feat: implement monitoring bootstrap

Makefile

@@ -181,7 +181,7 @@ endif
 
 .PHONY: install
 install: manifests kustomize ## Install CRDs into the K8s cluster specified in ~/.kube/config.
-	$(KUSTOMIZE) build config/crd | kubectl apply -f -
+	$(KUSTOMIZE) build config/crd | kubectl apply -f - --server-side
 
 .PHONY: uninstall
 uninstall: manifests kustomize ## Uninstall CRDs from the K8s cluster specified in ~/.kube/config. Call with ignore-not-found=true to ignore resource not found errors during deletion.
@@ -190,7 +190,7 @@ uninstall: manifests kustomize ## Uninstall CRDs from the K8s cluster specified
 .PHONY: deploy
 deploy: manifests kustomize ## Deploy controller to the K8s cluster specified in ~/.kube/config.
 	cd config/manager && $(KUSTOMIZE) edit set image controller=${IMAGE_REPO}/greptimedb-operator:${IMAGE_TAG}
-	$(KUSTOMIZE) build config/default | kubectl apply -f -
+	$(KUSTOMIZE) build config/default | kubectl apply -f - --server-side
 
 .PHONY: undeploy
 undeploy: ## Undeploy controller from the K8s cluster specified in ~/.kube/config. Call with ignore-not-found=true to ignore resource not found errors during deletion.

apis/v1alpha1/common.go

@@ -650,6 +650,13 @@ func (in *S3Storage) GetSecretName() string {
 	return ""
 }
 
+func (in *S3Storage) GetRoot() string {
+	if in != nil {
+		return in.Root
+	}
+	return ""
+}
+
 // OSSStorage defines the Aliyun OSS storage specification.
 type OSSStorage struct {
 	// The data will be stored in the bucket.
@@ -682,6 +689,13 @@ func (in *OSSStorage) GetSecretName() string {
 	return ""
 }
 
+func (in *OSSStorage) GetRoot() string {
+	if in != nil {
+		return in.Root
+	}
+	return ""
+}
+
 // GCSStorage defines the Google GCS storage specification.
 type GCSStorage struct {
 	// The data will be stored in the bucket.
@@ -714,6 +728,13 @@ func (in *GCSStorage) GetSecretName() string {
 	return ""
 }
 
+func (in *GCSStorage) GetRoot() string {
+	if in != nil {
+		return in.Root
+	}
+	return ""
+}
+
 // PrometheusMonitorSpec defines the PodMonitor configuration.
 type PrometheusMonitorSpec struct {
 	// Enabled indicates whether the PodMonitor is enabled.

apis/v1alpha1/constants.go

@@ -19,8 +19,8 @@ const (
 	// DefaultVersion is the default version of the GreptimeDB.
 	DefaultVersion = "Unknown"
 
-	// DefautlHealthEndpoint is the default health endpoint for the liveness probe.
-	DefautlHealthEndpoint = "/health"
+	// DefaultHealthEndpoint is the default health endpoint for the liveness probe.
+	DefaultHealthEndpoint = "/health"
 
 	// DefaultHTTPPort is the default HTTP port for the GreptimeDB.
 	DefaultHTTPPort int32 = 4000
@@ -58,8 +58,20 @@ const (
 	// DefaultInitializerImage is the default image for the GreptimeDB initializer.
 	DefaultInitializerImage = "greptime/greptimedb-initializer:latest"
 
+	// DefaultGreptimeDBImage is the default image for the GreptimeDB.
+	DefaultGreptimeDBImage = "greptime/greptimedb:latest"
+
 	// DefaultLoggingLevel is the default logging level for the GreptimeDB.
 	DefaultLoggingLevel = LoggingLevelInfo
+
+	// DefaultVectorImage is the default image for the vector.
+	DefaultVectorImage = "timberio/vector:nightly-alpine"
+
+	// DefaultVectorCPURequest is the default CPU request for the vector.
+	DefaultVectorCPURequest = "50m"
+
+	// DefaultVectorMemoryRequest is the default memory request for the vector.
+	DefaultVectorMemoryRequest = "128Mi"
 )
 
 // The following constants are the constant configuration for the GreptimeDBCluster and GreptimeDBStandalone.

apis/v1alpha1/defaulting.go

@@ -15,10 +15,12 @@
 package v1alpha1
 
 import (
+	"fmt"
 	"strings"
 
 	"dario.cat/mergo"
 	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
 	"k8s.io/apimachinery/pkg/util/intstr"
 	"k8s.io/utils/pointer"
 )
@@ -43,6 +45,11 @@ func (in *GreptimeDBCluster) SetDefaults() error {
 		return err
 	}
 
+	// Merge the logging settings into the GreptimeDBClusterSpec.
+	if err := in.mergeLogging(); err != nil {
+		return err
+	}
+
 	return nil
 }
 
@@ -54,7 +61,7 @@ func (in *GreptimeDBCluster) defaultSpec() *GreptimeDBClusterSpec {
 				LivenessProbe: &corev1.Probe{
 					ProbeHandler: corev1.ProbeHandler{
 						HTTPGet: &corev1.HTTPGetAction{
-							Path: DefautlHealthEndpoint,
+							Path: DefaultHealthEndpoint,
 							Port: intstr.FromInt32(DefaultHTTPPort),
 						},
 					},
@@ -78,6 +85,29 @@ func (in *GreptimeDBCluster) defaultSpec() *GreptimeDBClusterSpec {
 
 	defaultSpec.Logging = defaultLogging()
 
+	if in.GetMonitoring().IsEnabled() {
+		defaultSpec.Monitoring = &MonitoringSpec{
+			Standalone:     in.defaultMonitoringStandaloneSpec(),
+			LogsCollection: &LogsCollectionSpec{},
+			Vector: &VectorSpec{
+				Image: DefaultVectorImage,
+				Resource: corev1.ResourceRequirements{
+					Requests: corev1.ResourceList{
+						corev1.ResourceCPU:    resource.MustParse(DefaultVectorCPURequest),
+						corev1.ResourceMemory: resource.MustParse(DefaultVectorMemoryRequest),
+					},
+					Limits: corev1.ResourceList{
+						corev1.ResourceCPU:    resource.MustParse(DefaultVectorCPURequest),
+						corev1.ResourceMemory: resource.MustParse(DefaultVectorMemoryRequest),
+					},
+				},
+			},
+		}
+
+		// Set the default logging format to JSON if monitoring is enabled.
+		defaultSpec.Logging.Format = LogFormatJSON
+	}
+
 	return defaultSpec
 }
 
@@ -86,6 +116,7 @@ func (in *GreptimeDBCluster) defaultFrontend() *FrontendSpec {
 		ComponentSpec: ComponentSpec{
 			Template: &PodTemplateSpec{},
 			Replicas: pointer.Int32(DefaultReplicas),
+			Logging:  &LoggingSpec{},
 		},
 		RPCPort:        DefaultRPCPort,
 		HTTPPort:       DefaultHTTPPort,
@@ -106,6 +137,7 @@ func (in *GreptimeDBCluster) defaultMeta() *MetaSpec {
 		ComponentSpec: ComponentSpec{
 			Template: &PodTemplateSpec{},
 			Replicas: pointer.Int32(DefaultReplicas),
+			Logging:  &LoggingSpec{},
 		},
 		RPCPort:              DefaultMetaRPCPort,
 		HTTPPort:             DefaultHTTPPort,
@@ -118,6 +150,7 @@ func (in *GreptimeDBCluster) defaultDatanode() *DatanodeSpec {
 		ComponentSpec: ComponentSpec{
 			Template: &PodTemplateSpec{},
 			Replicas: pointer.Int32(DefaultReplicas),
+			Logging:  &LoggingSpec{},
 		},
 		RPCPort:  DefaultRPCPort,
 		HTTPPort: DefaultHTTPPort,
@@ -130,11 +163,43 @@ func (in *GreptimeDBCluster) defaultFlownodeSpec() *FlownodeSpec {
 		ComponentSpec: ComponentSpec{
 			Template: &PodTemplateSpec{},
 			Replicas: pointer.Int32(DefaultReplicas),
+			Logging:  &LoggingSpec{},
 		},
 		RPCPort: DefaultRPCPort,
 	}
 }
 
+func (in *GreptimeDBCluster) defaultMonitoringStandaloneSpec() *GreptimeDBStandaloneSpec {
+	standalone := new(GreptimeDBStandalone)
+	standalone.Spec = *standalone.defaultSpec()
+
+	if image := in.GetBaseMainContainer().GetImage(); image != "" {
+		standalone.Spec.Base.MainContainer.Image = image
+	} else {
+		standalone.Spec.Base.MainContainer.Image = DefaultGreptimeDBImage
+	}
+
+	standalone.Spec.Version = getVersionFromImage(standalone.Spec.Base.MainContainer.Image)
+
+	if osp := in.GetObjectStorageProvider(); osp != nil {
+		standalone.Spec.ObjectStorageProvider = osp.DeepCopy()
+
+		if root := osp.GetS3Storage().GetRoot(); root != "" {
+			standalone.Spec.ObjectStorageProvider.S3.Root = fmt.Sprintf("%s/monitoring", root)
+		}
+
+		if root := osp.GetOSSStorage().GetRoot(); root != "" {
+			standalone.Spec.ObjectStorageProvider.OSS.Root = fmt.Sprintf("%s/monitoring", root)
+		}
+
+		if root := osp.GetGCSStorage().GetRoot(); root != "" {
+			standalone.Spec.ObjectStorageProvider.GCS.Root = fmt.Sprintf("%s/monitoring", root)
+		}
+	}
+
+	return &standalone.Spec
+}
+
 func (in *GreptimeDBCluster) mergeTemplate() error {
 	if err := in.mergeFrontendTemplate(); err != nil {
 		return err
@@ -211,6 +276,50 @@ func (in *GreptimeDBCluster) mergeFlownodeTemplate() error {
 	return nil
 }
 
+func (in *GreptimeDBCluster) mergeLogging() error {
+	if logging := in.GetMeta().GetLogging(); logging != nil {
+		if err := mergo.Merge(logging, in.GetLogging().DeepCopy()); err != nil {
+			return err
+		}
+		if in.GetMonitoring().IsEnabled() {
+			// Set the default logging format to JSON if monitoring is enabled.
+			logging.Format = LogFormatJSON
+		}
+	}
+
+	if logging := in.GetDatanode().GetLogging(); logging != nil {
+		if err := mergo.Merge(logging, in.GetLogging().DeepCopy()); err != nil {
+			return err
+		}
+		if in.GetMonitoring().IsEnabled() {
+			// Set the default logging format to JSON if monitoring is enabled.
+			logging.Format = LogFormatJSON
+		}
+	}
+
+	if logging := in.GetFrontend().GetLogging(); logging != nil {
+		if err := mergo.Merge(logging, in.GetLogging().DeepCopy()); err != nil {
+			return err
+		}
+		if in.GetMonitoring().IsEnabled() {
+			// Set the default logging format to JSON if monitoring is enabled.
+			logging.Format = LogFormatJSON
+		}
+	}
+
+	if logging := in.GetFlownode().GetLogging(); logging != nil {
+		if err := mergo.Merge(logging, in.GetLogging().DeepCopy()); err != nil {
+			return err
+		}
+		if in.GetMonitoring().IsEnabled() {
+			// Set the default logging format to JSON if monitoring is enabled.
+			logging.Format = LogFormatJSON
+		}
+	}
+
+	return nil
+}
+
 func (in *GreptimeDBStandalone) SetDefaults() error {
 	if in == nil {
 		return nil
@@ -235,7 +344,7 @@ func (in *GreptimeDBStandalone) defaultSpec() *GreptimeDBStandaloneSpec {
 				LivenessProbe: &corev1.Probe{
 					ProbeHandler: corev1.ProbeHandler{
 						HTTPGet: &corev1.HTTPGetAction{
-							Path: DefautlHealthEndpoint,
+							Path: DefaultHealthEndpoint,
 							Port: intstr.FromInt32(DefaultHTTPPort),
 						},
 					},