Refactor to match current design doc (#109)

* Refactor cloud switch conditions into interfaces within the `cloud` pkg * Update all bucket URLs to use hashes * Add support for `params.json` (fixes #114) --------- Co-authored-by: Sam Stoelinga <[email protected]>
substratusai · Jul 21, 2023 · a94bc63 · a94bc63
1 parent f7901d5
commit a94bc63
Show file tree

Hide file tree

Showing 69 changed files with 1,265 additions and 792 deletions.
diff --git a/.gitignore b/.gitignore
@@ -71,5 +71,8 @@ dist/
 
 # envsubst rendered templates
 gcpmanager-skaffold.yaml
+
+/controllermanager
+/gcpmanager
 gcpmanager-dependencies.yaml
 skaffold-dependencies.sh
diff --git a/Makefile b/Makefile
@@ -1,6 +1,6 @@
 
 # Image URL to use all building/pushing image targets
-IMG ?= docker.io/substratusai/controller-manager:v0.4.2-alpha
+IMG ?= docker.io/substratusai/controller-manager:v0.5.0-alpha
 # ENVTEST_K8S_VERSION refers to the version of kubebuilder assets to be downloaded by envtest binary.
 ENVTEST_K8S_VERSION = 1.26.1
 
@@ -114,8 +114,8 @@ dev: manifests kustomize install-crds
 	go run ./cmd/controllermanager/main.go
 
 .PHONY: run
-run: manifests generate fmt vet ## Run a controller from your host.
-	go run ./cmd/controllermanager/main.go
+run: ## Run a controller from your host.
+	go run ./cmd/controllermanager/main.go --config-dump-path=/tmp/substratus-config.yaml
 
 # If you wish built the manager image targeting other platforms you can use the --platform flag.
 # (i.e. docker build --platform linux/arm64 ). However, you must enable docker buildKit for it.

diff --git a/api/v1/common_types.go b/api/v1/common_types.go
@@ -76,3 +76,7 @@ type GPUResources struct {
 	// Count is the number of GPUs.
 	Count int64 `json:"count,omitempty"`
 }
+
+type ArtifactsStatus struct {
+	URL string `json:"url,omitempty"`
+}
diff --git a/api/v1/dataset_types.go b/api/v1/dataset_types.go
@@ -10,9 +10,6 @@ type DatasetSpec struct {
 	// Command to run in the container.
 	Command []string `json:"command,omitempty"`
 
-	// Filename is the name of the file when it is downloaded.
-	Filename string `json:"filename"`
-
 	// Image that contains dataset loading code and dependencies.
 	Image Image `json:"image"`
 
@@ -23,6 +20,10 @@ type DatasetSpec struct {
 	Params map[string]intstr.IntOrString `json:"params,omitempty"`
 }
 
+func (d *Dataset) GetParams() map[string]intstr.IntOrString {
+	return d.Spec.Params
+}
+
 func (d *Dataset) GetImage() *Image {
 	return &d.Spec.Image
 }
@@ -39,6 +40,10 @@ func (d *Dataset) SetStatusReady(r bool) {
 	d.Status.Ready = r
 }
 
+func (d *Dataset) GetStatusArtifacts() ArtifactsStatus {
+	return d.Status.Artifacts
+}
+
 func (d *Dataset) SetStatusImage(us ImageStatus) {
 	d.Status.Image = us
 }
@@ -56,11 +61,11 @@ type DatasetStatus struct {
 	// Conditions is the list of conditions that describe the current state of the Dataset.
 	Conditions []metav1.Condition `json:"conditions,omitempty"`
 
-	// URL of the loaded data.
-	URL string `json:"url,omitempty"`
+	// Artifacts status.
+	Artifacts ArtifactsStatus `json:"artifacts,omitempty"`
 
-	// Upload contains details the controller returns from a requested signed upload URL.
-	Image ImageStatus `json:"upload,omitempty"`
+	// Image contains the status of the image. Upload URL is reported here.
+	Image ImageStatus `json:"image,omitempty"`
 }
 
 //+kubebuilder:resource:categories=ai

diff --git a/api/v1/model_types.go b/api/v1/model_types.go
@@ -28,6 +28,10 @@ type ModelSpec struct {
 	Params map[string]intstr.IntOrString `json:"params,omitempty"`
 }
 
+func (m *Model) GetParams() map[string]intstr.IntOrString {
+	return m.Spec.Params
+}
+
 func (m *Model) GetImage() *Image {
 	return &m.Spec.Image
 }
@@ -44,6 +48,10 @@ func (m *Model) SetStatusReady(r bool) {
 	m.Status.Ready = r
 }
 
+func (m *Model) GetStatusArtifacts() ArtifactsStatus {
+	return m.Status.Artifacts
+}
+
 func (m *Model) SetStatusImage(us ImageStatus) {
 	m.Status.Image = us
 }
@@ -61,11 +69,11 @@ type ModelStatus struct {
 	// Conditions is the list of conditions that describe the current state of the Model.
 	Conditions []metav1.Condition `json:"conditions,omitempty"`
 
-	// URL of model artifacts.
-	URL string `json:"url,omitempty"`
+	// Artifacts status.
+	Artifacts ArtifactsStatus `json:"artifacts,omitempty"`
 
-	// Upload contains details the controller returns from a requested signed upload URL.
-	Image ImageStatus `json:"upload,omitempty"`
+	// Image contains the status of the image. Upload URL is reported here.
+	Image ImageStatus `json:"image,omitempty"`
 }
 
 //+kubebuilder:resource:categories=ai

diff --git a/api/v1/notebook_types.go b/api/v1/notebook_types.go
@@ -29,6 +29,10 @@ type NotebookSpec struct {
 	Params map[string]intstr.IntOrString `json:"params,omitempty"`
 }
 
+func (n *Notebook) GetParams() map[string]intstr.IntOrString {
+	return n.Spec.Params
+}
+
 func (n *Notebook) GetImage() *Image {
 	return &n.Spec.Image
 }
@@ -62,8 +66,8 @@ type NotebookStatus struct {
 	// Conditions is the list of conditions that describe the current state of the Notebook.
 	Conditions []metav1.Condition `json:"conditions,omitempty"`
 
-	// Upload contains details the controller returns from a requested signed upload URL.
-	Image ImageStatus `json:"upload,omitempty"`
+	// Image contains the status of the image. Upload URL is reported here.
+	Image ImageStatus `json:"image,omitempty"`
 }
 
 //+kubebuilder:resource:categories=ai

diff --git a/api/v1/server_types.go b/api/v1/server_types.go
@@ -28,8 +28,8 @@ type ServerStatus struct {
 	// Conditions is the list of conditions that describe the current state of the Server.
 	Conditions []metav1.Condition `json:"conditions,omitempty"`
 
-	// Upload contains details the controller returns from a requested signed upload URL.
-	Image ImageStatus `json:"upload,omitempty"`
+	// Image contains the status of the image. Upload URL is reported here.
+	Image ImageStatus `json:"image,omitempty"`
 }
 
 //+kubebuilder:resource:categories=ai

diff --git a/api/v1/zz_generated.deepcopy.go b/api/v1/zz_generated.deepcopy.go
diff --git a/cmd/controllermanager/main.go b/cmd/controllermanager/main.go
@@ -1,13 +1,17 @@
 package main
 
 import (
+	"bytes"
+	"context"
 	"flag"
+	"io/ioutil"
 	"os"
 
 	// Import all Kubernetes client auth plugins (e.g. Azure, GCP, OIDC, etc.)
 	// to ensure that exec-entrypoint and run can make use of them.
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/credentials/insecure"
+	"gopkg.in/yaml.v2"
 	_ "k8s.io/client-go/plugin/pkg/client/auth"
 
 	"k8s.io/apimachinery/pkg/runtime"
@@ -40,6 +44,8 @@ func main() {
 	var metricsAddr string
 	var enableLeaderElection bool
 	var probeAddr string
+	var configDumpPath string
+	flag.StringVar(&configDumpPath, "config-dump-path", "", "The filepath to dump the running config to.")
 	flag.StringVar(&metricsAddr, "metrics-bind-address", ":8080", "The address the metric endpoint binds to.")
 	flag.StringVar(&probeAddr, "health-probe-bind-address", ":8081", "The address the probe endpoint binds to.")
 	flag.BoolVar(&enableLeaderElection, "leader-elect", false,
@@ -79,12 +85,21 @@ func main() {
 
 	// NOTE: NewCloudContext() will look up environment variables (intended for local development)
 	// and if they are not specified, it will try to use metadata servers on the cloud.
-	cloudContext, err := cloud.NewContext()
+	cld, err := cloud.New(context.Background())
 	if err != nil {
-		setupLog.Error(err, "unable to determine cloud context")
+		setupLog.Error(err, "unable to determine cloud configuration")
 		os.Exit(1)
 	}
 
+	if configDumpPath != "" {
+		if err := dumpConfigToFile(configDumpPath, struct {
+			Cloud cloud.Cloud
+		}{Cloud: cld}); err != nil {
+			setupLog.Error(err, "unable to dump config to path")
+			os.Exit(1)
+		}
+	}
+
 	// TODO(any): setup TLS
 	conn, err := grpc.Dial(
 		// "localhost:10080",
@@ -100,15 +115,19 @@ func main() {
 	gc := sci.NewControllerClient(conn)
 
 	if err = (&controller.ModelReconciler{
-		Client:       mgr.GetClient(),
-		Scheme:       mgr.GetScheme(),
-		CloudContext: cloudContext,
+		Client: mgr.GetClient(),
+		Scheme: mgr.GetScheme(),
+		Cloud:  cld,
 		ContainerImageReconciler: &controller.ContainerImageReconciler{
-			Scheme:                 mgr.GetScheme(),
-			Client:                 mgr.GetClient(),
-			CloudContext:           cloudContext,
-			CloudManagerGrpcClient: gc,
-			Kind:                   "Model",
+			Scheme: mgr.GetScheme(),
+			Client: mgr.GetClient(),
+			Cloud:  cld,
+			SCI:    gc,
+			Kind:   "Model",
+		},
+		ParamsReconciler: &controller.ParamsReconciler{
+			Scheme: mgr.GetScheme(),
+			Client: mgr.GetClient(),
 		},
 	}).SetupWithManager(mgr); err != nil {
 		setupLog.Error(err, "unable to create controller", "controller", "Model")
@@ -118,11 +137,11 @@ func main() {
 		Client: mgr.GetClient(),
 		Scheme: mgr.GetScheme(),
 		ContainerImageReconciler: &controller.ContainerImageReconciler{
-			Scheme:                 mgr.GetScheme(),
-			Client:                 mgr.GetClient(),
-			CloudContext:           cloudContext,
-			CloudManagerGrpcClient: gc,
-			Kind:                   "Server",
+			Scheme: mgr.GetScheme(),
+			Client: mgr.GetClient(),
+			Cloud:  cld,
+			SCI:    gc,
+			Kind:   "Server",
 		},
 	}).SetupWithManager(mgr); err != nil {
 		setupLog.Error(err, "unable to create controller", "controller", "Server")
@@ -133,26 +152,34 @@ func main() {
 		Client: mgr.GetClient(),
 		Scheme: mgr.GetScheme(),
 		ContainerImageReconciler: &controller.ContainerImageReconciler{
-			Scheme:                 mgr.GetScheme(),
-			Client:                 mgr.GetClient(),
-			CloudContext:           cloudContext,
-			CloudManagerGrpcClient: gc,
-			Kind:                   "Notebook",
+			Scheme: mgr.GetScheme(),
+			Client: mgr.GetClient(),
+			Cloud:  cld,
+			SCI:    gc,
+			Kind:   "Notebook",
+		},
+		ParamsReconciler: &controller.ParamsReconciler{
+			Scheme: mgr.GetScheme(),
+			Client: mgr.GetClient(),
 		},
 	}).SetupWithManager(mgr); err != nil {
 		setupLog.Error(err, "unable to create controller", "controller", "Notebook")
 		os.Exit(1)
 	}
 	if err = (&controller.DatasetReconciler{
-		Client:       mgr.GetClient(),
-		Scheme:       mgr.GetScheme(),
-		CloudContext: cloudContext,
+		Client: mgr.GetClient(),
+		Scheme: mgr.GetScheme(),
+		Cloud:  cld,
 		ContainerImageReconciler: &controller.ContainerImageReconciler{
-			Scheme:                 mgr.GetScheme(),
-			Client:                 mgr.GetClient(),
-			CloudContext:           cloudContext,
-			CloudManagerGrpcClient: gc,
-			Kind:                   "Dataset",
+			Scheme: mgr.GetScheme(),
+			Client: mgr.GetClient(),
+			Cloud:  cld,
+			SCI:    gc,
+			Kind:   "Dataset",
+		},
+		ParamsReconciler: &controller.ParamsReconciler{
+			Scheme: mgr.GetScheme(),
+			Client: mgr.GetClient(),
 		},
 	}).SetupWithManager(mgr); err != nil {
 		setupLog.Error(err, "unable to create controller", "controller", "Dataset")
@@ -175,3 +202,14 @@ func main() {
 		os.Exit(1)
 	}
 }
+
+func dumpConfigToFile(path string, config interface{}) error {
+	var buf bytes.Buffer
+	if err := yaml.NewEncoder(&buf).Encode(config); err != nil {
+		return err
+	}
+	if err := ioutil.WriteFile(path, buf.Bytes(), 0644); err != nil {
+		return err
+	}
+	return nil
+}
diff --git a/cmd/gcpmanager/main_test.go b/cmd/gcpmanager/main_test.go
@@ -15,17 +15,20 @@ package main_test
 //		$URL
 
 // the following function was successfully used to exercise gcpmanager.Server.CreateSignedURL()
-// func invokeManually(s *gcpmanager.Server) {
+// func invokeManually(storageClient *storage.Client) {
 // 	payload := sci.CreateSignedURLRequest{
 // 		BucketName:        "substratus-ai-001-substratus-notebooks",
 // 		ObjectName:        "notebook.tar.gz",
 // 		ExpirationSeconds: 300,
 // 	}
+// 	serv := gcpmanager.Server{
+// 		StorageClient: storageClient,
+// 	}
 // 	fmt.Println("calling CreateSignedURL with payload:")
 
-// 	resp, err := s.CreateSignedURL(context.Background(), &payload)
-// 	if err != nil {
-// 		log.Fatalf("failed to create signed URL: %v", err)
-// 	}
-// 	fmt.Printf("signed URL: %v\n", resp.Url)
-// }
+//		resp, err := serv.CreateSignedURL(context.Background(), &payload)
+//		if err != nil {
+//			log.Fatalf("failed to create signed URL: %v", err)
+//		}
+//		fmt.Printf("signed URL: %v\n", resp.Url)
+//	}