broker failed to restart the worker instance of Pulsar IO.

[viggo-rift]

When I use Pulsar Admin to create a Kafka IO connector and then restart the broker, the previously started connectors fail to start. I checked the broker startup log and found that when FunctionActioner#startFunction is executed, an HTTP 503 Service Unavailable error is reported. Is the startup and deployment of these management streams, admin APIs, and Functions, as well as the restart of Worker instances based on metadata, carried out asynchronously? Can the sequence of these two not be guaranteed? Is this a problem?