[Multi_K8s-Plugin] Primary Rollout

pkg/app/pipedv1/plugin/kubernetes_multicluster/config/application.go

@@ -56,7 +56,6 @@ type KubernetesApplicationSpec struct {
 	// The label will be configured to variant manifests used to distinguish them.
 	VariantLabel KubernetesVariantLabel `json:"variantLabel"`
 
-	// TODO: Define fields for KubernetesApplicationSpec.
 }
 
 func (s *KubernetesApplicationSpec) UnmarshalJSON(data []byte) error {
@@ -201,6 +200,32 @@ type K8sCanaryRolloutStageOptions struct {
 // K8sCanaryCleanStageOptions contains all configurable values for a K8S_CANARY_CLEAN stage.
 type K8sCanaryCleanStageOptions struct{}
 
+// K8sPrimaryRolloutStageOptions contains all configurable values for a K8S_PRIMARY_ROLLOUT stage.
+type K8sPrimaryRolloutStageOptions struct {
+	// Suffix that should be used when naming the PRIMARY variant's resources.
+	// Default is "primary".
+	Suffix string `json:"suffix" default:"primary"`
+	// Whether the PRIMARY service should be created.
+	CreateService bool `json:"createService"`
+	// Whether the PRIMARY variant label should be added to manifests if they were missing.
+	AddVariantLabelToSelector bool `json:"addVariantLabelToSelector"`
+	// Whether the resources that are no longer defined in Git should be removed or not.
+	Prune bool `json:"prune"`
+}
+
+func (o *K8sPrimaryRolloutStageOptions) UnmarshalJSON(data []byte) error {
+	type alias K8sPrimaryRolloutStageOptions
+	var a alias
+	if err := json.Unmarshal(data, &a); err != nil {
+		return err
+	}
+	*o = K8sPrimaryRolloutStageOptions(a)
+	if err := defaults.Set(o); err != nil {
+		return err
+	}
+	return nil
+}
+
 // K8sResourcePatch represents a patch operation for a Kubernetes resource.
 type K8sResourcePatch struct {
 	// The target of the patch operation.

pkg/app/pipedv1/plugin/kubernetes_multicluster/deployment/pipeline.go

@@ -30,13 +30,16 @@ const (
 	StageK8sMultiCanaryRollout = "K8S_CANARY_ROLLOUT"
 	// StageK8sMultiCanaryClean represents the state where all canary resources should be removed.
 	StageK8sMultiCanaryClean = "K8S_CANARY_CLEAN"
+	// StageK8sMultiPrimaryRollout represents the state where the new version is promoted as PRIMARY to all targets.
+	StageK8sMultiPrimaryRollout = "K8S_PRIMARY_ROLLOUT"
 )
 
 var allStages = []string{
 	StageK8sMultiSync,
 	StageK8sMultiRollback,
 	StageK8sMultiCanaryRollout,
 	StageK8sMultiCanaryClean,
+	StageK8sMultiPrimaryRollout,
 }
 
 const (
@@ -48,6 +51,8 @@ const (
 	StageDescriptionK8sMultiCanaryRollout = "Rollout the new version as CANARY to all targets"
 	// StageDescriptionK8sMultiCanaryClean represents the description of the K8sCanaryClean stage.
 	StageDescriptionK8sMultiCanaryClean = "Remove all canary resources"
+	// StageDescriptionK8sMultiPrimaryRollout represents the description of the K8sPrimaryRollout stage.
+	StageDescriptionK8sMultiPrimaryRollout = "Rollout the new version as PRIMARY to all targets"
 )
 
 func buildQuickSyncPipeline(autoRollback bool) []sdk.QuickSyncStage {

pkg/app/pipedv1/plugin/kubernetes_multicluster/deployment/plugin.go

@@ -73,6 +73,8 @@ func (p *Plugin) ExecuteStage(ctx context.Context, _ *sdk.ConfigNone, dts []*sdk
 		return &sdk.ExecuteStageResponse{
 			Status: p.executeK8sMultiCanaryCleanStage(ctx, input, dts),
 		}, nil
+	case StageK8sMultiPrimaryRollout:
+		return &sdk.ExecuteStageResponse{Status: p.executeK8sMultiPrimaryRolloutStage(ctx, input, dts)}, nil
 	default:
 		return nil, errors.New("unimplemented or unsupported stage")
 	}

pkg/app/pipedv1/plugin/kubernetes_multicluster/deployment/primary.go

@@ -0,0 +1,263 @@
+// Copyright 2025 The PipeCD Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package deployment
+
+import (
+	"cmp"
+	"context"
+	"encoding/json"
+	"fmt"
+	"time"
+
+	"golang.org/x/sync/errgroup"
+
+	sdk "github.com/pipe-cd/piped-plugin-sdk-go"
+
+	kubeconfig "github.com/pipe-cd/pipecd/pkg/app/pipedv1/plugin/kubernetes_multicluster/config"
+	"github.com/pipe-cd/pipecd/pkg/app/pipedv1/plugin/kubernetes_multicluster/provider"
+	"github.com/pipe-cd/pipecd/pkg/app/pipedv1/plugin/kubernetes_multicluster/toolregistry"
+)
+
+func (p *Plugin) executeK8sMultiPrimaryRolloutStage(ctx context.Context, input *sdk.ExecuteStageInput[kubeconfig.KubernetesApplicationSpec], dts []*sdk.DeployTarget[kubeconfig.KubernetesDeployTargetConfig]) sdk.StageStatus {
+	lp := input.Client.LogPersister()
+
+	cfg, err := input.Request.TargetDeploymentSource.AppConfig()
+	if err != nil {
+		lp.Errorf("Failed while decoding application config (%v)", err.Error())
+		return sdk.StageStatusFailure
+	}
+
+	var stageCfg kubeconfig.K8sPrimaryRolloutStageOptions
+	if len(input.Request.StageConfig) > 0 {
+		if err := json.Unmarshal(input.Request.StageConfig, &stageCfg); err != nil {
+			lp.Errorf("Failed while unmarshalling stage config (%v)", err)
+			return sdk.StageStatusFailure
+		}
+	}
+
+	type targetConfig struct {
+		deployTarget *sdk.DeployTarget[kubeconfig.KubernetesDeployTargetConfig]
+		multiTarget  *kubeconfig.KubernetesMultiTarget
+	}
+
+	deployTargetMap := make(map[string]*sdk.DeployTarget[kubeconfig.KubernetesDeployTargetConfig])
+	targetConfigs := make([]targetConfig, 0, len(dts))
+
+	for _, target := range dts {
+		deployTargetMap[target.Name] = target
+	}
+
+	// If no multi-targets are specified, roll out primary to all deploy targets.
+	if len(cfg.Spec.Input.MultiTargets) == 0 {
+		for _, dt := range dts {
+			targetConfigs = append(targetConfigs, targetConfig{
+				deployTarget: dt,
+				multiTarget:  nil,
+			})
+		}
+	} else {
+		for _, multiTarget := range cfg.Spec.Input.MultiTargets {
+			dt, ok := deployTargetMap[multiTarget.Target.Name]
+			if !ok {
+				lp.Infof("Ignore multi target '%s': not matched any deployTarget", multiTarget.Target.Name)
+				continue
+			}
+			targetConfigs = append(targetConfigs, targetConfig{
+				deployTarget: dt,
+				multiTarget:  &multiTarget,
+			})
+		}
+	}
+
+	eg, ctx := errgroup.WithContext(ctx)
+	for _, tc := range targetConfigs {
+		eg.Go(func() error {
+			lp.Infof("Start primary rollout for target %s", tc.deployTarget.Name)
+			status := p.primaryRollout(ctx, input, tc.deployTarget, tc.multiTarget, stageCfg)
+			if status == sdk.StageStatusFailure {
+				return fmt.Errorf("failed to primary rollout for target %s", tc.deployTarget.Name)
+			}
+			return nil
+		})
+	}
+
+	if err := eg.Wait(); err != nil {
+		lp.Errorf("Failed while rolling out primary (%v)", err)
+		return sdk.StageStatusFailure
+	}
+
+	return sdk.StageStatusSuccess
+}
+
+func (p *Plugin) primaryRollout(
+	ctx context.Context,
+	input *sdk.ExecuteStageInput[kubeconfig.KubernetesApplicationSpec],
+	dt *sdk.DeployTarget[kubeconfig.KubernetesDeployTargetConfig],
+	multiTarget *kubeconfig.KubernetesMultiTarget,
+	stageCfg kubeconfig.K8sPrimaryRolloutStageOptions,
+) sdk.StageStatus {
+	lp := input.Client.LogPersister()
+
+	cfg, err := input.Request.TargetDeploymentSource.AppConfig()
+	if err != nil {
+		lp.Errorf("Failed while loading application config (%v)", err)
+		return sdk.StageStatusFailure
+	}
+
+	var (
+		appCfg         = cfg.Spec
+		variantLabel   = appCfg.VariantLabel.Key
+		primaryVariant = appCfg.VariantLabel.PrimaryValue
+	)
+
+	toolRegistry := toolregistry.NewRegistry(input.Client.ToolRegistry())
+	loader := provider.NewLoader(toolRegistry)
+
+	lp.Infof("Loading manifests at commit %s for handling", input.Request.TargetDeploymentSource.CommitHash)
+	manifests, err := p.loadManifests(ctx, &input.Request.Deployment, cfg.Spec, &input.Request.TargetDeploymentSource, loader, input.Logger, multiTarget)
+	if err != nil {
+		lp.Errorf("Failed while loading manifests (%v)", err)
+		return sdk.StageStatusFailure
+	}
+	lp.Successf("Successfully loaded %d manifests", len(manifests))
+
+	if len(manifests) == 0 {
+		lp.Error("This application has no Kubernetes manifests to handle")
+		return sdk.StageStatusFailure
+	}
+
+	// Generate the manifests for applying.
+	lp.Info("Start generating manifests for PRIMARY variant")
+	primaryManifests, err := generatePrimaryManifests(appCfg, manifests, stageCfg, variantLabel, primaryVariant)
+	if err != nil {
+		lp.Errorf("Unable to generate manifests for PRIMARY variant (%v)", err)
+		return sdk.StageStatusFailure
+	}
+	lp.Successf("Successfully generated %d manifests for PRIMARY variant", len(primaryManifests))
+
+	addVariantLabelsAndAnnotations(primaryManifests, variantLabel, primaryVariant)
+
+	if err := annotateConfigHash(primaryManifests); err != nil {
+		lp.Errorf("Unable to set %q annotation into the workload manifest (%v)", provider.AnnotationConfigHash, err)
+		return sdk.StageStatusFailure
+	}
+
+	deployTargetConfig := dt.Config
+
+	// Resolve kubectl version: multiTarget > spec > deployTarget
+	kubectlVersion := cmp.Or(appCfg.Input.KubectlVersion, deployTargetConfig.KubectlVersion)
+	if multiTarget != nil {
+		kubectlVersion = cmp.Or(multiTarget.KubectlVersion, kubectlVersion)
+	}
+
+	kubectlPath, err := toolRegistry.Kubectl(ctx, kubectlVersion)
+	if err != nil {
+		lp.Errorf("Failed while getting kubectl tool (%v)", err)
+		return sdk.StageStatusFailure
+	}
+
+	kubectl := provider.NewKubectl(kubectlPath)
+	applier := provider.NewApplier(kubectl, appCfg.Input, deployTargetConfig, input.Logger)
+
+	lp.Info("Start rolling out PRIMARY variant...")
+	if err := applyManifests(ctx, applier, primaryManifests, appCfg.Input.Namespace, lp); err != nil {
+		lp.Errorf("Failed while applying manifests (%v)", err)
+		return sdk.StageStatusFailure
+	}
+
+	if !stageCfg.Prune {
+		lp.Info("Resource GC was skipped because prune was not configured")
+		return sdk.StageStatusSuccess
+	}
+
+	// Wait for all applied manifests to be stable.
+	// In theory, we don't need to wait for them to be stable before going to the next step
+	// but waiting for a while reduces the number of Kubernetes changes in a short time.
+	lp.Info("Waiting for the applied manifests to be stable")
+	select {
+	case <-time.After(15 * time.Second):
+		break
+	case <-ctx.Done():
+		break
+	}
+
+	// Find the running resources that are not defined in Git.
+	lp.Info("Start finding all running PRIMARY resources but no longer defined in Git")
+	namespacedLiveResources, clusterScopedLiveResources, err := provider.GetLiveResources(ctx, kubectl, deployTargetConfig.KubeConfigPath, input.Request.Deployment.ApplicationID, fmt.Sprintf("%s=%s", variantLabel, primaryVariant))
+	if err != nil {
+		lp.Errorf("Failed while getting live resources (%v)", err)
+		return sdk.StageStatusFailure
+	}
+
+	if len(namespacedLiveResources)+len(clusterScopedLiveResources) == 0 {
+		lp.Info("There is no data about live resource so no resource will be removed")
+		return sdk.StageStatusSuccess
+	}
+
+	lp.Successf("Successfully loaded %d live resources", len(namespacedLiveResources)+len(clusterScopedLiveResources))
+
+	removeKeys := provider.FindRemoveResources(primaryManifests, namespacedLiveResources, clusterScopedLiveResources)
+	if len(removeKeys) == 0 {
+		lp.Info("There are no live resources should be removed")
+		return sdk.StageStatusSuccess
+	}
+
+	lp.Infof("Start pruning %d resources", len(removeKeys))
+	deletedCount := deleteResources(ctx, lp, applier, removeKeys)
+	lp.Successf("Successfully deleted %d resources", deletedCount)
+
+	return sdk.StageStatusSuccess
+}
+
+// generatePrimaryManifests generates manifests for the PRIMARY variant.
+// It deep-copies the input manifests, adds the variant label to workload selectors
+// if requested, and generates a variant Service manifest if requested.
+func generatePrimaryManifests(appCfg *kubeconfig.KubernetesApplicationSpec, manifests []provider.Manifest, stageCfg kubeconfig.K8sPrimaryRolloutStageOptions, variantLabel, variant string) ([]provider.Manifest, error) {
+	suffix := variant
+	if stageCfg.Suffix != "" {
+		suffix = stageCfg.Suffix
+	}
+
+	primaryManifests := provider.DeepCopyManifests(manifests)
+
+	// Add the variant label to workload selectors if requested.
+	if stageCfg.AddVariantLabelToSelector {
+		workloads := findWorkloadManifests(primaryManifests, nil)
+		for _, m := range workloads {
+			if err := ensureVariantSelectorInWorkload(m, variantLabel, variant); err != nil {
+				return nil, fmt.Errorf("unable to check/set %q in selector of workload %s (%w)", variantLabel+": "+variant, m.Key().ReadableString(), err)
+			}
+		}
+	}
+
+	// Generate Service manifests for the PRIMARY variant if requested.
+	if stageCfg.CreateService {
+		serviceName := appCfg.Service.Name
+		services := findManifests(provider.KindService, serviceName, primaryManifests)
+		if len(services) == 0 {
+			return nil, fmt.Errorf("unable to find any service for PRIMARY variant")
+		}
+		// Deep-copy the services to avoid mutating the shared primaryManifests slice entries.
+		services = provider.DeepCopyManifests(services)
+
+		generatedServices, err := generateVariantServiceManifests(services, variantLabel, variant, suffix)
+		if err != nil {
+			return nil, fmt.Errorf("failed to generate service manifests: %w", err)
+		}
+		primaryManifests = append(primaryManifests, generatedServices...)
+	}
+
+	return primaryManifests, nil
+}