Prerequisites

• Search the current open issues

What are you trying to do that currently feels hard or impossible?

I'm from the Google Cloud Dataproc team, and I'd like to add support for one of our runtime engines, Serverless Spark (full name Google Cloud Serverless for Apache Spark).

Mostly, the exposed tools should be thin wrappers around our batch and session CRUD APIs, as well as some undocumented APIs for fetching debugging information.

Sample tools:

• batch_list: list all batches in a project/location
• batch_get: get info on a single batch
• batch_create: create a batch (Spark workload)
• operation_cancel: cancel a long-running operation, such as a batch
• batch_metrics_get: get runtime metrics for a batch that may be useful for debugging

Example CUJs include:

• Integrating with coding agents to build and launch batches from project source code
• Checking the status of running batches.
• Debugging performance or other runtime issues with current or historical batches.

Suggested Solution(s)

No response

Alternatives Considered

No response

Additional Details

No response