Observability Security And Ops

An invocation appears stuck. The operator restarts the worker pod hoping to 'unstick' it instead of inspecting why it is suspended.

Inspect the invocation via the CLI / admin introspection (its status, which durable promise or timer it is awaiting, its journal position) to learn WHY it is suspended — a long sleep, an unresolved awakeable, a retrying step. Restarting the worker does nothing for a durably-suspended invocation. Di…

A multi-service flow fails somewhere and the operator has no correlated trace, so they grep logs across services by hand.

Enable Restate's tracing export (OpenTelemetry) so cross-service invocations, ctx.run steps, and awaits are correlated under one trace per invocation; drill into the failing span instead of grepping. Propagate trace context into ctx.run side effects. [REQUIRES-VERIFICATION] for the exact OTel expor…

The Restate ingress is exposed so external clients can invoke services, with no authentication in front of it — anyone who can reach it can invoke any handler.

Put authentication / authorization in front of the ingress (API gateway, mTLS, signed requests, or the documented Restate auth mechanism) so only authorized callers invoke handlers; an open ingress lets anyone trigger durable side effects. Scope credentials to the minimum services needed. [REQUIRES…