feat(streaming): all stateful executors can try spill when chunk comes

[wcy-fdu]

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

What's changed and what's your intention?

Previously we have introduced mem table spill for hash_join, dynamic filter and materialized executor, which can reduce oom risk.
Other stateful exeutors, like top_n may also suffer from big mem table, so this pr add spill for all stateful executors.

perf test result: http://metabase.risingwave-cloud.xyz/dashboard/241-nexmark-blackhole-1cn-anti-affinity-rw-avg-source-throughput?start_date=2023-12-06&namespace=

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.

[codecov]

Codecov Report

Attention: 11 lines in your changes are missing coverage. Please review.

Comparison is base (5af41d3) 68.31% compared to head (200990b) 68.30%.
Report is 18 commits behind head on main.

Files	Patch %	Lines
...c/stream/src/executor/source/fs_source_executor.rs	0.00%	6 Missing ⚠️
...tream/src/executor/top_n/group_top_n_appendonly.rs	0.00%	3 Missing ⚠️
src/stream/src/executor/hash_agg.rs	87.50%	1 Missing ⚠️
src/stream/src/executor/simple_agg.rs	80.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #13751      +/-   ##
==========================================
- Coverage   68.31%   68.30%   -0.01%     
==========================================
  Files        1527     1527              
  Lines      262575   262615      +40     
==========================================
+ Hits       179378   179380       +2     
- Misses      83197    83235      +38     

Flag	Coverage Δ
rust	68.30% <72.50%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kwannoel]

Thanks for the PR. Can we run benchmark on this to make sure no performance degradation?

[wcy-fdu]

Thanks for the PR. Can we run benchmark on this to make sure no performance degradation?

Sure, I will perf this pr with daily benchmark later.

[stdrc]

LGTM, wait for benchmark

miss action

[fuyufjh]

Please fill the PR description..

[wcy-fdu]

Wait for the perf test. https://buildkite.com/risingwave-test/nexmark-benchmark/builds/2623

[kwannoel]

LGTM as well. Waiting for results.

[wcy-fdu]

perf test result: http://metabase.risingwave-cloud.xyz/dashboard/241-nexmark-blackhole-1cn-anti-affinity-rw-avg-source-throughput?start_date=2023-12-06&namespace=

• q3: 882->830
• q5: 140->176
• q7-rewrite: 975->926
• q8: 727->753
• q15: 487->503
• q16: 141->151
• q20: 394->410
• q101: 403->428
• q102: 255->270
• q103: 614->641
• q105: 300->340

In short, The performance of q3 and q7-rewrite has slightly decreased, and the other performance listed here has improved. The performance of the remaining unlisted queries remains basically unchanged.

Not sure if this is expected behavior, but the performance of most queries has actually improved. cc @stdrc @st1page

[st1page]

perf test result: http://metabase.risingwave-cloud.xyz/dashboard/241-nexmark-blackhole-1cn-anti-affinity-rw-avg-source-throughput?start_date=2023-12-06&namespace=

• q3: 882->830
• q5: 140->176
• q7-rewrite: 975->926
• q8: 727->753
• q15: 487->503
• q16: 141->151
• q20: 394->410
• q101: 403->428
• q102: 255->270
• q103: 614->641
• q105: 300->340

In short, The performance of q3 and q7-rewrite has slightly decreased, and the other performance listed here has improved. The performance of the remaining unlisted queries remains basically unchanged.

Not sure if this is expected behavior, but the performance of most queries has actually improved. cc @stdrc @st1page

q3 and q7-rewrite's degression LGTM and the performance is just similar to the nightly1203

I used to think the performance increase is because the memtable consumes less memory and the operator cache can consume more. But the metrics give the opposite phenomenon
q5 this branch

q5 nightly1205

I guess it is because the backpressure becomes more sensitive and make spikes smoother
q5 this branch

q5 nightly1205

Generally, the PR LGTM and I do not see negative effects...

[wcy-fdu]

spikes

I see, thanks. So let's merge.