Scheduled triggers

Triggers ask the Envoy scheduler to start a job on a cron schedule. A trigger can run the full job or target a task node.

Create a trigger

Open Triggers and provide:

• a recognizable name;
• a job;
• optionally, a task node;
• a cron expression;
• an IANA time zone such as Europe/London;
• enabled or disabled state.

Save the trigger and verify the displayed next run.

Cron schedules

Envoy uses a five-field cron expression:

minute hour day-of-month month day-of-week

Examples:

0 2 * * *       Every day at 02:00
30 6 * * 1-5    Weekdays at 06:30
0 */4 * * *     Every four hours

The selected time zone controls how local clock time is interpreted. Use an IANA zone instead of a fixed UTC offset when daylight-saving behavior matters.

Full-job and task triggers

A full-job trigger starts the graph according to its dependencies.

A task trigger targets one DAG node. Use task triggers for independently repeatable stages, not as a substitute for a correctly modeled dependency graph. Confirm whether the selected node requires upstream output before scheduling it in isolation.

Enable and disable

Disable a trigger when maintenance or an external dependency makes new runs undesirable. Disabling affects future dispatch; it does not stop an already-started run.

Deleting a trigger removes the schedule, not the job or existing run history.

Scheduler operation

The control plane scheduler loads enabled triggers, calculates due times, and records scheduling state. The resulting job run is executed through the configured local or remote executor path.

When a deployment has no eligible executor capacity, the trigger can be accepted but execution cannot progress normally. Monitor both trigger state and the job run.

Operational checklist

• Verify the displayed next run after every schedule edit.
• Use the deployment's intended IANA time zone.
• Avoid starting the same non-idempotent job from several triggers.
• Make task outputs idempotent before adding retries or frequent schedules.
• Disable schedules before an upgrade that requires draining active work.
• Review resulting job runs and alerts; a saved trigger is not proof of successful execution.