Alex Chi Z.

Also I guess this issue happens because the future is being run on another core than Instant::now is captured.

For planner test, please refer to https://github.com/risingwavelabs/risingwave/tree/main/src/frontend/test_runner, where we use yaml files to do regression test on plans. We can do the same for fragmenter test.

That's cool! cc @xxchan this might be related to the TPC-H Q18 issue. PTAL if you're interested. Seems to be originated from nexmark connector. No problem.

dup w/ https://github.com/singularity-data/risingwave/issues/3607, I've proposed it long before!

+1. No shuffle exchange looks bad when there's scale-in / scale-out. If we don't want to handle this as a special case when scale, this proposal looks good.

may also close https://github.com/singularity-data/risingwave/issues/3607

> I found that the compute node has the problem of a sharp increase in memory Do you know when the sharp increate starts? e.g. when create materialize view, or...

What's your configuration? Would you please paste your docker-compose file here?

Also the full process of running this? e.g. what MV did you create, how's the source look like.

min / max's implementation is significantly different from other aggregators. I guess it is the root cause of your issue.