout, error) {
+func PromoteRollout(rolloutIf clientset.RolloutInterface, name string, skipCurrentStep, skipAllSteps, full bool) (*v1alpha1.Rollout, error) {
 	ctx := context.TODO()
 	ro, err := rolloutIf.Get(ctx, name, metav1.GetOptions{})
 	if err != nil {
@@ -97,15 +92,37 @@ func PromoteRollout(rolloutIf clientset.RolloutInterface, name string, skipCurre
 			return nil, fmt.Errorf(skipFlagWithNoStepCanaryError)
 		}
 	}
-	patch := getPatch(ro, skipCurrentStep, skipAllSteps)
-	ro, err = rolloutIf.Patch(ctx, name, types.MergePatchType, patch, metav1.PatchOptions{})
-	if err != nil {
-		return nil, err
+
+	// This function is intended to be compatible with Rollouts v0.9 and Rollouts v0.10+, the latter
+	// of which uses CRD status subresources. When using status subresource, status must be updated
+	// separately from spec. Since we don't know which version is installed in the cluster, we
+	// attempt status patching first. If it errors with NotFound, it indicates that status
+	// subresource is not used (v0.9), at which point we need to use the unified patch that updates
+	// both spec and status. Otherwise, we proceed with a spec only patch.
+	// NOTE: this logic can be simplified in the future once we  can safely assume users are using
+	// v0.10+, and wish to drop support for v0.9 and below
+	specPatch, statusPatch, unifiedPatch := getPatches(ro, skipCurrentStep, skipAllSteps, full)
+	if statusPatch != nil {
+		ro, err = rolloutIf.Patch(ctx, name, types.MergePatchType, statusPatch, metav1.PatchOptions{}, "status")
+		if err != nil {
+			if !k8serrors.IsNotFound(err) {
+				return nil, err
+			}
+			// we got a NotFound error. status subresource is not being used, so perform unifiedPatch
+			specPatch = unifiedPatch
+		}
+	}
+	if specPatch != nil {
+		ro, err = rolloutIf.Patch(ctx, name, types.MergePatchType, specPatch, metav1.PatchOptions{})
+		if err != nil {
+			return nil, err
+		}
 	}
 	return ro, nil
 }
 
-func getPatch(rollout *v1alpha1.Rollout, skipCurrentStep, skipAllStep bool) []byte {
+func getPatches(rollout *v1alpha1.Rollout, skipCurrentStep, skipAllStep, full bool) ([]byte, []byte, []byte) {
+	var specPatch, statusPatch, unifiedPatch []byte
 	switch {
 	case skipCurrentStep:
 		_, index := replicasetutil.GetCurrentCanaryStep(rollout)
@@ -114,10 +131,22 @@ func getPatch(rollout *v1alpha1.Rollout, skipCurrentStep, skipAllStep bool) []by
 		if *index < int32(len(rollout.Spec.Strategy.Canary.Steps)) {
 			*index++
 		}
-		return []byte(fmt.Sprintf(setCurrentStepIndex, *index))
+		statusPatch = []byte(fmt.Sprintf(setCurrentStepIndex, *index))
+		unifiedPatch = statusPatch
 	case skipAllStep:
-		return []byte(fmt.Sprintf(setCurrentStepIndex, len(rollout.Spec.Strategy.Canary.Steps)))
+		statusPatch = []byte(fmt.Sprintf(setCurrentStepIndex, len(rollout.Spec.Strategy.Canary.Steps)))
+		unifiedPatch = statusPatch
+	case full:
+		statusPatch = []byte(fmt.Sprintf(fullPromotionPatch, rollout.Status.CurrentPodHash))
+		unifiedPatch = statusPatch
 	default:
-		return []byte(unpausePatch)
+		if rollout.Spec.Paused {
+			specPatch = []byte(unpausePatch)
+		}
+		if len(rollout.Status.PauseConditions) > 0 {
+			statusPatch = []byte(clearPauseConditionsPatch)
+		}
+		unifiedPatch = []byte(unpauseAndClearPauseConditionsPatch)
 	}
+	return specPatch, statusPatch, unifiedPatch
 }

pkg/kubectl-argo-rollouts/cmd/retry/retry.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 
 	"github.com/spf13/cobra"
+	k8serrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	types "k8s.io/apimachinery/pkg/types"
 
@@ -82,7 +83,12 @@ func NewCmdRetryRollout(o *options.ArgoRolloutsOptions) *cobra.Command {
 // RetryRollout retries a rollout after it's been aborted
 func RetryRollout(rolloutIf clientset.RolloutInterface, name string) (*v1alpha1.Rollout, error) {
 	ctx := context.TODO()
-	return rolloutIf.Patch(ctx, name, types.MergePatchType, []byte(retryRolloutPatch), metav1.PatchOptions{})
+	// attempt using status subresource, first
+	ro, err := rolloutIf.Patch(ctx, name, types.MergePatchType, []byte(retryRolloutPatch), metav1.PatchOptions{}, "status")
+	if err != nil && k8serrors.IsNotFound(err) {
+		ro, err = rolloutIf.Patch(ctx, name, types.MergePatchType, []byte(retryRolloutPatch), metav1.PatchOptions{})
+	}
+	return ro, err
 }
 
 // NewCmdRetryExperiment returns a new instance of an `argo rollouts retry experiment` command