refactor: rewrite activestandby task to remove external dioscuri requirement

DEPRECATIONS.md

@@ -12,7 +12,7 @@ All deprecations are listed below, with the most recent announcements at the top
 
 ### Lagoon v2.17.0
 release link: https://github.com/uselagoon/lagoon/releases/tag/v2.17.0
-* (insert any planned deprecations here)
+* This release introduces a new active/standby task image that does not require the use of the [dioscuri controller](https://github.com/amazeeio/dioscuri). Dioscuri is deprecated and will eventually be removed from the `lagoon-remote` helm chart. If you use active/standby functionality in your clusters, you should upgrade to lagoon v2.17.0 and update your remote clusters to the version of the `lagoon-remote` helm chart the v2.17.0 release says to use (see release notes for v2.17.0)
 
 ### Lagoon v2.16.0
 release link: https://github.com/uselagoon/lagoon/releases/tag/v2.16.0

Makefile

@@ -52,6 +52,11 @@ UPSTREAM_TAG ?= latest
 # edge is the most current merged change
 BUILD_DEPLOY_IMAGE_TAG ?= edge
 
+# OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG and OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY
+# set this to a particular build image if required, defaults to nothing to consume what the chart provides
+OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG=
+OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY=
+
 # To build k3d with Calico instead of Flannel, set this to true. Note that the Calico install in lagoon-charts is always
 # disabled for use with k3d, as the cluster needs it on creation.
 USE_CALICO_CNI ?= false
@@ -499,6 +504,8 @@ k3d/test: k3d/cluster helm/repos $(addprefix local-dev/,$(K3D_TOOLS)) build
 			HELM=$$(realpath ../local-dev/helm) KUBECTL=$$(realpath ../local-dev/kubectl) \
 			JQ=$$(realpath ../local-dev/jq) \
 			OVERRIDE_BUILD_DEPLOY_DIND_IMAGE=uselagoon/build-deploy-image:${BUILD_DEPLOY_IMAGE_TAG} \
+			$$([ $(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG) ] && echo 'OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG=$(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG)') \
+			$$([ $(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY) ] && echo 'OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY=$(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY)') \
 			OVERRIDE_ACTIVE_STANDBY_TASK_IMAGE=$$IMAGE_REGISTRY/task-activestandby:$(SAFE_BRANCH_NAME) \
 			IMAGE_REGISTRY=$$IMAGE_REGISTRY \
 			SKIP_INSTALL_REGISTRY=true \
@@ -531,6 +538,8 @@ k3d/setup: k3d/cluster helm/repos $(addprefix local-dev/,$(K3D_TOOLS)) build
 			HELM=$$(realpath ../local-dev/helm) KUBECTL=$$(realpath ../local-dev/kubectl) \
 			JQ=$$(realpath ../local-dev/jq) \
 			OVERRIDE_BUILD_DEPLOY_DIND_IMAGE=uselagoon/build-deploy-image:${BUILD_DEPLOY_IMAGE_TAG} \
+			$$([ $(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG) ] && echo 'OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG=$(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG)') \
+			$$([ $(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY) ] && echo 'OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY=$(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY)') \
 			OVERRIDE_ACTIVE_STANDBY_TASK_IMAGE=$$IMAGE_REGISTRY/task-activestandby:$(SAFE_BRANCH_NAME) \
 			IMAGE_REGISTRY=$$IMAGE_REGISTRY
 
@@ -584,6 +593,8 @@ k3d/dev: build
 			HELM=$$(realpath ../local-dev/helm) KUBECTL=$$(realpath ../local-dev/kubectl) \
 			JQ=$$(realpath ../local-dev/jq) \
 			OVERRIDE_BUILD_DEPLOY_DIND_IMAGE=uselagoon/build-deploy-image:${BUILD_DEPLOY_IMAGE_TAG} \
+			$$([ $(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG) ] && echo 'OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG=$(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGETAG)') \
+			$$([ $(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY) ] && echo 'OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY=$(OVERRIDE_BUILD_DEPLOY_CONTROLLER_IMAGE_REPOSITORY)') \
 			OVERRIDE_ACTIVE_STANDBY_TASK_IMAGE=$$IMAGE_REGISTRY/task-activestandby:$(SAFE_BRANCH_NAME) \
 			IMAGE_REGISTRY=$$IMAGE_REGISTRY
 

node-packages/commons/src/tasks.ts

@@ -1081,28 +1081,6 @@ const restoreConfig = (name, backupId, backupS3Config, restoreS3Config) => {
   return config;
 };
 
-// creates the route/ingress migration config
-const migrateHosts = (destinationNamespace, sourceNamespace) => {
-  const randId = Math.random().toString(36).substring(7);
-  const migrateName = `host-migration-${randId}`;
-  let config = {
-    apiVersion: 'dioscuri.amazee.io/v1',
-    kind: 'HostMigration',
-    metadata: {
-      name: migrateName,
-      annotations: {
-          'dioscuri.amazee.io/migrate':'true'
-      }
-    },
-    spec: {
-      destinationNamespace: destinationNamespace,
-      activeEnvironment: sourceNamespace,
-    },
-  };
-
-  return config;
-};
-
 export const getTaskProjectEnvironmentVariables =async (projectName: string, environmentId: number) => {
   // inject variables into tasks the same way it is in builds
   // this makes variables available to tasks the same way for consumption
@@ -1323,18 +1301,15 @@ export const createMiscTask = async function(taskData: any) {
           const restoreBytes = new Buffer(JSON.stringify(restoreConf).replace(/\\n/g, "\n")).toString('base64')
           miscTaskData.misc.miscResource = restoreBytes
           break;
-        case 'deploytarget:route:migrate':
+        case 'deploytarget:task:activestandby':
           // handle setting up the task configuration for running the active/standby switch
           // this uses the `advanced task` system in the controllers
-          // first generate the migration CRD
-          const migrateConf = migrateHosts(
-            makeSafe(taskData.data.productionEnvironment.openshiftProjectName),
-            makeSafe(taskData.data.environment.openshiftProjectName))
           // generate out custom json payload to send to the advanced task
           var jsonPayload: any = {
             productionEnvironment: taskData.data.productionEnvironment.name,
             standbyEnvironment: taskData.data.environment.name,
-            crd: migrateConf
+            sourceNamespace: makeSafe(taskData.data.environment.openshiftProjectName),
+            destinationNamespace: makeSafe(taskData.data.productionEnvironment.openshiftProjectName)
           }
           // encode it
           const jsonPayloadBytes = new Buffer(JSON.stringify(jsonPayload).replace(/\\n/g, "\n")).toString('base64')

services/actions-handler/handler/controller_tasks.go

@@ -34,7 +34,7 @@ func (m *Messenger) handleTask(ctx context.Context, messageQueue *mq.MessageQueu
 	l := lclient.New(m.LagoonAPI.Endpoint, "actions-handler", &token, false)
 
 	switch message.Meta.Key {
-	case "kubernetes:route:migrate", "deploytarget:route:migrate":
+	case "kubernetes:route:migrate", "deploytarget:route:migrate", "deploytarget:task:activestandby":
 		// if the result is from an active/standby task, handle updating the project here
 		switch message.Meta.JobStatus {
 		case "complete", "succeeded":
@@ -58,11 +58,17 @@ func (m *Messenger) handleTask(ctx context.Context, messageQueue *mq.MessageQueu
 			json.Unmarshal(decodeData, advTask)
 			// then prepare the patch operation
 			updateProject := schema.UpdateProjectPatchInput{
-				ProductionEnvironment:        &advTask.StandbyProductionEnvironment, // these are inverted because of how the task works
-				StandbyProductionEnvironment: &advTask.ProductionEnvironment,        // these are inverted because of how the task works
+				ProductionEnvironment:        &advTask.ProductionEnvironment,
+				StandbyProductionEnvironment: &advTask.StandbyProductionEnvironment,
 				ProductionRoutes:             &advTask.ProductionRoutes,
 				StandbyRoutes:                &advTask.StandbyRoutes,
 			}
+			switch message.Meta.Key {
+			case "kubernetes:route:migrate", "deploytarget:route:migrate":
+				// the old task had these inverted, so this keeps that inversion in place for now
+				updateProject.ProductionEnvironment = &advTask.StandbyProductionEnvironment
+				updateProject.StandbyProductionEnvironment = &advTask.ProductionEnvironment
+			}
 			// update the project in the api
 			updatedProject, err := lagoon.UpdateProject(ctx, int(project.ID), updateProject, l)
 			if err != nil {

services/api/src/resources/deployment/resolvers.ts

@@ -1247,7 +1247,7 @@ export const switchActiveStandby: ResolverFn = async (
     data.task.id = sourceTaskData.addTask.id.toString();
 
     // queue the task to trigger the migration
-    await createMiscTask({ key: 'route:migrate', data });
+    await createMiscTask({ key: 'task:activestandby', data });
 
     // return the task id and remote id
     var retData = {

taskimages/activestandby/Dockerfile

@@ -5,6 +5,7 @@ COPY . /go/src/github.com/uselagoon/lagoon/taskimages/activestandby/
 WORKDIR /go/src/github.com/uselagoon/lagoon/taskimages/activestandby/
 
 # Build
+RUN CGO_ENABLED=0 GOOS=linux GOARCH=${ARCH} go test -v ./...
 RUN CGO_ENABLED=0 GOOS=linux GOARCH=${ARCH} go build -a -o taskrunner .
 
 # Use distroless as minimal base image to package the binary

taskimages/activestandby/README.md

@@ -1,3 +1,11 @@
 # tasks-activestandby
 
-This image is used by the activestandby task when using the Lagoon Kubernetes Controllers
+This image is used by the activestandby task. The remote-controller has knowledge of this task and will create a role binding between the two namespaces to allow them to temporarily talk and create/edit/delete resources between them as part of the task process.
+
+The resulting payload contains information about the actions that were performed, which are sent back to Lagoon via the message queue to be reflected in the API.
+
+The basic idea is that when activestandby is triggered, it collects the ingress in both namespaces that match the labels `dioscuri.amazee.io/migrate=true` and `activestandby.lagoon.sh/migrate=true` and will then perform the action of storing information about them, removing them from the source namespace, and then creating them in the destination namespace.
+
+Part of this process also involves copying secrets and certificates if they are present, so that they are also available in the destination namespace.
+
+When the process of migrating is taking place, all ingress have a new label added which is `activestandby.lagoon.sh/migrating=true`, which at the end of the migration process is set to `false`. This label will only be true while migrations are taking place. This allows external systems to be aware of the migration if they need to take any action, or prevent some actions from taking place.

taskimages/activestandby/dioscuri/certificates.go

@@ -0,0 +1,43 @@
+package dioscuri
+
+import (
+	"context"
+	"fmt"
+
+	certv1 "github.com/cert-manager/cert-manager/pkg/apis/certmanager/v1"
+	networkv1 "k8s.io/api/networking/v1"
+	"k8s.io/apimachinery/pkg/types"
+	client "sigs.k8s.io/controller-runtime/pkg/client"
+)
+
+// copy any certificate into a slice of certificates
+func copyCertificates(ctx context.Context, c client.Client, ingress *networkv1.Ingress) []*certv1.Certificate {
+	var certificates []*certv1.Certificate
+	for _, tls := range ingress.Spec.TLS {
+		certificate := &certv1.Certificate{}
+		err := c.Get(ctx, types.NamespacedName{Namespace: ingress.ObjectMeta.Namespace, Name: tls.SecretName}, certificate)
+		if err != nil {
+			break
+		}
+		certificates = append(certificates, certificate)
+		fmt.Println(fmt.Sprintf(">> Copying certificate %s in namespace %s", certificate.ObjectMeta.Name, certificate.ObjectMeta.Namespace))
+	}
+	return certificates
+}
+
+// create any certificates in the destination namespace
+func createCertificates(ctx context.Context, c client.Client, destinationNamespace string, certificates []*certv1.Certificate) error {
+	for _, certificate := range certificates {
+		certificate.ObjectMeta.Namespace = destinationNamespace
+		certificate.ResourceVersion = ""
+		certificate.SelfLink = ""
+		certificate.UID = ""
+		err := c.Create(ctx, certificate)
+		if err != nil {
+			break
+		}
+		// secrets = append(secrets, certificate)
+		fmt.Println(fmt.Sprintf(">> Creating certificate %s in namespace %s", certificate.ObjectMeta.Name, certificate.ObjectMeta.Namespace))
+	}
+	return nil
+}