Analytics 2 documented workflows
Dataproc
Cluster and job workflows with optional local runtime execution.
Documented workflows
2 workflows- clusters.create/list/get/delete
- jobs submit/list/get
Service boundaries
- Autoscaling and Kubernetes cluster mode are not complete.
- Runtime execution requires Docker socket access; without it, cluster metadata remains available but Spark/Hadoop processes are not started.
- jobs submit/list/get: Cluster jobs run in the Dataproc runtime container; jobs without a cluster use ephemeral serverless containers. Execution requires Docker socket access and a compatible runtime image; otherwise only metadata paths are available.
- Audited metadata survives restart while the LocalCloud volume is retained
- External runtime/data-plane side effects, if any, have separate lifecycle and recovery limits.
Next steps
- AI agent local testing: prompts and endpoint routing for local workflows.
- Local cloud for AI agents: give agent-written cloud code a credentialless target.
- Compatibility: compare all 27 service guides.
- Configuration: service selection, tiers, and runtime precedence.