Deprecate dev image & use image from arg instead.

Fix #9

cmd/dev-start.go

@@ -29,11 +29,12 @@ import (
 
 // startCmd represents the `dev start` command
 var startCmd = &cobra.Command{
-	Use:   "start",
+	Use:   "start [docker-image]",
 	Short: "Start a dev session.",
-	Long:  "Start a dev session.",
+	Long:  "Start a dev session using the provider docker image.",
+	Args:  cobra.ExactArgs(1),
 	Run: func(cmd *cobra.Command, args []string) {
-		if err := start(); err != nil {
+		if err := start(args[0]); err != nil {
 			fmt.Println("ERROR:", err)
 		}
 	},
@@ -43,12 +44,12 @@ func init() {
 	devCmd.AddCommand(startCmd)
 }
 
-func start() error {
+func start(devImage string) error {
 	fmt.Println("→ Starting a remote session...")
 	// Image to run.
 	image := viper.GetString("image")
 	// Command to run.
-	command := []string{"kuda_dev_start"}
+	command := []string{"kuda_dev_start", devImage}
 
 	// Add the CWD to the volumes mounted in Docker.
 	dir, err := os.Getwd()

docs/kuda/cli.md

@@ -32,17 +32,24 @@ Deletes the remote cluster.
 ### → Start
 
 ```bash
-kuda dev start
+kuda dev start [base-image]
 ```
 
 Starts a remote dev session with the current working directory.
 
+**Example:** `kuda dev start nvidia/cuda:10.1-base`
+
 This command:
 
 - Provisions a node with GPU on the cluster & install the nvidia driver
 - Starts a development pod based on the Deep Learning VM
 - Synchronise the directory provided as parameter with the remote node
 
+List of recommended `base-image`:
+
+- all images from [nvidia/cuda](https://hub.docker.com/r/nvidia/cuda/)
+- gcloud's [Deep Learning containers](https://cloud.google.com/ai-platform/deep-learning-containers/docs/choosing-container)
+
 ### → Stop
 
 ```bash

docs/kuda/remote_development.md

@@ -19,9 +19,11 @@ Developping on remote sessions offers many other advantages such as:
 Start a remote dev session that will be provisioned on your cluster.
 
 ```bash
-kuda dev start
+kuda dev start nvidia/cuda:10.1-base
 ```
 
+`nvidia/cuda:10.1-base` Is the docker image to use as base. It allows you to specify which version of CUDA and CuDNN you need. You can find a list of suggested images in the kuda dev [reference page](https://docs.kuda.dev/kuda/cli#dev).
+
 This command will start the remote session and synchronize the CWD \(current working directory\) with the remote instance.
 
 Upon started, it will also print the cluster's IP address / port to use later on. Make not of that as we'll refer to it later as `<your-dev-session-external-ip:port>`

images/providers/gcp/.config.sh

@@ -26,8 +26,6 @@ export KUDA_DEFAULT_GPU="${KUDA_DEFAULT_GPU:-k80}"
 export KUDA_DEFAULT_USE_PREEMPTIBLE="${KUDA_DEFAULT_USE_PREEMPTIBLE:-false}"
 
 export KUDA_DEV_APP_NAME="${KUDA_DEV_APP_NAME:-kuda-dev}"
-export KUDA_DEV_APP_IMAGE_NAME="${KUDA_DEV_APP_IMAGE_NAME:-kuda-dev-base}"
-export KUDA_DEV_APP_IMAGE="${KUDA_DEV_APP_IMAGE:-gcr.io/kuda-project/dev:1.0.0}"
 export KUDA_DEV_SYNC_PATH="${KUDA_DEV_SYNC_PATH:-/app_home}"
 
 # Disable prompts.

images/providers/gcp/dev_start.sh

@@ -13,11 +13,9 @@ gcloud container clusters get-credentials $KUDA_GCP_CLUSTER_NAME
 #   --num-nodes 1 \
 #   --quiet
 
-# TODO: Support multiple sessions.
-# tmp_uuid=$(od -x /dev/urandom | head -1 | awk '{print $2$3}')
-# KUDA_DEV_APP_NAME=kuda-dev-$tmp_uuid
-
-# exit 1
+uid=$(head /dev/urandom | tr -dc A-Za-z0-9 | head -c 7)
+dev_app_name=$KUDA_DEV_APP_NAME-$uid
+dev_image=$1
 
 # Launch.
 # TODO: materialize as external yaml file and customize
@@ -26,39 +24,41 @@ cat <<EOF | kubectl apply -f -
 apiVersion: v1
 kind: Service
 metadata:
-  name: $KUDA_DEV_APP_NAME
+  name: $dev_app_name
   labels:
-    app: $KUDA_DEV_APP_NAME
+    app: $dev_app_name
 spec:
   ports:
   - name: http
     port: 8000
     targetPort: 80
   selector:
-    app: $KUDA_DEV_APP_NAME
+    app: $dev_app_name
 ---
 apiVersion: apps/v1
 kind: Deployment
 metadata:
-  name: $KUDA_DEV_APP_NAME
+  name: $dev_app_name
 spec:
   replicas: 1
   selector:
     matchLabels:
-      app: $KUDA_DEV_APP_NAME
+      app: $dev_app_name
   template:
     metadata:
       labels:
-        app: $KUDA_DEV_APP_NAME
+        app: $dev_app_name
     spec:
       containers:
-        - name: $KUDA_DEV_APP_IMAGE_NAME
-          image: $KUDA_DEV_APP_IMAGE
+        - name: kuda-dev-image
+          image: $dev_image
+          imagePullPolicy: Always
           resources:
             limits:
               nvidia.com/gpu: 1
           ports:
           - containerPort: 80
+          command: ["/bin/bash"]
           env:
             - name: POD_NAME
               valueFrom:
@@ -85,7 +85,7 @@ cat <<EOF | kubectl apply -f -
 apiVersion: networking.istio.io/v1alpha3
 kind: Gateway
 metadata:
-  name: $KUDA_DEV_APP_NAME-gateway
+  name: $dev_app_name-gateway
 spec:
   selector:
     istio: ingressgateway # use Istio default gateway implementation
@@ -95,17 +95,17 @@ spec:
       name: http
       protocol: HTTP
     hosts:
-    - "$KUDA_DEV_APP_NAME.example.com"
+    - "$dev_app_name.example.com"
 ---
 apiVersion: networking.istio.io/v1alpha3
 kind: VirtualService
 metadata:
-  name: $KUDA_DEV_APP_NAME
+  name: $dev_app_name
 spec:
   hosts:
-  - "$KUDA_DEV_APP_NAME.example.com"
+  - "$dev_app_name.example.com"
   gateways:
-  - $KUDA_DEV_APP_NAME-gateway
+  - $dev_app_name-gateway
   http:
   - match:
     - uri:
@@ -114,7 +114,7 @@ spec:
     - destination:
         port:
           number: 8000
-        host: $KUDA_DEV_APP_NAME
+        host: $dev_app_name
 EOF
 
 # Initialize ksync.
@@ -149,7 +149,7 @@ sp="⣷⣯⣟⡿⢿⣻⣽⣾"
 while true; do
   printf "\b${sp:i++%${#sp}:1}"
   min=1
-  status=$(kubectl get deployment $KUDA_DEV_APP_NAME -o jsonpath={.status.availableReplicas})
+  status=$(kubectl get deployment $dev_app_name -o jsonpath={.status.availableReplicas})
   if [ ! -z "$status" ]; then
     if [ "$status" -ge 1 ]; then
       break
@@ -164,7 +164,7 @@ ksync watch --daemon=true
 
 # Watch file changes.
 ksync create \
-  --selector=app=$KUDA_DEV_APP_NAME \
+  --selector=app=$dev_app_name \
   --reload=false \
   $KUDA_DEV_SYNC_PATH \
   $KUDA_DEV_SYNC_PATH
@@ -180,5 +180,5 @@ echo "$ingress_host:$ingress_port"
 echo
 
 # Launch remote shell.
-pod_name=$(kubectl get pods -o name | grep -m1 $KUDA_DEV_APP_NAME | cut -d'/' -f 2)
+pod_name=$(kubectl get pods -o name | grep -m1 $dev_app_name | cut -d'/' -f 2)
 kubectl exec -it $pod_name -- /bin/bash