Deployment Overview

This chapter separates cluster installation from optional services. Deploy and verify the Curvine Master and Workers first. Add the Transfer service only when Load or Export jobs should run outside the Master process.

Deployment Sequence

1. Choose a cluster form.
   • Use Quick Start for a local evaluation.
   • Use Standalone Mode for a single-host cluster.
   • Use bare metal or Kubernetes for a distributed cluster.
2. Verify that Master and Worker processes are healthy and that required UFS mounts can be listed with cv mount --check.
3. If the cluster needs independent Load and Export scheduling, deploy the Transfer Service.
4. Enable Transfer only in the clients that should use it. Existing clients can continue to use the legacy Master job path during a staged rollout.

Documentation Boundaries

• Cluster deployment pages describe Master, Worker, storage, and networking.
• Transfer pages describe the independent scheduler, its state store, and client routing.
• The configuration reference is the single source for configuration defaults. Deployment pages show only the settings required for each topology.

Production Changes

Treat upgrades as a separate operation. Verify version compatibility and configuration changes in a non-production cluster before changing the running cluster. For Transfer, use a shared MySQL store before running more than one instance.