Updated documentation to adapt to new gRPC services

docs/development/new_cp_support.md

@@ -4,22 +4,33 @@ Steps to be followed while implementing a new (hyperscale) provider are mentione
 
 ### Setting up your repository
 
-1. Fork [this repository](https://github.com/gardener/machine-controller-manager-provider-sampleprovider) on your github.
-1. Rename the repository from `machine-controller-manager-provider-sampleprovider` to `machine-controller-manager-provider-{provider-name}` by updating the same in your github repository settings.
-1. Create directories as required and nagivate into this path. {your-github-username} could also be {github-project}.
+1. Create a new empty repository named `machine-controller-manager-provider-{provider-name}` on github. Do not initialize this repository with a README.
+1. Copy the remote repository `URL` (HTTPS/SSH) to this repository which is displayed once you create this repository.
+1. Now on your local system, create directories as required. {your-github-username} given below could also be {github-project} depending on where you have created the new repository.
     ```bash
     mkdir -p $GOPATH/src/github.com/{your-github-username}
+    ```
+1. Navigate to this created directory.
+    ```bash
     cd $GOPATH/src/github.com/{your-github-username}
     ```
-1. Clone this newly forked repository.
+1. Clone [this repository](https://github.com/gardener/machine-controller-manager-provider-sampleprovider) on your local machine.
+    ```bash
+    git clone [email protected]:gardener/machine-controller-manager-provider-sampleprovider.git
+    ```
+1. Rename the directory from `machine-controller-manager-provider-sampleprovider` to `machine-controller-manager-provider-{provider-name}`.
     ```bash
-    git clone https://github.com/{your-github-username}/machine-controller-manager-provider-{provider-name}
+    mv machine-controller-manager-provider-sampleprovider machine-controller-manager-provider-{provider-name}
     ```
-1. Navigate into the newly created repository.
+1. Navigate into the newly created directory.
     ```bash
     cd machine-controller-manager-provider-{provider-name}
     ```
-1. Rename github project from `gardener` to `{github-org}` or `{your-github-username}` where ever you have forked the repository above. Use the hack script given below to do the same.
+1. Update the remote `origin` URL to the newly created repository's URL you had copied above.
+    ```bash
+    git remote set-url origin [email protected]:{your-github-username}/machine-controller-manager-provider-{provider-name}.git
+    ```
+1. Rename github project from `gardener` to `{github-org/your-github-username}` where ever you have cloned the repository above. Use the hack script given below to do the same.
     ```bash
     make rename-project PROJECT_NAME={github-org/your-github-username}
     eg:
@@ -36,35 +47,70 @@ Steps to be followed while implementing a new (hyperscale) provider are mentione
 1. Now commit your changes and push it upstream.
     ```bash
     git add -A
-    git commit
+    git commit -m "Renamed SampleProvider"
     git push origin master
     ```
 
 ### Code changes required
 
-1. Update the `pkg/{provider-name}/apis/provider-spec.go` specification file to reflect the structure of the objects exchanged between the machine-controller-manager and the driver. It typically contains the machine details and secrets (if required). Follow the sample spec provided already in the file. A sample provider specification can be found [here](https://github.com/prashanth26/machine-controller-manager-provider-gcp/blob/master/pkg/gcp/apis/provider-spec.go). 
-    ```bash
-    vim pkg/{provider-name}/apis/provider-spec.go
-    ```
-1. Fill in the create/list/delete methods `pkg/{provider-name}/machineserver.go` to create/list/delete VMs on your cloud provider. A sample provider implementation for these methods can be found [here](https://github.com/prashanth26/machine-controller-manager-provider-gcp/blob/master/pkg/gcp/machineserver.go).
-    ```bash
-    vim pkg/{provider-name}/machineserver.go
-    ```
+1. Update the `pkg/{provider-name}/apis/provider-spec.go` specification file to reflect the structure of the objects exchanged between the machine-controller-manager (AKA cmi-client) and the driver (AKA cmi-server). It typically contains the machine details and secrets (if required). Follow the sample spec provided already in the file. A sample provider specification can be found [here](https://github.com/prashanth26/machine-controller-manager-provider-gcp/blob/master/pkg/gcp/apis/provider-spec.go).
+1. Fill in the methods described at `pkg/{provider-name}/machineserver.go` to manage VMs on your cloud provider.
+    - Fill in the required methods CREATE/GET/DELETE and LIST methods.
+    - The request and response parameters for each of the methods to be implemented are well documented as comments and sample codes at `pkg/{provider-name}/machineserver.go`.
+    - Optional methods like SHUTDOWN are optional as implicit, however we strongly recommend you to implement them as well.
+    - A sample provider implementation for these methods can be found [here](https://github.com/prashanth26/machine-controller-manager-provider-gcp/blob/master/pkg/gcp/machineserver.go).
+1. Perform validation of APIs that you have described and make it a part of your methods as required.
+1. Write unit tests to make it work with your implementation by running `make test`.
 1. Re-generate the vendors, to update any new vendors imported.
     ```bash
     make revendor
     ```
+1. Update the sample YAML files on `kubernetes/` directory to provide sample files through which the working of the cmi-server can be tested.
+1. Update `README.md` to reflect any additional changes
 
 ### Testing your code changes
 
 Make sure `$KUBECONFIG` points to the cluster where you wish to manage machines. `$NAMESPACE` represents the namespaces where MCM is looking for machine objects.
 
-1. Run the provider driver
-```bash
-   go run app/controller/main.go --endpoint=tcp://127.0.0.1:8080
-```
-1. Run the machine-controller-manager in the `cmi-client` branch
+1. On the first terminal,
+    - Run the cmi-server (driver) using the command below.
+        ```bash
+        go run app/controller/cmi-server.go --endpoint=tcp://127.0.0.1:8080
+        ```
+1. On the second terminal,
+    - Clone the [latest MCM code](https://github.com/gardener/machine-controller-manager/tree/cmi-client) and switch to the `cmi-client` branch.
+    - Deploy the required CRDs from the machine-controller-manager repo,
+        ```bash
+        kubectl apply -f kubernetes/crds.yaml
+        ```
+    - Run the machine-controller-manager in the `cmi-client` branch
+        ```bash
+        go run cmd/machine-controller-manager/controller_manager.go --control-kubeconfig=$KUBECONFIG --target-kubeconfig=$KUBECONFIG --namespace=$NAMESPACE
+        ```
+1. On the third terminal,
+    - Fill in the object files given below and deploy them as described below.
+    - Deploy the `machine-class`
+        ```bash
+        kubectl apply -f kubernetes/machine-class.yaml
+        ```
+    - Deploy the `kubernetes secret` if required.
+        ```bash
+        kubectl apply -f kubernetes/secret.yaml
+        ```
+    - Deploy the `machine` object and make sure it joins the cluster successfully.
+        ```bash
+        kubectl apply -f kubernetes/machine.yaml
+        ```
+    - Once machine joins, you can test by deploying a machine-deployment.
+    - Deploy the `machine-deployment` object and make sure it joins the cluster successfully.
+        ```bash
+        kubectl apply -f kubernetes/machine-deployment.yaml
+
+### Releasing your docker image
+
+1. Make sure you have logged into gcloud/docker using the CLI.
+1. To release your docker image, run the following.
 ```bash
-   go run cmd/machine-controller-manager/controller_manager.go --control-kubeconfig=$KUBECONFIG --target-kubeconfig=$KUBECONFIG --namespace=$NAMESPACE
+    make release IMAGE_REPOSITORY=<link-to-image-repo>
 ```
-1. Create machineClass, secrets, machines and machinedeployment
+1. A sample kubernetes deploy file can be found at `kubernetes/deployment.yaml`. Modify the same to deploy your cmi-server.

pkg/controller/machine.go

@@ -23,9 +23,12 @@ import (
 	"fmt"
 	"time"
 
+	"google.golang.org/grpc/codes"
+
 	"github.com/golang/glog"
+	"google.golang.org/grpc/status"
 
-	"k8s.io/api/core/v1"
+	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/labels"
 	"k8s.io/apimachinery/pkg/selection"
 	"k8s.io/apimachinery/pkg/util/sets"
@@ -145,7 +148,7 @@ func (c *controller) reconcileClusterMachine(machine *v1alpha1.Machine) error {
 	//driver := driver.NewDriver(machine.Spec.ProviderID, secretRef, machine.Spec.Class.Kind, //MachineClass, machine.Name)
 
 	// TODO: Make cloud-provider configurable via machine.spec.provider or machineclass.provider
-	driver := driver.NewCmiDriverClient(machine.Spec.ProviderID, "AWS", secretRef, MachineClass, machine.Name)
+	driver := driver.NewCMIDriverClient(machine.Spec.ProviderID, "AWS", secretRef, MachineClass, machine.Name)
 
 	actualProviderID, err := machine.Spec.ProviderID, nil
 	if err != nil {
@@ -313,7 +316,7 @@ func (c *controller) updateMachineState(machine *v1alpha1.Machine) (*v1alpha1.Ma
 	Machine operations - Create, Update, Delete
 */
 
-func (c *controller) machineCreate(machine *v1alpha1.Machine, driver driver.CmiDriverClient) error {
+func (c *controller) machineCreate(machine *v1alpha1.Machine, driver driver.CMIDriverClient) error {
 	glog.V(2).Infof("Creating machine %s, please wait!", machine.Name)
 
 	actualProviderID, nodeName, err := driver.CreateMachine()
@@ -413,15 +416,17 @@ func (c *controller) machineUpdate(machine *v1alpha1.Machine, actualProviderID s
 	return nil
 }
 
-func (c *controller) machineDelete(machine *v1alpha1.Machine, driver driver.CmiDriverClient) error {
+func (c *controller) machineDelete(machine *v1alpha1.Machine, driver driver.CMIDriverClient) error {
 	var err error
 
+	// If Finalizers are present
 	if finalizers := sets.NewString(machine.Finalizers...); finalizers.Has(DeleteFinalizerName) {
 		glog.V(2).Infof("Deleting Machine %s", machine.Name)
 
-		// If machine was created on the cloud provider
-		machineID, _ := driver.GetExisting()
+		// Getting the machine-ID on the cloud provider
+		machineID := driver.MachineID
 
+		// If machine status has not been set to terminating
 		if machine.Status.CurrentStatus.Phase != v1alpha1.MachineTerminating {
 			lastOperation := v1alpha1.LastOperation{
 				Description:    "Deleting machine from cloud provider",
@@ -446,6 +451,7 @@ func (c *controller) machineDelete(machine *v1alpha1.Machine, driver driver.CmiD
 			}
 		}
 
+		// If machine is created on the cloud provider
 		if machineID != "" {
 			timeOutDuration := c.safetyOptions.MachineDrainTimeout.Duration
 			// Timeout value obtained by subtracting last operation with expected time out period
@@ -481,12 +487,27 @@ func (c *controller) machineDelete(machine *v1alpha1.Machine, driver driver.CmiD
 					c.updateMachineStatus(machine, lastOperation, machine.Status.CurrentStatus)
 
 					// Machine still tries to terminate after drain failure
-					glog.Warningf("Drain failed for machine %s - \nBuf:%v \nErrBuf:%v \nErr-Message:%v", machine.Name, buf, errBuf, err)
+					glog.Warningf("Drain failed for machine %q - \nBuf:%v \nErrBuf:%v \nErr-Message:%v", machine.Name, buf, errBuf, err)
 					return err
 				}
-				glog.V(2).Infof("Drain successful for machine %s - %v %v", machine.Name, buf, errBuf)
+				glog.V(2).Infof("Drain successful for machine %q", machine.Name)
+			}
+
+			// Check for existance of machine at cloud provider
+			_, err := driver.GetMachine(machineID)
+			if err == nil || status.Code(err) == codes.Unimplemented {
+				// Either there is no err
+				// or the GetMachine is not implemented
+				// then continue
+
+				err = driver.ShutDownMachine(machineID)
+				if err == nil || status.Code(err) == codes.Unimplemented {
+					// Either there is no err
+					// or the ShutDownMachine is not implemented
+					// then continue
+					err = driver.DeleteMachine(machineID)
+				}
 			}
-			err = driver.DeleteMachine()
 		}
 
 		if err != nil {

pkg/controller/machine_safety.go

@@ -430,17 +430,18 @@ func (c *controller) checkMachineClass(
 	}
 
 	// Dummy driver object being created to invoke GetVMs
-	dvr := driver.NewCmiDriverClient(
+	dvr := driver.NewCMIDriverClient(
 		"",
 		classKind,
 		secret,
 		machineClass,
 		"",
 	)
 
-	listOfVMs, err := dvr.GetVMs("")
+	listOfVMs, err := dvr.ListMachines()
 	if err != nil {
-		glog.Warningf("Failed to list VMs at provider. Err - %s", err)
+		glog.Warningf("Orphan VM handler is not running. Failed to list VMs at provider. Err - %s", err)
+		return
 	}
 
 	// Making sure that its not a VM just being created, machine object not yet updated at API server
@@ -469,16 +470,14 @@ func (c *controller) checkMachineClass(
 
 			// Re-check VM object existence
 			// before deleting orphan VM
-			result, _ := dvr.GetVMs(machineID)
-			for reMachineID := range result {
-				if reMachineID == machineID {
-					// Get latest version of machine object and verfiy again
-					machine, err := c.controlMachineClient.Machines(c.namespace).Get(machineName, metav1.GetOptions{})
-					if (err != nil && apierrors.IsNotFound(err)) || machine.Spec.ProviderID != machineID {
-						vm := make(map[string]string)
-						vm[machineID] = machineName
-						c.deleteOrphanVM(vm, secret, classKind, machineClass)
-					}
+			exists, _ := dvr.GetMachine(machineID)
+			if exists {
+				// Get latest version of machine object and verfiy again
+				machine, err := c.controlMachineClient.Machines(c.namespace).Get(machineName, metav1.GetOptions{})
+				if (err != nil && apierrors.IsNotFound(err)) || machine.Spec.ProviderID != machineID {
+					vm := make(map[string]string)
+					vm[machineID] = machineName
+					c.deleteOrphanVM(vm, secret, classKind, machineClass)
 				}
 			}
 
@@ -559,15 +558,15 @@ func (c *controller) deleteOrphanVM(vm driver.VMs, secretRef *corev1.Secret, kin
 		machineName = v
 	}
 
-	dvr := driver.NewCmiDriverClient(
+	dvr := driver.NewCMIDriverClient(
 		machineID,
 		kind,
 		secretRef,
 		machineClass,
 		machineName,
 	)
 
-	err := dvr.DeleteMachine()
+	err := dvr.DeleteMachine(machineID)
 	if err != nil {
 		glog.Errorf("Error while deleting VM on CP - %s. Shall retry in next safety controller sync.", err)
 	} else {