RayService: Dev RayService CR and Controller logic (ray-project#287)

* Implement RayService CRD and controller * Update config and controller * update * update * refactor * goimports * goimports * address comments * address comments * address comments * address comments * address comments * address comments * address comments * add unit tests * fix ut * Update yaml * update ut * update
lowang-bh · Jun 9, 2022 · 5429751 · 5429751
1 parent a793574
commit 5429751
Show file tree

Hide file tree

Showing 14 changed files with 12,856 additions and 23 deletions.
diff --git a/ray-operator/Makefile b/ray-operator/Makefile
@@ -4,7 +4,8 @@ COMMIT_SHA1     := $(shell git rev-parse HEAD )
 # Image URL to use all building/pushing image targets
 IMG ?= controller:latest
 # Produce CRDs that work back to Kubernetes 1.11 (no version conversion)
-CRD_OPTIONS ?= "crd:maxDescLen=100,trivialVersions=true,preserveUnknownFields=false,generateEmbeddedObjectMeta=true"
+# Golang forbids converting float from Json to struct. Need allowDangerousTypes=true to unblock build.
+CRD_OPTIONS ?= "crd:maxDescLen=100,trivialVersions=true,preserveUnknownFields=false,generateEmbeddedObjectMeta=true,allowDangerousTypes=true"
 
 # Get the currently used golang install path (in GOPATH/bin, unless GOBIN is set)
 ifeq (,$(shell go env GOBIN))

diff --git a/ray-operator/apis/ray/v1alpha1/rayservice_types.go b/ray-operator/apis/ray/v1alpha1/rayservice_types.go
@@ -7,16 +7,80 @@ import (
 // EDIT THIS FILE!  THIS IS SCAFFOLDING FOR YOU TO OWN!
 // NOTE: json tags are required.  Any new fields you add must have json tags for the fields to be serialized.
 
+type ServiceStatus string
+
+const (
+	FailToGetOrCreateRayCluster  ServiceStatus = "FailToGetOrCreateRayCluster"
+	WaitForDashboard             ServiceStatus = "WaitForDashboard"
+	FailServeDeploy              ServiceStatus = "FailServeDeploy"
+	FailGetServeDeploymentStatus ServiceStatus = "FailGetServeDeploymentStatus"
+	Running                      ServiceStatus = "Running"
+	Restarting                   ServiceStatus = "Restarting"
+	FailDeleteRayCluster         ServiceStatus = "FailDeleteRayCluster"
+)
+
 // RayServiceSpec defines the desired state of RayService
 type RayServiceSpec struct {
-	// INSERT ADDITIONAL SPEC FIELDS - desired state of cluster
 	// Important: Run "make" to regenerate code after modifying this file
+	ServeConfigSpecs []ServeConfigSpec `json:"serveConfigs,omitempty"`
+	RayClusterSpec   RayClusterSpec    `json:"rayClusterConfig,omitempty"`
+}
+
+// ServeConfigSpec defines the desired state of RayService
+// Reference to https://docs.ray.io/en/latest/ray-core/package-ref.html#ray-remote.
+type ServeConfigSpec struct {
+	Name                      string             `json:"name"`
+	ImportPath                string             `json:"importPath"`
+	InitArgs                  []string           `json:"initArgs,omitempty"`
+	InitKwargs                map[string]string  `json:"initKwargs,omitempty"`
+	NumReplicas               *int32             `json:"numReplicas,omitempty"`
+	RoutePrefix               string             `json:"routePrefix,omitempty"`
+	MaxConcurrentQueries      *int32             `json:"maxConcurrentQueries,omitempty"`
+	UserConfig                map[string]string  `json:"userConfig,omitempty"`
+	AutoscalingConfig         map[string]string  `json:"autoscalingConfig,omitempty"`
+	GracefulShutdownWaitLoopS *int32             `json:"gracefulShutdownWaitLoopS,omitempty"`
+	GracefulShutdownTimeoutS  *int32             `json:"gracefulShutdownTimeoutS,omitempty"`
+	HealthCheckPeriodS        *int32             `json:"healthCheckPeriodS,omitempty"`
+	HealthCheckTimeoutS       *int32             `json:"healthCheckTimeoutS,omitempty"`
+	RayActorOptions           RayActorOptionSpec `json:"rayActorOptions,omitempty"`
+}
+
+// RayActorOptionSpec defines the desired state of RayActor
+type RayActorOptionSpec struct {
+	RuntimeEnv        map[string][]string `json:"runtimeEnv,omitempty"`
+	NumCpus           *float64            `json:"numCpus,omitempty"`
+	NumGpus           *float64            `json:"numGpus,omitempty"`
+	Memory            *int32              `json:"memory,omitempty"`
+	ObjectStoreMemory *int32              `json:"objectStoreMemory,omitempty"`
+	Resources         map[string]string   `json:"resources,omitempty"`
+	AcceleratorType   string              `json:"acceleratorType,omitempty"`
 }
 
 // RayServiceStatus defines the observed state of RayService
 type RayServiceStatus struct {
-	// INSERT ADDITIONAL STATUS FIELD - define observed state of cluster
 	// Important: Run "make" to regenerate code after modifying this file
+	ServiceStatus    ServiceStatus           `json:"serviceStatus,omitempty"`
+	ServeStatuses    []ServeDeploymentStatus `json:"serveDeploymentStatuses,omitempty"`
+	RayClusterName   string                  `json:"rayClusterName,omitempty"`
+	RayClusterStatus RayClusterStatus        `json:"rayClusterStatus,omitempty"`
+}
+
+// ServeDeploymentStatuses defines the current states of all Serve Deployments
+type ServeDeploymentStatuses struct {
+	Statuses []ServeDeploymentStatus `json:"statuses,omitempty"`
+}
+
+// ServeDeploymentStatus defines the current state of Serve Deployment
+type ServeDeploymentStatus struct {
+	// Name, Status, Message are from Ray Dashboard to represent the state of a serve deployment.
+	Name string `json:"name,omitempty"`
+	// TODO: change status type to enum
+	Status         string      `json:"status,omitempty"`
+	Message        string      `json:"message,omitempty"`
+	LastUpdateTime metav1.Time `json:"lastUpdateTime,omitempty"`
+	// Keep track of how long the service is healthy.
+	// Update when Serve Deployment is healthy or first time convert to unhealthy from healthy.
+	HealthLastUpdateTime metav1.Time `json:"healthLastUpdateTime,omitempty"`
 }
 
 //+kubebuilder:object:root=true