Skip to main content
Recurring runs are delegated to an orchestrator. Two are supported, and they are interchangeable: the client for each implements the same method surface, so nothing in the calling code changes when you switch.
Only these two values change behavior. Anything else is read as an unrecognized setting and falls back to mage — there is no third backend.
The setting is read by the listing endpoints too, not only when a schedule is created. A deployment that flips the flag without restarting keeps reading the previous orchestrator, and the runs created on the new one never appear.
Orchestrator screen listing agents with their model, schedule and run actions

The orchestrator screen: one row per agent, with its model, its schedule and a manual trigger.

Scheduling a run

Interval shortcuts

Mage accepts the @interval shortcuts directly. The th2etl client translates them into five-field cron expressions before creating the trigger, so the same request body works against either backend.

Managing schedules and runs

The endpoints keep the pipelines naming from the Mage era. They are orchestrator agnostic: the same calls reach th2etl when the flag is set.

What changes between the two

The mapping matters if you inspect the orchestrator directly rather than through this API.
Identifiers are opaque to callers — read one from a listing and pass it back. Do not compare them numerically: Mage ids are integers while th2etl uses names, and a numeric comparison against a th2etl id silently fails to match.

Configuration

See Configuration for the variables each orchestrator needs. When ORCHESTRATOR is unset, the server talks to Mage — so a deployment that means to use th2etl must set it explicitly.