Scheduling guide
Compose | Kubernetes | workerStopTimeoutDeployment and shutdown — Scheduling & Temporal
Generated environments and Temporal-aware worker drain order.
When to use
Use this page while designing, implementing, operating, or troubleshooting scheduled and durable execution.
Behavior
- Compose provisions PostgreSQL, Temporal Server, namespace initialization, admin tools, and UI before services.
- Kubernetes output includes chart values, PostgreSQL, admin/UI, and metrics discovery.
- Shutdown first drains native network handlers, then ordinary Sources and Worker admission, followed by producer pools/timers, nested graph work, Sinks, and finally durable clients.
Temporal-aware shutdown order
NetworkHTTP and gRPC listenersReject new requests and drain accepted handlers while graph dependencies remain alive.
SourcesCron and ordinary DataSourcesStop root admission and drain accepted source work.
WorkersTemporal admissionStop polling new tasks only after ordinary Sources drain.
ProducersPools, timers, storageDrain accepted producer work before checking graph tasks.
GraphNested workDrain ParallelCall tasks after producers quiesce.
SinksDataSinksFinish accepted outbound submissions.
ClientsTemporal connector coreClose durable clients after Sinks no longer need them.
TelemetryMetrics, traces, logsFlush last.
Timeouts properties
Related but distinct limits.
PropertyTypeDescription
service.shutdownTimeoutmillisecondsGlobal graceful-shutdown deadline.
Minimum:0connector.workerStopTimeoutmillisecondsWorker drain cap; zero inherits remaining service deadline.
Default:0Minimum: 0Product-level example
service.shutdownTimeout: 30000
temporalConnector.workerStopTimeout: 0