[Feature] Improving arrow flight sql with Kubernetes setup

[jonasbrami]

Search before asking

• I had searched in the issues and found no similar issues.

Description

Related to this merge request #43281.

I am afraid of the performance degradation of using Arrow Flight SQL from outside the Kubernetes cluster because from my understanding, it would be like setting enable_parallel_result_sink = False . Would it be theoretically possible, instead of having random BE resolution for the reverse proxy, have a smarter reverse proxy that could be, for example, based on path prefix (doris-url/be1, doris-url/be2) ? Then those doris-url/be{i} would become the "advertised" urls for the grpc client and would be reachable from outside the kubernetes cluster. Do you think it would be a good idea?

Use case

Faster queries when using arrow flight sql from outside Kubernetes cluster

Related issues

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct