Summary

• add temporalio.contrib.gcp.OpenTelemetryPlugin, a thin GCP-focused wrapper around the existing OpenTelemetry plugin
• export workflow traces and Core metrics over OTLP/gRPC to a local collector sidecar, with Cloud Run-aware service-name defaults
• default Core metric exports to 60 seconds, matching the coordinated Java/Go/Python and upstream OpenTelemetry SDK cadence, while preserving metric_periodicity= overrides
• keep tracer-provider and runtime ownership explicit, including trace flush/shutdown helpers and worker-stop opt-in
• document separate Google collector pipelines: cumulative metrics without batching, traces with a dedicated batch processor
• add the gcp-opentelemetry dependency extra, usage documentation, and focused tests

The worker does not use Google Cloud clients, exporters, or resource detection directly. The collector is responsible for authenticating, enriching telemetry with GCP resource attributes, and exporting it to Google Cloud.

Google Managed Service for Prometheus must receive each cumulative OTLP metric export separately. A collector batch processor can combine a periodic export with a shutdown-time export and cause Duplicate TimeSeries rejection regardless of the configured periodicity. Trace batching remains safe and recommended.

Related implementations

• Java: Add Google Cloud OpenTelemetry plugin sdk-java#2955
• Go: Add Google Cloud OpenTelemetry plugin sdk-go#2467
• Python Cloud Run sample: Add GCP Cloud Run OpenTelemetry sample samples-python#329

Validation

• 59 original focused and regression tests passed
• 13 GCP plugin tests passed after the cadence change, including a literal 60-second default assertion and an explicit 30-second override assertion
• ruff check and ruff format --check for the GCP implementation/tests
• pyright, mypy, and basedpyright for the GCP implementation/tests
• documentation generation and uv lock --check
• Cloud Run worker-pool and Temporal Cloud E2E verified workflow execution, the unoverridden 60-second metrics cadence, metrics and trace export, GCP resource attribution, and clean shutdown without duplicate-series or exporter errors

The Python runtime currently has no public explicit Core-metrics force-flush API. The no-batch metrics guidance prevents a periodic and runtime-shutdown export from being combined if such an export occurs.