Paul Lin

@hadoopkandy Thanks for updating! I have one concern regarding the change: in the current version, if one fetch times out, the engine returns `hasNext==ture`, it's up to the client to...

> when timeout is reached, `hasNext` should be `true` as the streaming query is not finished. `beeline` supports `--incremental=true` to print the fetched records immediately. How about adding a `exit-on-fetch-timeout`...

> I'm not very familiar with the streaming process, does Flink/Beam have such a switch to control the behavior of timeout reached but query not completed? Flink SQL Gateway exits...

> The engine should not see the configured `kyuubi.frontend.thrift.binary.bind.port` I find the engines use `kyuubiConf.setIfMissing(KyuubiConf.FRONTEND_THRIFT_BINARY_BIND_PORT, 0)`. Could we just simply override `kyuubi.frontend.thrift.binary.bind.port` with `0` on the engine side?

@pan3793 Understood. But weird enough, I couldn't find the codes related to the filtering. Could you give me some pointers? Another possible approach is filtering config options with `isServerOnly`, WDYT?

ping @pan3793

@bobo495 Thanks for reporting the problem. I agree that the current implementation of the result retrieval of the Flink engine is improvable. The final solution would be refactoring it to...

@yuruguo Thanks a lot for your help! In another way, we could turn to the incremental fetching approach, which solves the problem once for all. I've started some related work...

@weaksloth Thanks for your feedback. I agree Flink engine is not production ready yet and we have a plan to improve this situation (see https://github.com/apache/kyuubi/issues/2100). If things go well, we...

@weaksloth you're right. I've confirmed that the connection would be blocked by Flink insert operations. The problem lies in https://issues.apache.org/jira/browse/FLINK-24461, which refactors the fetching of Flink table results and breaks...