Skip to content
ReplicaDBReplicaDB

CLI guide

The standalone CLI performs direct bulk replication from one source to one sink, or through a sequential multi-table catalog. It is Spring-free, runs on Java 17 or newer, and does not require a ReplicaDB metadata database.

Follow a workflow

Section titled “Follow a workflow”

The CLI option reference is generated from the maintained option evidence and preserves exact flag names. The options-file examples show safe environment substitution without resolving secrets into source control.

What one invocation does

Section titled “What one invocation does”
  1. Parse the options file, then apply command-line overrides.
  2. Validate the source, sink, mode, table selection, and staging constraints.
  3. Open connector managers and run their pre-transfer preparation.
  4. Partition the current source table across jobs workers and write rows to the sink or its staging area.
  5. Run the mode-specific sink operation, cleanup, and connector shutdown.

A multi-table catalog repeats that complete lifecycle for each table pair in numeric order. It does not turn one invocation into a durable workflow: keep the options, exit code, logs, destination checks, and last committed watermark with the external scheduler or run record that launched it.

Choose the next guide

Section titled “Choose the next guide”
  • Start with complete when replacing the destination is acceptable.
  • Choose complete-atomic when readers must not see the ordinary replacement window and the sink supports transactional staging.
  • Choose incremental when a source predicate or committed watermark can identify changed rows and the sink has the keys required to merge them.
  • Use source.query for a result set that cannot be expressed as one table, but accept that automatic watermark and multi-table features are unavailable.

Scope and limits

Section titled “Scope and limits”

The CLI is a batch replication tool. It does not provide change-data-capture, durable server-side scheduling, remote run cancellation, or automatic resume after a process is interrupted. Use the managed server when those operations need durable shared state.