Why does the message "not leader or follower" appear after writing to Kafka using a Java client?

[even0306]

I checked the leader status in the pod and found that both were present. Then I tried to push messages using kafka-producer-perf-test.sh in the pod, but the connection failed again.

This is the temporary authentication file I created inside the pod

So far, no errors have been reported in user-operator and topic-operator.

[im-konge]

You would get errors in User Operator or Topic Operator in case that you are doing something wrong with the KafkaUser or KafkaTopic custom resources. The producer is not related to this.
I don't think that the "not leader or follower" is the issue here. You are getting failures creating the client itself. For the SCRAM-SHA-512 I'm usually using the sasl.mechanism=SCRAM-SHA-512 and not "scram-sha-512", IDK if that's the issue.
Also, are you using the creds created for SCRAM-SHA-512 KafkaUser (just to be sure)?

[even0306]

I have been dealing with various errors in the past two days, so I always pay attention to the logs of user-operator and topic-operator. There is no error when creating kafkauser and topic. The authentication error of kafka-console-producer.sh has been solved. Thanks, it turns out that sasl.mechanism=SCRAM-SHA-512 cannot be included in "".
As for the error of not leader or follower, do you think it is more likely to be a problem with the Java client, or is there any missing configuration on the kafka server?

[even0306]

I found the reason. It was because I didn't know enough about the connection establishment mechanism of Kafka. I tried to access it from outside the cluster. I added three fqdns in /etc/hosts for resolution, but I transferred the fqdns of the three pods to one ip of bootstrap, so it could only find one and not the other two. If this problem does not exist in the cluster, I have now established three broker services separately, and then configured them separately in the external /etc/hosts, and it succeeded. Thanks

[im-konge]

Yeah I guess it's there based on the way you are connecting to Kafka. Great that you find the way to fix it and that it's working now :)