[bug] ProducerBusy
XLPE opened this issue · 1 comments
Pulsar clients periodically call the internalCreatePartitionsProducers method, which we found to be prone to timeouts, triggering the checkServerError method to close the connection and trigger producer reconnection. Since the Pulsar broker is not able to detect the disconnection of the client, the client frequently receives an error message stating that "xxx is already connected to the topic" upon reconnection attempts. This situation can last for several minutes, until the server closes the connection due to keepalive timeout, after which the client can reconnect.
Expected behavior
The invocation of internalCreatePartitionsProducers should not affect the connection of producers.
Actual behavior
The invocation of internalCreatePartitionsProducers causes all connections to be disconnected, resulting in producers being unable to connect to the broker for a long time.