feat(udf): support client side load balancer for UDF

[KeXiangWang]

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

Support client side load balancer, which read IP list from DNS server and load balance requests to the IPs.

Background:

• RW's UDF client use gRPC protocol to exchange data with UDF servers.
• gRPC is based on HTTP2.0, requests of which reuse a single TCP connection.
• L4 balancer make routing decisions based on information at the TCP or UDP level without understanding higher-level protocols, such as HTTP2.0.
• L4 is commonly used in K8S environment.
• If one RW client tries to connect a group of UDF servers behind an L4 balancer, only one server will be accessed.

#15138

Checklist

• I have written necessary rustdoc comments
• I have added necessary unit tests and integration tests
• I have added test labels as necessary. See details.
• I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
• My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
• All checks passed in ./risedev check (or alias, ./risedev c)
• My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)

• My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

• My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

If this PR includes changes that directly affect users or other significant modifications relevant to the community, kindly draft a release note to provide a concise summary of these changes. Please prioritize highlighting the impact these changes will have on users.

[BugenZhao]

May I ask whether there's any downside for always enabling the periodical service discovery? IIUC, if there's no loading balancing under the given domain name, it'll degrade to the normal case and everything should still work well.

[KeXiangWang]

if there's no loading balancing under the given domain name, it'll degrade to the normal case and everything should still work well.

Yes, it will degrade to normal case.

May I ask whether there's any downside for always enabling the periodical service discovery?

If there's a L7 load balancer behind, the client-side load balancing is redundant, which issues an extra thread to refresh the IP list and has some extra cost on routing requests.

[BugenZhao]

which issues an extra thread to refresh the IP list and has some extra cost on routing requests

I guess this is quite acceptable. Also IIUC, the interval for DNS lookup only impacts the responsiveness to the changes behind the L4 load balancer and does not hurt the functionality of loading balancing itself. Therefore, using a longer interval should also be fine.

[KeXiangWang]

Update to always enabling client-side LB. The DNS refresh interval is set to be 5 seconds.

[kwannoel]

Rest LGTM

[wangrunji0408]

LGTM!

[BugenZhao]

By inspecting the source, it appears that .channel() is actually synchronous. We may use .now_or_never().unwrap() on the future here.

[kwannoel]

Hmm will it be much higher cost to do block_on + await here? I prefer not to rely on internal implementations of library.

[BugenZhao]

futures::executor::block_on is fragile as well. If the future implementation leverages I/O capabilities of tokio inside, it will also panic.

[KeXiangWang]

Updated to .now_or_never().unwrap()

[wangrunji0408]

LGTM!