[Feature] LeastActiveLoadBalance does not properly utilize weights.

[howie-xu]

Pre-check

• I am sure that all the content I provide is in English.

Search before asking

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

Apache Dubbo Component

Java SDK (apache/dubbo)

Descriptions

In LeastActiveLoadBalance, instances are first filtered by the minimum number of active connections. If there is more than one instance with the minimum number of active connections, a random selection is made based on weights. This design seems somewhat unreasonable, and the warmup feature doesn't have much effect (a batch of instances that are restarted or scaled up will still suddenly receive a large number of requests). Could we use the weights more effectively during the minimum connection filtering phase, for example, by using active * afterWarmup to determine the least active instance?

Related issues

No response

Are you willing to submit a pull request to fix on your own?

• Yes I am willing to submit a pull request on my own!

Code of Conduct

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

[wcy666103]

Please submit in English Only

[AlbumenJ]

Yes, we need a new algorithm

[howie-xu]

Can the actual number of active be directly changed to "active * afterWarmup" in the LeastActiveLoadBalance?

[AlbumenJ]

Can the actual number of active be directly changed to "active * afterWarmup" in the LeastActiveLoadBalance?

This may be not well enough. In active * afterWarmup, the newly launched node will get lower score and will be more high weighed.