buildscripts: Convert GAE CI to Cloud Build (1.76.x backport) #12631
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The Google App Engine build now requires Java 17, because the App Engine libraries are now using Java 17 bytecode. The Kokoro environment doesn't include Java 17, and while we could make some custom pools to resolve it, it is easier to swap to Cloud Build than to fight and maintain the Kokoro images. With Cloud Build we can also restrict permissions easier, as the same workers aren't used for multiple tasks. However, the Gradle App Engine plugin doesn't support choosing a service account for GAE, so I swapped to using gcloud app deploy. Although we'll be using restricted service accounts, we'll configure Cloud Build to require a "/gcbrun" GitHub comment except for owners and collaborators of the repository, similar to the "kokoro:run" label today. I swapped the Gradle code to use project properties instead of system properties, as we really should have been using project properties to begin with and I didn't want to add new system properties. The sleep has probably been unnecessary since the turndown of GAE Java 7, when the architecture of GAE changed considerably. But today it is very possible a new instance is spun up for that request and GAE does a warmup request, so the delay seems unlikely to help anything and was excessive at 20 seconds. The Cloud Build file _doesn't_ include GAE in its name because it can do more than GAE testing and it is easy to run things in parallel in Cloud Build (although they share the worker). In particular, some of the Android tests may make sense to migrate away from Kokoro. We're using e2-standard-16 for Kokoro and it takes about 10 minutes. With the default Cloud Build worker e2-standard-2 it takes 20 minutes, and with e2-highcpu-8 it takes 10 minutes with 4 minutes spent on app deploy. The expectation is to run this with a Java-CI-specific service account, so we have configure logging ourselves. I chose CLOUD_LOGGING_ONLY because it was easy, but we'll want to configure GCS in the future to allow external contributors to see the logs.
kannanjgithub
approved these changes
Jan 27, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The Google App Engine build now requires Java 17, because the App Engine libraries are now using Java 17 bytecode. The Kokoro environment doesn't include Java 17, and while we could make some custom pools to resolve it, it is easier to swap to Cloud Build than to fight and maintain the Kokoro images. With Cloud Build we can also restrict permissions easier, as the same workers aren't used for multiple tasks.
However, the Gradle App Engine plugin doesn't support choosing a service account for GAE, so I swapped to using gcloud app deploy.
Although we'll be using restricted service accounts, we'll configure Cloud Build to require a "/gcbrun" GitHub comment except for owners and collaborators of the repository, similar to the "kokoro:run" label today.
I swapped the Gradle code to use project properties instead of system properties, as we really should have been using project properties to begin with and I didn't want to add new system properties. The sleep has probably been unnecessary since the turndown of GAE Java 7, when the architecture of GAE changed considerably. But today it is very possible a new instance is spun up for that request and GAE does a warmup request, so the delay seems unlikely to help anything and was excessive at 20 seconds.
The Cloud Build file doesn't include GAE in its name because it can do more than GAE testing and it is easy to run things in parallel in Cloud Build (although they share the worker). In particular, some of the Android tests may make sense to migrate away from Kokoro.
We're using e2-standard-16 for Kokoro and it takes about 10 minutes. With the default Cloud Build worker e2-standard-2 it takes 20 minutes, and with e2-highcpu-8 it takes 10 minutes with 4 minutes spent on app deploy.
The expectation is to run this with a Java-CI-specific service account, so we have configure logging ourselves. I chose CLOUD_LOGGING_ONLY because it was easy, but we'll want to configure GCS in the future to allow external contributors to see the logs.
See https://console.cloud.google.com/cloud-build/builds/966f7962-aa17-4e6c-b41c-b8e07d0feddb?project=grpc-testing for an example run (if you have access to the project).
Backport of #12423. This isn't that important to backport; we can just ignore the CI status. But it's "one less thing."