AWS: cleanup & fix dev

cmd/setup-aws.go

@@ -47,7 +47,7 @@ func init() {
 
 func setupAWS() error {
 	const provider = "aws"
-	const providerVersion = "0.1.0"
+	const providerVersion = "2.0.0"
 
 	// Set provider config.
 	viper.Set("provider", provider)

providers/aws/.config.sh

@@ -1,5 +1,3 @@
-set -e
-
 echo
 echo -e "\e[1m \e[34mKuda AWS provider \e[36mv$(cat /kuda_cmd/VERSION) \e[0m"
 echo

providers/aws/Dockerfile

@@ -1,12 +1,5 @@
 FROM alpine:3.10
 
-# Set dependencies versions.
-ARG EKSCTL_VERSION=0.7.0
-ARG KUBECTL_VERSION=1.16.2
-ARG SKAFFOLD_VERSION=0.40.0
-ARG ISTIO_VERSION=1.3.3
-ARG HELM_VERSION=2.15.0
-
 # Install base apps & docker.
 RUN apk add --update-cache --no-cache \
   bash \
@@ -23,27 +16,34 @@ RUN apk add --update-cache --no-cache \
 RUN pip install awscli
 
 # Install eksctl
+ARG EKSCTL_VERSION=0.7.0
 ARG EKSCTL_HOST="https://github.com/weaveworks/eksctl/releases/download"
 RUN curl --silent \
   --location "${EKSCTL_HOST}/${EKSCTL_VERSION}/eksctl_$(uname -s)_amd64.tar.gz" \
   | tar xz -C /tmp
 RUN mv /tmp/eksctl /usr/local/bin
 
 # Install kubectl
+ARG KUBECTL_VERSION=1.16.2
 ARG KUBECTL_HOST="https://storage.googleapis.com/kubernetes-release/release"
 RUN curl -LO "${KUBECTL_HOST}/v${KUBECTL_VERSION}/bin/linux/amd64/kubectl" && \
   chmod +x ./kubectl && \
   mv ./kubectl /usr/local/bin/kubectl
 
 # Install Helm
+ARG HELM_VERSION=2.15.1
 RUN curl -L https://git.io/get_helm.sh | DESIRED_VERSION="v${HELM_VERSION}" bash
 
 # Download Istio CRDs
+ARG ISTIO_VERSION=1.3.3
 RUN curl -L https://git.io/getLatestIstio | ISTIO_VERSION="${ISTIO_VERSION}" sh -
 
 # Install Skaffold.
-ARG SKAFFOLD_HOST="https://storage.googleapis.com/skaffold/releases"
+ARG SKAFFOLD_VERSION=0.40.0
+# ARG SKAFFOLD_HOST="https://storage.googleapis.com/skaffold/releases"
 # RUN curl -Lo skaffold "${SKAFFOLD_HOST}/v${SKAFFOLD_VERSION}/skaffold-linux-amd64" && \
+# Temporarily use a custom skaffold build that fixes a kaniko issue #731 : 
+# https://github.com/GoogleContainerTools/skaffold/issues/731
 RUN curl -Lo skaffold "https://storage.googleapis.com/artifacts.gpu-sh.appspot.com/releases/latest/skaffold-linux-amd64" && \
   chmod +x skaffold && \
   mv skaffold /usr/local/bin

providers/aws/README.md

@@ -5,30 +5,43 @@ It uses [ECR](https://aws.amazon.com/ecr) to privately store the app images.
 
 # Status
 
-| Command | Status |
-| - | - |
-| setup | ✔ |
-| delete | ✔ |
-| get | Not Started |
-| app dev | WIP |
-| app deploy | WIP |
-| app delete | ✔ |
+| Command      |  Status |
+| ------------ | ------- |
+| `setup`      | ✔       |
+| `delete`     | ✔       |
+| `app dev`    | ✔       |
+| `app deploy` | ✔       |
+| `app delete` | ✔       |
+
+Functionalities:
+
+| Functionality       |  Status     |
+| ------------------- | ----------- |
+| Dev file sync       | ✔           |
+| GPU node autoscaler | WIP         |
+| Https               | Not started |
+| Dns                 | Not started |
+| Monitoring          | Not started |
+
+Notes:
+
+- There is an issue with the nodes autoscaler
 
 # Prerequisites
 
 - You must have subscribed to [EKS-optimize AMI with GPU support](https://aws.amazon.com/marketplace/pp/B07GRHFXGM)
 - You must have an increased limit of at least 1 instance of type p2.xlarge. You can make requests [here](http://aws.amazon.com/contact-us/ec2-request)
 - You must have an [aws configuration](https://docs.aws.amazon.com/cli/latest/userguide/cli-configure-files.html) on your local machine in `~/.aws/` with credentials that have authorization for:
-    - cloudformation
-    - ec2
-    - ec2 autoscaling
-    - eks
-    - iam
-    - api
-    - ecr
-    - elb
+  - cloudformation
+  - ec2
+  - ec2 autoscaling
+  - eks
+  - iam
+  - api
+  - ecr
+  - elb
 
-# Configuration
+<!-- # Configuration
 
 You can override the following settings by adding them as flags of the `kuda setup` command (ex: `kuda setup aws ... --aws_...=mycluster`).
 
@@ -37,5 +50,4 @@ You can override the following settings by adding them as flags of the `kuda set
 | | | |
 
 
-# Limitations
-
+# Limitations -->

providers/aws/app_dev.sh

@@ -4,68 +4,41 @@ set -e
 
 source $KUDA_CMD_DIR/.config.sh
 
+# Config.
+cluster_name=$KUDA_AWS_CLUSTER_NAME
+aws_region=$KUDA_AWS_CLUSTER_REGION
 aws_account_id="$(aws sts get-caller-identity | jq -r .Account)"
-ecr_domain="$aws_account_id.dkr.ecr.$KUDA_AWS_CLUSTER_REGION.amazonaws.com"
-
+ecr_domain="$aws_account_id.dkr.ecr.$aws_region.amazonaws.com"
 app_name=$1
 app_registry="$ecr_domain/$app_name"
-app_image="$app_registry:$app_version"
 namespace="default"
-
 app_cache_name=$app_name-cache
 
-echo $app_image
-
-# Create Container Registry if it doesn't exists.
-if [ -z "$(aws ecr describe-repositories --region $KUDA_AWS_CLUSTER_REGION | grep $app_name)" ]; then
+# Create container registries if they doesn't exists.
+repos="$(aws ecr describe-repositories --region $aws_region)"
+if [ -z "$(echo $repos | grep $app_name)" ]; then
   aws ecr create-repository \
     --repository-name $app_name \
-    --region $KUDA_AWS_CLUSTER_REGION
+    --region $aws_region
 else
   echo "Container Registry $app_registry already exists"
 fi
-
-# Create the cache registry if it doesn't exists.
-if [ -z "$(aws ecr describe-repositories --region $KUDA_AWS_CLUSTER_REGION | grep $app_cache_name)" ]; then
+# Cache registry.
+if [ -z "$(echo $repos | grep $app_cache_name)" ]; then
   aws ecr create-repository \
     --repository-name $app_cache_name \
-    --region $KUDA_AWS_CLUSTER_REGION
+    --region $aws_region
 else
   echo "Container Registry $app_registry-cache already exists"
 fi
 
 # Retrieve cluster token.
 aws eks update-kubeconfig \
-  --name $KUDA_AWS_CLUSTER_NAME \
-  --region $KUDA_AWS_CLUSTER_REGION
+  --name $cluster_name \
+  --region $aws_region
 
 # Login Container Registry.
-# aws ecr get-login --region $KUDA_AWS_CLUSTER_REGION --no-include-email | bash
-
-#TODO: Build & Push image using Kaniko.
-
-# Write Knative service config.
-# cat <<EOF | kubectl apply -f -
-# apiVersion: serving.knative.dev/v1alpha1
-# kind: Service
-# metadata:
-#   name: $app_name
-#   namespace: default
-# spec:
-#   template:
-#     spec:
-#       nodeSelector:
-#         nvidia.com/gpu: "true"
-#       tolerations:
-#         - key: "nvidia.com/gpu"
-#           operator: "Exists"
-#           effect: "NoSchedule"
-#       containers:
-#         - image: $app_image
-#           resources:
-#             limits:
-#               nvidia.com/gpu: 1
-# EOF
+aws ecr get-login --region $aws_region --no-include-email | bash
 
 # Write Knative service config.
 echo "
@@ -78,20 +51,22 @@ spec:
   template:
     spec:
       # nodeSelector:
-      #   nvidia.com/gpu: true
+      #   nvidia.com/gpu: 'true'
       # tolerations:
-      #   - key: nvidia.com/gpu
-      #     operator: Exists
-      #     effect: NoSchedule
+      #   - key: 'nvidia.com/gpu'
+      #     operator: 'Exists'
+      #     effect: 'NoSchedule'
       containers:
         - image: $app_registry
-          resources:
-            limits:
-              nvidia.com/gpu: 1
+          # resources:
+          #   limits:
+          #     nvidia.com/gpu: 1
+          env:
+            - name: KUDA_DEV
+              value: 'true'
 " >.kuda-app.k8.yaml
 
-# cat <<EOF | skaffold dev -n $namespace -f -
-cat <<EOF | skaffold run -v debug -n $namespace -f -
+cat <<EOF | skaffold dev -n $namespace -f -
 apiVersion: skaffold/v1beta17
 kind: Config
 build:
@@ -102,8 +77,6 @@ build:
           - src: '**/*'
             dest: .
       kaniko:
-        buildArgs:
-          verbosity: debug
         buildContext:
           localDir: {}
         cache: