Fix up mlflow-server (#29)

- Fix issues that causes failure when relating the pod defaults interface
  1) KeyError when getting object storage interface data
  2) Typo for getting mlflow port config value
  3) Fail to load `mlflow-requirements.txt` because it wasn't packed in the charm
- Add namespace to MLFLOW_S3_ENDPOINT_URL, or else it would fail to log model following the example script
- Removed unused scripts
- Updated readme
  - follow official quick start guide to deploy kubeflow
  - move example model and script into an .ipynb
  - add step to copy pod defaults to user's namespace (temporary workaround)
  - create minio bucket in example model script if default bucket is not created 
  - add quick instructions on how to read data from minio
- Added tests

.github/workflows/integration.yaml

@@ -7,123 +7,65 @@ on:
   pull_request:
 
 jobs:
-  unit-test:
-    name: Unit Test
+  lint:
+    name: Lint
     runs-on: ubuntu-latest
     strategy:
+      fail-fast: false
       matrix:
-        charm: [server]
-
+        charm:
+          - server
     steps:
-    - uses: actions/checkout@v2
-    - run: sudo apt update
-    - run: sudo apt install tox
-    - run: tox -e ${{ matrix.charm }}
+      - uses: actions/checkout@v2
+      - run: python3 -m pip install tox
+      - run: tox -e ${{ matrix.charm }}-lint
 
-  deploy:
-    name: Test
+  unit:
+    name: Unit tests
     runs-on: ubuntu-latest
-
+    strategy:
+      fail-fast: false
+      matrix:
+        charm:
+          - server
     steps:
-    - name: Check out repo
-      uses: actions/checkout@v2
-
-    - uses: balchua/microk8s-actions@v0.2.2
-      with:
-        channel: '1.20/stable'
-        addons: '["dns", "storage", "rbac", "metallb:10.64.140.43-10.64.140.49"]'
-
-    - name: Install dependencies
-      run: |
-        set -eux
-        sudo snap install charm --classic
-        sudo snap install juju --classic
-        sudo snap install juju-bundle --classic
-        sudo snap install juju-wait --classic
-        sudo snap install charmcraft --classic --channel latest/candidate
-        sudo apt update
-        sudo apt install -y firefox-geckodriver tox
-
-    # Avoid race condition with storage taking a long time to initialize
-    - name: Wait for storage
-      run: |
-        sg microk8s -c 'microk8s kubectl rollout status deployment/hostpath-provisioner -n kube-system'
-
-    - name: Bootstrap
-      run: |
-        set -eux
-        sg microk8s -c 'juju bootstrap microk8s uk8s'
-        juju add-model mlflow
-
-    - name: Deploy charm dependencies
-      timeout-minutes: 15
-      run: |
-        set -eux
-        juju model-config update-status-hook-interval=15s
-        juju deploy istio-gateway --channel=1.5/stable --trust istio-ingressgateway
-        juju deploy istio-pilot --channel=1.5/stable --config default-gateway=kubeflow-gateway
-        juju relate istio-pilot:istio-pilot istio-ingressgateway:istio-pilot
-        sleep 30
-        kubectl patch role -n mlflow istio-ingressgateway-operator -p '{"apiVersion":"rbac.authorization.k8s.io/v1","kind":"Role","metadata":{"name":"istio-ingressgateway-operator"},"rules":[{"apiGroups":["*"],"resources":["*"],"verbs":["*"]}]}'
-        juju wait -wvt 300 --retry_errors 20
-        juju model-config update-status-hook-interval=5m
-
-        # https://bugs.launchpad.net/juju/+bug/1921739
-        (i=600; while ! juju wait -wvt 30 ; do ((--i)) || exit; sleep 1; done)
-
-    - name: Deploy MLflow
-      run: |
-        set -eux
-        sg microk8s -c 'juju bundle deploy --build --destructive-mode'
-        juju relate mlflow-server istio-pilot
-        juju wait -wvt 900 --retry_errors 20
-
-    - name: Test MLflow
-      run: sg microk8s -c 'tox -e selenium'
+      - uses: actions/checkout@v2
+      - run: python3 -m pip install tox
+      - run: tox -e ${{ matrix.charm }}-unit
 
-    - run: kubectl get all -A
-      if: failure()
-
-    - run: kubectl get virtualservices -A
-      if: failure()
-
-    - run: juju status
-      if: failure()
-
-    - name: Get mlflow workload logs
-      run: kubectl logs --tail 100 -nmlflow -lapp.kubernetes.io/name=mlflow-server
-      if: failure()
-
-    - name: Get mlflow operator logs
-      run: kubectl logs --tail 100 -nmlflow -loperator.juju.is/name=mlflow-server
-      if: failure()
-
-    - name: Generate inspect tarball
-      run: >
-        sg microk8s <<EOF
-          microk8s inspect | \
-          grep -Po "Report tarball is at \K.+" | \
-          xargs -I {} cp {} inspection-report-${{ strategy.job-index }}.tar.gz
-        EOF
-      if: failure()
-
-    - name: Upload inspect tarball
-      uses: actions/upload-artifact@v2
-      with:
-        name: inspection-reports
-        path: ./inspection-report-${{ strategy.job-index }}.tar.gz
-      if: failure()
-
-    - name: Upload selenium screenshots
-      uses: actions/upload-artifact@v2
-      with:
-        name: selenium-screenshots
-        path: /tmp/selenium-*.png
-      if: failure()
-
-    - name: Upload HAR logs
-      uses: actions/upload-artifact@v2
-      with:
-        name: selenium-har
-        path: /tmp/selenium-*.har
-      if: failure()
+  integration:
+    name: Integration tests (microk8s)
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        charm:
+          - server
+    steps:
+      - uses: actions/checkout@v2
+      - name: Setup operator environment
+        uses: charmed-kubernetes/actions-operator@main
+        with:
+          provider: microk8s
+          channel: 1.21/stable
+          charmcraft-channel: latest/candidate
+      - run: sg microk8s -c 'microk8s enable metallb:10.64.140.43-10.64.140.49'
+      - run: sudo apt install -y firefox-geckodriver
+
+      - run: |
+          sg microk8s -c "tox -e ${{ matrix.charm }}-integration"
+
+      # Collect debug logs if failed
+      - name: Dump Juju/k8s logs on failure
+        uses: canonical/charm-logdump-action@main
+        if: failure()
+        with:
+          app: ${{ matrix.charm }}
+          model: testing
+
+      - name: Upload HAR logs
+        uses: actions/upload-artifact@v2
+        with:
+          name: selenium-har
+          path: /tmp/selenium-*.har
+        if: failure()