This message was deleted.

This does not happen in Pulsar Client CLI

A connection timeout error indicates a network issue. Is the Pulsar proxy reachable from the environment where you are running Flink? Does the code work using a Broker IP instead of the proxy?

@David K So we have client application, pulsar proxies and pulsar brokers all in separate VPCs. We do not want to allow clients to directly access brokers from client application VPC, so we have not paired client application and broker VPCs. Pairing is done between client application and proxy VPCs. We are giving proxy end point in Flink sink operator. Our expectation is for pulsar Flink connector to connect to proxy instances only and not broker instances. But the ip you see in failure logs is a broker ip. Since VPC pairing is not there, this connection is failing. If we use Pulsar CLI Client on application instances to consume data from a pulsar topic using proxy end point, it consumes without any issue, which means it internally does not make any direct connection to pulsar brokers. Why the behaviour is different in Flink connector?

Thanks for the detailed explanation. I was going to have you double-check your proxy configuration to ensure it has the correct URL for the brokers, but that doesn’t appear to be the case. I will forward this to our engineering team. Can you tell me the version of the Pulsar proxy and Flink connector? What is your security configuration, e.g. mTLS for wire encryption? What authN mechanism are you using?

cc @Yubiao Feng It looks like a similar issue we countered recently.