Update cloud build configs to deploy playground sandboxes

Fix kaniko cache invalidation issue

Author: aomarks

Dockerfile

@@ -44,6 +44,13 @@ RUN npx lerna run build --scope lit-dev-api --stream && \
 # Site content
 COPY packages/lit-dev-content/ ./packages/lit-dev-content/
 ARG PLAYGROUND_SANDBOX
+# Kaniko doesn't include ARG values in the layer cache key (see
+# https://github.com/GoogleContainerTools/kaniko/pull/1085). This is different
+# to normal Docker behavior, which would invalidate anything after the ARG
+# declaration if the value changes. So, we need to write it to the file system
+# to force a cache invalidation. Otherwise, we might re-use the most recent
+# Eleventy build output, even when the playground sandbox URL has changed.
+RUN echo "$PLAYGROUND_SANDBOX" > playground-sandbox
 RUN npx lerna run build --scope lit-dev-content --stream
 
 # Run the web service on container startup.

cloudbuild-main.yaml

@@ -8,29 +8,31 @@ steps:
   # https://docs.docker.com/engine/reference/commandline/build/
   # https://github.com/GoogleCloudPlatform/cloud-builders/tree/master/docker
   # https://cloud.google.com/build/docs/kaniko-cache
-  - id: Build
+  - id: build
     name: gcr.io/kaniko-project/executor
     args:
       - --dockerfile=./Dockerfile
-      - --destination=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/$_SERVICE_NAME:$COMMIT_SHA
+      - --destination=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/lit-dev:$COMMIT_SHA
       - --cache=true
       - --cache-ttl=168h # 1 week
+      # Bake in the playground sandbox url
+      - --build-arg=PLAYGROUND_SANDBOX=https://lit-dev-playground-5ftespv5na-uc.a.run.app/
 
-  # Create a new Cloud Run revision.
+  # Create a new Cloud Run revision for the main site.
   #
   # https://cloud.google.com/sdk/gcloud/reference/beta/run/deploy
   # https://github.com/GoogleCloudPlatform/cloud-sdk-docker
-  - id: Deploy
+  - id: deploy-main
     name: gcr.io/google.com/cloudsdktool/cloud-sdk
     entrypoint: gcloud
     args:
       - beta
       - run
       - deploy
-      - $_SERVICE_NAME
+      - lit-dev # Service name
       - '--region=$_DEPLOY_REGION'
       - '--platform=managed'
-      - '--image=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/$_SERVICE_NAME:$COMMIT_SHA'
+      - '--image=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/lit-dev:$COMMIT_SHA'
       - '--quiet'
       - '--no-traffic'
       - '--tag=master-$SHORT_SHA'
@@ -42,20 +44,80 @@ steps:
       - '--concurrency=40'
       - '--min-instances=1'
       - '--max-instances=1000'
+      # Serve the main site
+      - '--update-env-vars=MODE=main'
 
-  # Route traffic to new revision.
+  # Create a new Cloud Run revision for the Playground sandbox.
+  #
+  # https://cloud.google.com/sdk/gcloud/reference/beta/run/deploy
+  # https://github.com/GoogleCloudPlatform/cloud-sdk-docker
+  - id: deploy-playground
+    name: gcr.io/google.com/cloudsdktool/cloud-sdk
+    entrypoint: gcloud
+    waitFor:
+      # Don't wait for main site deploy
+      - build
+    args:
+      - beta
+      - run
+      - deploy
+      - lit-dev-playground # Service name
+      - '--region=$_DEPLOY_REGION'
+      - '--platform=managed'
+      - '--image=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/lit-dev:$COMMIT_SHA'
+      - '--quiet'
+      - '--no-traffic'
+      - '--tag=master-$SHORT_SHA'
+      # IMPORTANT: If you change --memory, be sure to also change
+      # --max-old-space-size in ./Dockerfile, and this same flag in
+      # ./cloudbuild-pr.yaml
+      - '--memory=1Gi'
+      - '--cpu=1'
+      - '--concurrency=40'
+      - '--min-instances=1'
+      - '--max-instances=1000'
+      # Serve the playground files
+      - '--update-env-vars=MODE=playground'
+
+  # Route traffic to new revision for the playground.
   #
   # https://cloud.google.com/sdk/gcloud/reference/beta/run/services
   # https://github.com/GoogleCloudPlatform/cloud-sdk-docker
-  - id: Route Traffic
+  - id: route-traffic-main
     name: gcr.io/google.com/cloudsdktool/cloud-sdk
     entrypoint: gcloud
+    waitFor:
+      # Route traffic simultaneously once both deploys are done
+      - deploy-main
+      - deploy-playground
     args:
       - beta
       - run
       - services
       - update-traffic
-      - $_SERVICE_NAME
+      - lit-dev # Service name
+      - '--region=$_DEPLOY_REGION'
+      - '--platform=managed'
+      - '--quiet'
+      - '--to-tags=master-$SHORT_SHA=100'
+
+  # Route traffic to new revision for the main site.
+  #
+  # https://cloud.google.com/sdk/gcloud/reference/beta/run/services
+  # https://github.com/GoogleCloudPlatform/cloud-sdk-docker
+  - id: route-traffic-playground
+    name: gcr.io/google.com/cloudsdktool/cloud-sdk
+    entrypoint: gcloud
+    waitFor:
+      # Route traffic simultaneously once both deploys are done
+      - deploy-main
+      - deploy-playground
+    args:
+      - beta
+      - run
+      - services
+      - update-traffic # Service name
+      - lit-dev-playground
       - '--region=$_DEPLOY_REGION'
       - '--platform=managed'
       - '--quiet'

cloudbuild-pr.yaml

@@ -12,29 +12,31 @@ steps:
   # https://docs.docker.com/engine/reference/commandline/build/
   # https://github.com/GoogleCloudPlatform/cloud-builders/tree/master/docker
   # https://cloud.google.com/build/docs/kaniko-cache
-  - id: Build
+  - id: build
     name: gcr.io/kaniko-project/executor
     args:
       - --dockerfile=./Dockerfile
-      - --destination=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/$_SERVICE_NAME:$COMMIT_SHA
+      - --destination=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/lit-dev:$COMMIT_SHA
       - --cache=true
       - --cache-ttl=168h # 1 week
+      # Bake in this revision's corresponding playground sandbox url
+      - --build-arg=PLAYGROUND_SANDBOX=https://pr$_PR_NUMBER-$SHORT_SHA---lit-dev-playground-5ftespv5na-uc.a.run.app/
 
-  # Create a new Cloud Run revision.
+  # Create a new Cloud Run revision for the main site.
   #
   # https://cloud.google.com/sdk/gcloud/reference/beta/run/deploy
   # https://github.com/GoogleCloudPlatform/cloud-sdk-docker
-  - id: Deploy
+  - id: deploy-main
     name: gcr.io/google.com/cloudsdktool/cloud-sdk
     entrypoint: gcloud
     args:
       - beta
       - run
       - deploy
-      - $_SERVICE_NAME
+      - lit-dev # Service name
       - '--region=$_DEPLOY_REGION'
       - '--platform=managed'
-      - '--image=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/$_SERVICE_NAME:$COMMIT_SHA'
+      - '--image=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/lit-dev:$COMMIT_SHA'
       - '--quiet'
       - '--no-traffic'
       - '--tag=pr$_PR_NUMBER-$SHORT_SHA'
@@ -46,6 +48,40 @@ steps:
       - '--concurrency=default' # unlimited
       - '--min-instances=0'
       - '--max-instances=1'
+      # Serve the main site
+      - '--update-env-vars=MODE=main'
+
+  # Create a new Cloud Run revision for the playground sandbox.
+  #
+  # https://cloud.google.com/sdk/gcloud/reference/beta/run/deploy
+  # https://github.com/GoogleCloudPlatform/cloud-sdk-docker
+  - id: deploy-playground
+    name: gcr.io/google.com/cloudsdktool/cloud-sdk
+    entrypoint: gcloud
+    waitFor:
+      # Don't wait for main site to deploy
+      - build
+    args:
+      - beta
+      - run
+      - deploy
+      - lit-dev-playground # Service name
+      - '--region=$_DEPLOY_REGION'
+      - '--platform=managed'
+      - '--image=$_GCR_HOSTNAME/$PROJECT_ID/$REPO_NAME/lit-dev:$COMMIT_SHA'
+      - '--quiet'
+      - '--no-traffic'
+      - '--tag=pr$_PR_NUMBER-$SHORT_SHA'
+      # IMPORTANT: If you change --memory, be sure to also change
+      # --max-old-space-size in ./Dockerfile, and this same flag in
+      # ./cloudbuild-main.yaml
+      - '--memory=1Gi'
+      - '--cpu=1'
+      - '--concurrency=default' # unlimited
+      - '--min-instances=0'
+      - '--max-instances=1'
+      # Serve the playground
+      - '--update-env-vars=MODE=playground'
 
 tags:
   - lit-dev