Merge pull request #184 from slintes/metal3_remediation

Metal3 remediation backport

Dockerfile

@@ -11,3 +11,6 @@ FROM registry.ci.openshift.org/ocp/4.13:base
 #    rpm -V $INSTALL_PKGS && \
 #    yum clean all
 COPY --from=builder /go/src/github.com/openshift/cluster-api-provider-baremetal/machine-controller-manager /
+COPY --from=builder /go/src/github.com/openshift/cluster-api-provider-baremetal/manifests /manifests
+
+LABEL io.openshift.release.operator=true

Makefile

@@ -3,6 +3,10 @@
 IMG ?= controller:latest
 GOPATH=$(shell go env GOPATH)
 
+MOCKGEN := $(GOPATH)/bin/mockgen
+$(MOCKGEN): # Build mockgen
+	go build -tags=tools -o $(GOPATH)/bin github.com/golang/mock/mockgen
+
 .PHONY: build
 build:
 	@mkdir -p bin
@@ -43,9 +47,23 @@ vet:
 
 # Generate code
 .PHONY: generate
-generate:
+generate: $(MOCKGEN)
 	go generate ./pkg/... ./cmd/...
 
+	$(MOCKGEN) \
+	  -destination=./pkg/baremetal/mocks/zz_generated.metal3remediation_manager.go \
+	  -source=./pkg/baremetal/metal3remediation_manager.go \
+		-package=baremetal_mocks \
+		-copyright_file=./hack/boilerplate.go.txt \
+		RemediationManagerInterface
+
+	$(MOCKGEN) \
+	  -destination=./pkg/baremetal/mocks/zz_generated.manager_factory.go \
+	  -source=./pkg/baremetal/manager_factory.go \
+		-package=baremetal_mocks \
+		-copyright_file=./hack/boilerplate.go.txt \
+		ManagerFactoryInterface
+
 .PHONY: generate-check
 generate-check:
 	./hack/generate.sh
@@ -67,3 +85,7 @@ vendor:
 	go mod tidy
 	go mod vendor
 	go mod verify
+
+.PHONY: metal3-crds
+metal3-crds:
+	./hack/fetch-metal3-crds.sh

cmd/manager/main.go

@@ -23,24 +23,35 @@ import (
 	"time"
 
 	bmoapis "github.com/metal3-io/baremetal-operator/apis/metal3.io/v1alpha1"
+	capm3apis "github.com/metal3-io/cluster-api-provider-metal3/api/v1beta1"
 	machinev1beta1 "github.com/openshift/api/machine/v1beta1"
 	"github.com/openshift/cluster-api-provider-baremetal/pkg/apis"
+	"github.com/openshift/cluster-api-provider-baremetal/pkg/baremetal"
 	"github.com/openshift/cluster-api-provider-baremetal/pkg/cloud/baremetal/actuators/machine"
 	"github.com/openshift/cluster-api-provider-baremetal/pkg/controller"
+	"github.com/openshift/cluster-api-provider-baremetal/pkg/controller/metal3remediation"
 	"github.com/openshift/cluster-api-provider-baremetal/pkg/manager/wrapper"
 	maomachine "github.com/openshift/machine-api-operator/pkg/controller/machine"
 	"k8s.io/apimachinery/pkg/runtime/schema"
 	"k8s.io/client-go/discovery"
 	"k8s.io/client-go/rest"
 	"k8s.io/klog/v2"
 	"k8s.io/klog/v2/klogr"
+	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
 	"sigs.k8s.io/controller-runtime/pkg/healthz"
 	logf "sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/manager"
 	"sigs.k8s.io/controller-runtime/pkg/manager/signals"
 )
 
+const (
+	// Align with MachineWebhookPort in https://github.com/openshift/machine-api-operator/blob/master/pkg/operator/sync.go
+	// or import from there when relevant change was merged
+	defaultWebhookPort    = 8440
+	defaultWebhookCertdir = "/etc/machine-api-operator/tls"
+)
+
 // The default durations for the leader election operations.
 var (
 	leaseDuration = 120 * time.Second
@@ -87,6 +98,16 @@ func main() {
 		leaseDuration,
 		"The duration that non-leader candidates will wait after observing a leadership renewal until attempting to acquire leadership of a led but unrenewed leader slot. This is effectively the maximum duration that a leader can be stopped before it is replaced by another candidate. This is only applicable if leader election is enabled.",
 	)
+
+	webhookEnabled := flag.Bool("webhook-enabled", true,
+		"Webhook server, enabled by default. When enabled, the manager will run a webhook server.")
+
+	webhookPort := flag.Int("webhook-port", defaultWebhookPort,
+		"Webhook Server port, only used when webhook-enabled is true.")
+
+	webhookCertdir := flag.String("webhook-cert-dir", defaultWebhookCertdir,
+		"Webhook cert dir, only used when webhook-enabled is true.")
+
 	flag.Parse()
 
 	log := logf.Log.WithName("baremetal-controller-manager")
@@ -146,6 +167,10 @@ func main() {
 		panic(err)
 	}
 
+	if err := capm3apis.AddToScheme(mgr.GetScheme()); err != nil {
+		panic(err)
+	}
+
 	// the manager wrapper will add an extra Watch to the controller
 	maomachine.AddWithActuator(wrapper.New(mgr), machineActuator)
 
@@ -154,6 +179,33 @@ func main() {
 		os.Exit(1)
 	}
 
+	// Set up the context that's going to be used in controllers and for the manager.
+	ctx := signals.SetupSignalHandler()
+
+	if err := (&metal3remediation.Metal3RemediationReconciler{
+		Client:         mgr.GetClient(),
+		ManagerFactory: baremetal.NewManagerFactory(mgr.GetClient()),
+		Log:            ctrl.Log.WithName("controllers").WithName("Metal3Remediation"),
+	}).SetupWithManager(ctx, mgr); err != nil {
+		log.Error(err, "unable to create controller", "controller", "Metal3Remediation")
+		os.Exit(1)
+	}
+
+	if *webhookEnabled {
+		mgr.GetWebhookServer().Port = *webhookPort
+		mgr.GetWebhookServer().CertDir = *webhookCertdir
+
+		if err := (&capm3apis.Metal3Remediation{}).SetupWebhookWithManager(mgr); err != nil {
+			log.Error(err, "unable to create webhook", "webhook", "Metal3Remediation")
+			os.Exit(1)
+		}
+
+		if err := (&capm3apis.Metal3RemediationTemplate{}).SetupWebhookWithManager(mgr); err != nil {
+			log.Error(err, "unable to create webhook", "webhook", "Metal3RemediationTemplate")
+			os.Exit(1)
+		}
+	}
+
 	if err := mgr.AddReadyzCheck("ping", healthz.Ping); err != nil {
 		klog.Fatal(err)
 	}
@@ -162,7 +214,7 @@ func main() {
 		klog.Fatal(err)
 	}
 
-	if err := mgr.Start(signals.SetupSignalHandler()); err != nil {
+	if err := mgr.Start(ctx); err != nil {
 		entryLog.Error(err, "unable to run manager")
 		os.Exit(1)
 	}

config/crd/infrastructure.cluster.x-k8s.io_metal3remediations.yaml

@@ -0,0 +1,178 @@
+---
+apiVersion: apiextensions.k8s.io/v1
+kind: CustomResourceDefinition
+metadata:
+  annotations:
+    controller-gen.kubebuilder.io/version: v0.9.1
+  creationTimestamp: null
+  name: metal3remediations.infrastructure.cluster.x-k8s.io
+spec:
+  group: infrastructure.cluster.x-k8s.io
+  names:
+    categories:
+    - cluster-api
+    kind: Metal3Remediation
+    listKind: Metal3RemediationList
+    plural: metal3remediations
+    shortNames:
+    - m3r
+    - m3remediation
+    singular: metal3remediation
+  scope: Namespaced
+  versions:
+  - additionalPrinterColumns:
+    - description: How many times remediation controller should attempt to remediate
+        the host
+      jsonPath: .spec.strategy.retryLimit
+      name: Retry limit
+      type: string
+    - description: How many times remediation controller has tried to remediate the
+        node
+      jsonPath: .status.retryCount
+      name: Retry count
+      type: string
+    - description: Timestamp of the last remediation attempt
+      jsonPath: .status.lastRemediated
+      name: Last Remediated
+      type: string
+    - description: Type of the remediation strategy
+      jsonPath: .spec.strategy.type
+      name: Strategy
+      type: string
+    - description: Phase of the remediation
+      jsonPath: .status.phase
+      name: Phase
+      type: string
+    name: v1alpha5
+    schema:
+      openAPIV3Schema:
+        description: Metal3Remediation is the Schema for the metal3remediations API.
+        properties:
+          apiVersion:
+            description: 'APIVersion defines the versioned schema of this representation
+              of an object. Servers should convert recognized schemas to the latest
+              internal value, and may reject unrecognized values. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#resources'
+            type: string
+          kind:
+            description: 'Kind is a string value representing the REST resource this
+              object represents. Servers may infer this from the endpoint the client
+              submits requests to. Cannot be updated. In CamelCase. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds'
+            type: string
+          metadata:
+            type: object
+          spec:
+            description: Metal3RemediationSpec defines the desired state of Metal3Remediation.
+            properties:
+              strategy:
+                description: Strategy field defines remediation strategy.
+                properties:
+                  retryLimit:
+                    description: Sets maximum number of remediation retries.
+                    type: integer
+                  timeout:
+                    description: Sets the timeout between remediation retries.
+                    type: string
+                  type:
+                    description: Type of remediation.
+                    type: string
+                type: object
+            type: object
+          status:
+            description: Metal3RemediationStatus defines the observed state of Metal3Remediation.
+            properties:
+              lastRemediated:
+                description: LastRemediated identifies when the host was last remediated
+                format: date-time
+                type: string
+              phase:
+                description: Phase represents the current phase of machine remediation.
+                  E.g. Pending, Running, Done etc.
+                type: string
+              retryCount:
+                description: RetryCount can be used as a counter during the remediation.
+                  Field can hold number of reboots etc.
+                type: integer
+            type: object
+        type: object
+    served: true
+    storage: false
+    subresources:
+      status: {}
+  - additionalPrinterColumns:
+    - description: How many times remediation controller should attempt to remediate
+        the host
+      jsonPath: .spec.strategy.retryLimit
+      name: Retry limit
+      type: string
+    - description: How many times remediation controller has tried to remediate the
+        node
+      jsonPath: .status.retryCount
+      name: Retry count
+      type: string
+    - description: Timestamp of the last remediation attempt
+      jsonPath: .status.lastRemediated
+      name: Last Remediated
+      type: string
+    - description: Type of the remediation strategy
+      jsonPath: .spec.strategy.type
+      name: Strategy
+      type: string
+    - description: Phase of the remediation
+      jsonPath: .status.phase
+      name: Phase
+      type: string
+    name: v1beta1
+    schema:
+      openAPIV3Schema:
+        description: Metal3Remediation is the Schema for the metal3remediations API.
+        properties:
+          apiVersion:
+            description: 'APIVersion defines the versioned schema of this representation
+              of an object. Servers should convert recognized schemas to the latest
+              internal value, and may reject unrecognized values. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#resources'
+            type: string
+          kind:
+            description: 'Kind is a string value representing the REST resource this
+              object represents. Servers may infer this from the endpoint the client
+              submits requests to. Cannot be updated. In CamelCase. More info: https://git.k8s.io/community/contributors/devel/sig-architecture/api-conventions.md#types-kinds'
+            type: string
+          metadata:
+            type: object
+          spec:
+            description: Metal3RemediationSpec defines the desired state of Metal3Remediation.
+            properties:
+              strategy:
+                description: Strategy field defines remediation strategy.
+                properties:
+                  retryLimit:
+                    description: Sets maximum number of remediation retries.
+                    type: integer
+                  timeout:
+                    description: Sets the timeout between remediation retries.
+                    type: string
+                  type:
+                    description: Type of remediation.
+                    type: string
+                type: object
+            type: object
+          status:
+            description: Metal3RemediationStatus defines the observed state of Metal3Remediation.
+            properties:
+              lastRemediated:
+                description: LastRemediated identifies when the host was last remediated
+                format: date-time
+                type: string
+              phase:
+                description: Phase represents the current phase of machine remediation.
+                  E.g. Pending, Running, Done etc.
+                type: string
+              retryCount:
+                description: RetryCount can be used as a counter during the remediation.
+                  Field can hold number of reboots etc.
+                type: integer
+            type: object
+        type: object
+    served: true
+    storage: true
+    subresources:
+      status: {}