CLI example - Custom Container MLFlow (#1270)

* Add MLFlow custom container

* edit

* Aligned scoring and paths

* Add endpoint yaml

* Revert readme changes

* Add acr for auto test

* Add test

* Fix suffix

* Remove the model hack for modelids

* Add sed for endpoint

* Editing

* Reapply dockerfile

* Fix acr build

* Edits for tests

* Remove traffic adjustment

.github/workflows/cli-scripts-deploy-custom-container-mlflow.yml

@@ -0,0 +1,31 @@
+name: cli-scripts-deploy-custom-container-mlflow
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: "0 0/4 * * *"
+  pull_request:
+    branches:
+      - main
+    paths:
+      - cli/deploy-custom-container-mlflow.sh
+      - .github/workflows/cli-scripts-deploy-custom-container-mlflow.yml
+      - cli/setup.sh
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+    - name: check out repo
+      uses: actions/checkout@v2
+    - name: azure login
+      uses: azure/login@v1
+      with:
+        creds: ${{secrets.AZ_CREDS}}
+    - name: setup
+      run: bash setup.sh
+      working-directory: cli
+      continue-on-error: true
+    - name: scripts installs
+      run: sudo apt-get upgrade -y && sudo apt-get install uuid-runtime jq -y
+    - name: test script script
+      run: set -e; bash -x deploy-custom-container-mlflow.sh
+      working-directory: cli

cli/deploy-custom-container-mlflow.sh

@@ -0,0 +1,122 @@
+#!/bin/bash
+
+set -e
+
+# <set_variables>
+export ENDPOINT_NAME="<ENDPOINT_NAME>"
+export ACR_NAME="<CONTAINER_REGISTRY_NAME>"
+# </set_variables>
+
+export ENDPOINT_NAME=endpt-moe-`echo $RANDOM`
+export ACR_NAME=$(az ml workspace show --query container_registry -o tsv | cut -d'/' -f9-)
+
+# Create subdir "mlflow_context" and set BASE_PATH to it
+# <initialize_build_context>
+export PARENT_PATH=endpoints/online/custom-container
+export BASE_PATH="$PARENT_PATH/mlflow_context"
+export ASSET_PATH=endpoints/online/mlflow
+rm -rf $BASE_PATH && mkdir $BASE_PATH
+# </initialize_build_context> 
+
+# Copy model directories, sample-requests, and Dockerfile 
+# <copy_assets>
+cp -r $ASSET_PATH/{lightgbm-iris,sklearn-diabetes} $BASE_PATH
+cp $ASSET_PATH/sample-request-*.json $BASE_PATH 
+cp $PARENT_PATH/mlflow.dockerfile $BASE_PATH/Dockerfile
+cp $PARENT_PATH/mlflow-endpoint.yml $BASE_PATH/endpoint.yaml
+sed -i "s/{{endpoint_name}}/$ENDPOINT_NAME/g;" $BASE_PATH/endpoint.yaml 
+# </copy_assets> 
+
+# Create two deployment yamls, store paths in SKLEARN_DEPLOYMENT and LIGHTGBM_DEPLOYMENT
+# <make_deployment_yamls> 
+make_deployment_yaml () {
+    DEPLOYMENT_ENV=$1
+    MODEL_NAME=$2
+    export ${DEPLOYMENT_ENV}="$BASE_PATH/mlflow-deployment-$MODEL_NAME.yaml"   
+    cp $PARENT_PATH/mlflow-deployment.yml ${!DEPLOYMENT_ENV}
+    sed -i "s/{{acr_name}}/$ACR_NAME/g;\
+            s/{{endpoint_name}}/$ENDPOINT_NAME/g;\
+            s/{{environment_name}}/mlflow-cc-$MODEL_NAME-env/g;\
+            s/{{model_name}}/$MODEL_NAME/g;\
+            s/{{deployment_name}}/$MODEL_NAME/g;" ${!DEPLOYMENT_ENV}
+}
+
+make_deployment_yaml SKLEARN_DEPLOYMENT sklearn-diabetes
+make_deployment_yaml LIGHTGBM_DEPLOYMENT lightgbm-iris 
+#</make_deployment_yaml>
+
+# <login_to_acr>
+az acr login -n ${ACR_NAME} 
+# </login_to_acr> 
+
+# <build_with_acr>
+az acr build --build-arg MLFLOW_MODEL_NAME=sklearn-diabetes -t azureml-examples/mlflow-cc-sklearn-diabetes:latest -r $ACR_NAME $BASE_PATH
+az acr build --build-arg MLFLOW_MODEL_NAME=lightgbm-iris -t azureml-examples/mlflow-cc-lightgbm-iris:latest -r $ACR_NAME $BASE_PATH 
+# </build_with_acr>
+
+# <create_endpoint>
+az ml online-endpoint create -f $BASE_PATH/endpoint.yaml 
+# </create_endpoint>
+
+endpoint_status=`az ml online-endpoint show --name $ENDPOINT_NAME --query "provisioning_state" -o tsv`
+
+echo $endpoint_status
+
+if [[ $endpoint_status == "Succeeded" ]]
+then
+  echo "Endpoint created successfully"
+else 
+  echo "Endpoint creation failed"
+  exit 1
+fi
+
+# <create_deployments>
+az ml online-deployment create -f $SKLEARN_DEPLOYMENT 
+az ml online-deployment create -f $LIGHTGBM_DEPLOYMENT  
+# </create_deployments>
+
+# <check_deploy_status>
+az ml online-deployment show --endpoint-name $ENDPOINT_NAME --name sklearn-diabetes
+az ml online-deployment show --endpoint-name $ENDPOINT_NAME --name lightgbm-iris
+# </check_deploy_status>
+
+check_deployment_status () {
+    deploy_name=$1
+    deploy_status=`az ml online-deployment show --endpoint-name $ENDPOINT_NAME --name $deploy_name --query "provisioning_state" -o tsv`
+    echo $deploy_status
+    if [[ $deploy_status == "Succeeded" ]]
+    then
+    echo "Deployment $deploy_name completed successfully"
+    else
+    echo "Deployment $deploy_name failed"
+    exit 1
+    fi
+}
+
+check_deployment_status sklearn-diabetes
+check_deployment_status lightgbm-iris
+
+# <test_online_endpoints_with_invoke>
+az ml online-endpoint invoke -n $ENDPOINT_NAME --deployment-name sklearn-diabetes --request-file "$BASE_PATH/sample-request-sklearn.json"
+az ml online-endpoint invoke -n $ENDPOINT_NAME --deployment-name lightgbm-iris --request-file "$BASE_PATH/sample-request-lightgbm.json"
+# </test_online_endpoints_with_invoke>
+
+# Get key
+echo "Getting access key..."
+KEY=$(az ml online-endpoint get-credentials -n $ENDPOINT_NAME --query primaryKey -o tsv)
+
+# Get scoring url
+echo "Getting scoring url..."
+SCORING_URL=$(az ml online-endpoint show -n $ENDPOINT_NAME --query scoring_uri -o tsv)
+echo "Scoring url is $SCORING_URL"
+
+# <test_online_endpoints_with_curl>
+curl -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" -H "azureml-model-deployment: sklearn-diabetes" -d @"$BASE_PATH/sample-request-sklearn.json"  $SCORING_URL
+curl -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" -H "azureml-model-deployment: lightgbm-iris" -d @"$BASE_PATH/sample-request-lightgbm.json"  $SCORING_URL
+# </test_online_endpoints_with_curl>
+
+# <delete_online_endpoint>
+az ml online-endpoint delete -y -n $ENDPOINT_NAME
+# </delete_online_endpoint>
+
+#rm -rf $BASE_PATH

cli/endpoints/online/custom-container/mlflow-deployment.yml

@@ -0,0 +1,21 @@
+$schema: https://azuremlschemas.azureedge.net/latest/managedOnlineDeployment.schema.json
+name: {{deployment_name}} 
+endpoint_name: {{endpoint_name}}
+model:
+  name: {{model_name}}
+  path: ./{{model_name}}/model
+environment: 
+  name: {{environment_name}} 
+  image: {{acr_name}}.azurecr.io/azureml-examples/mlflow-cc-{{model_name}}:latest
+  inference_config:
+    liveness_route:
+      path: /
+      port: 5001
+    readiness_route:
+      path: /
+      port: 5001
+    scoring_route:
+      path: /score
+      port: 5001
+instance_type: Standard_DS2_v2
+instance_count: 1

cli/endpoints/online/custom-container/mlflow-endpoint.yml

@@ -0,0 +1,3 @@
+$schema: https://azuremlschemas.azureedge.net/latest/managedOnlineEndpoint.schema.json
+name: {{endpoint_name}}
+auth_mode: key

cli/endpoints/online/custom-container/mlflow.dockerfile

@@ -0,0 +1,24 @@
+FROM mcr.microsoft.com/azureml/mlflow-ubuntu18.04-py37-cpu-inference:latest
+
+USER dockeruser
+
+ARG MLFLOW_MODEL_NAME=model_name
+
+# Conda is already installed
+ENV CONDA_ENV_DIR=/opt/miniconda/envs
+
+# Create a new conda environment and install the same version of the server
+COPY $MLFLOW_MODEL_NAME/model/conda.yaml /tmp/conda.yml
+RUN conda env create -n userenv -f /tmp/conda.yml && \
+    export SERVER_VERSION=$(pip show azureml-inference-server-http | grep Version | sed -e 's/.*: //')  && \ 
+    $CONDA_ENV_DIR/userenv/bin/pip install azureml-inference-server-http==$SERVER_VERSION
+
+# Update environment variables to default to the new conda env
+ENV AZUREML_CONDA_ENVIRONMENT_PATH="$CONDA_ENV_DIR/userenv" 
+ENV PATH="$AZUREML_CONDA_ENVIRONMENT_PATH/bin:$PATH"
+ENV LD_LIBRARY_PATH="$AZUREML_CONDA_ENVIRONMENT_PATH/lib:$LD_LIBRARY_PATH"
+
+ENV MLFLOW_MODEL_FOLDER=model
+
+EXPOSE 5001
+CMD ["runsvdir", "/var/runit"]