PLAT-2802: Set metrics server up to be installed in agent cfg (#76)

* Set metrics server up to be installed in agent cfg * Allow deployer image override in PRs * Make it less likely we need to scale up * AttachVolume needs to match instance tags Chose a tag that's on both volumes and instances instead, and it's also independent of any sort of "deployment tracking" so hopefully leaves us free to change that schema if desired. * Sigh at black
dominodatalab · Oct 11, 2021 · 59c49bc · 59c49bc
1 parent 95a00cf
commit 59c49bc
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 2 deletions.
diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
@@ -28,6 +28,14 @@ jobs:
       uses: actions/setup-node@v2
       with:
         node-version: 16
+    - name: Determine deployer image
+      env:
+        PR_BODY: ${{ github.event.pull_request.body }}
+      run: |
+        export PR_DEPLOYER_IMAGE=$(echo $PR_BODY | grep -oP "deployer_image: \K\S+")
+        export DEPLOYER_IMAGE=${PR_DEPLOYER_IMAGE:-$DEPLOYER_IMAGE}
+        echo Using deployer image: $DEPLOYER_IMAGE
+        echo "DEPLOYER_IMAGE=$DEPLOYER_IMAGE" >> $GITHUB_ENV
     - name: Install dependencies
       run: |
         pip install -r requirements.txt
@@ -60,7 +68,7 @@ jobs:
         BASE_DOMAIN: ${{ secrets.DELTA_BASE_DOMAIN }}
       run: |
         export NAME=cdk-deploy-${GITHUB_SHA:0:6}
-        ./util.py generate_config_template --name $NAME --aws-region=us-west-2 --aws-account-id=$AWS_ACCOUNT_ID --dev --registry-username $REGISTRY_USERNAME --registry-password $REGISTRY_PASSWORD --hostname $NAME.$BASE_DOMAIN --acm-cert-arn $ACM_CERT_ARN --disable-flow-logs > config.yaml
+        ./util.py generate_config_template --name $NAME --aws-region=us-west-2 --aws-account-id=$AWS_ACCOUNT_ID --dev --platform-nodegroups 2 --registry-username $REGISTRY_USERNAME --registry-password $REGISTRY_PASSWORD --hostname $NAME.$BASE_DOMAIN --acm-cert-arn $ACM_CERT_ARN --disable-flow-logs > config.yaml
         ./util.py load_config -f ./config.yaml
     - name: Test default config (single and nested stacks)
       env:

diff --git a/cdk/domino_cdk/agent.py b/cdk/domino_cdk/agent.py
@@ -74,6 +74,7 @@ def generate_install_config(
                 "bucket": buckets["registry"].bucket_name,
             }
         },
+        "metrics_server": {"install": True},
         "gpu": {"enabled": True},
         "helm": {
             "cache_path": "charts",

diff --git a/cdk/domino_cdk/provisioners/eks/eks_iam.py b/cdk/domino_cdk/provisioners/eks/eks_iam.py
@@ -107,7 +107,9 @@ def provision(self, stack_name: str, cluster_name: str, r53_zone_ids: List[str],
                         "ec2:ModifyVolume",
                     ],
                     resources=["*"],
-                    conditions={"StringLike": {"aws:ResourceTag/deploy_id": cluster_name}},
+                    conditions={
+                        "StringLike": {f"aws:ResourceTag/kubernetes.io/cluster/{stack_name}": "owned"}
+                    },  # We can't use token cluster_name here, but should be the same
                 ),
                 iam.PolicyStatement(
                     effect=iam.Effect.ALLOW,