This server runs on Google Cloud Run, and provides the most recent test results from the Dart CI (continuous integration) testing visible at https://ci.chromium.org/p/dart/g/be/console
The service takes about two minutes to load all the existing latest results.json files from all the configurations, so it's reponse to the first request after deploying will take at lest that long. When deploying, a health check verifies that it is responding to web requests, and this health check has a 4 minute timeout. The instance started for the health check does not seem to remain alive, so the first real request has the same startup delay.
To generate the Dart code reading the records from results.json, run protogen on the .proto file declaring them:
protoc --dart_out=gen ../common/result.proto
To build the server and deploy to cloud run on staging, run
gcloud builds submit --project=dart-ci-staging --tag gcr.io/dart-ci-staging/current_results gcloud run deploy --project=dart-ci-staging current-results --image gcr.io/dart-ci-staging/current_results --platform=managed --allow-unauthenticated
Unauthenticated access is allowed only while the server being developed does not modify any data, remove that flag before adding APIs that modify data.
To build the server and deploy to cloud run on production, run
gcloud builds submit --project=dart-ci --tag gcr.io/dart-ci/current_results gcloud run deploy --project=dart-ci current-results --image gcr.io/dart-ci/current_results --platform=managed
The server reads the current results.json and flaky.json from cloud storage at gs://dart-test-results/configuration/master/[configuration name]/[build number]/ and stores selected fields from those records in memory. (Flaky.json reads not implemented).
The server exposes a service to Pub/Sub on the same project, which is triggered by any object creation on the dart-test-results bucket. If the object is a /configuration/master/[configuration name]/latest file, the build number is read from that file and the cached results.json and flaky.json are updated from that build.
A gRPC service is defined, with a query message type and a response message type defined by protocol buffers. It will allow queries by test name or test name prefix and by configuration or configuration set.
APIs will be defined to query current flakiness data, and to revoke the flakiness mark on test-configuration pairs. When designing the API to revoke flakiness, security and auditing concerns must be addressed by requiring authentication.