planner: fix wrong prepare plan after isolation read changed

[lzmhhh123]

What problem does this PR solve?

Issue Number: close #16256

Problem Summary: After isolation read changed, the prepared statement should be re-prepared.

What is changed and how it works?

What's Changed: clear prepared plan when isolation read changes.

Related changes

• Need to cherry-pick to the release branch

Check List

Tests

• Unit test
• Integration test

Side effects

• Performance regression
  • Consumes more CPU

Release note

• Make plan cache consider isolation read engines.

[crazycs520]

How about not use reflect.DeepEqual?

func prepare() (map[string]struct{}, map[string]struct{}) {
	m1 := make(map[string]struct{}, 3)
	m1["tikv"] = struct{}{}
	m1["tidb"] = struct{}{}
	m1["tiflash"] = struct{}{}
	m2 := make(map[string]struct{}, 3)
	m2["tikv"] = struct{}{}
	m2["tidb"] = struct{}{}
	m2["tiflash"] = struct{}{}
	return m1, m2
}

func Benchmark_cmp1(b *testing.B) {
	m1, m2 := prepare()
	for i := 0; i < b.N; i++ {
		reflect.DeepEqual(m1, m2)
	}
}

func Benchmark_cmp2(b *testing.B) {
	m1, m2 := prepare()
	for i := 0; i < b.N; i++ {
		checkEqualMap(m1, m2)
	}
}

func checkEqualMap(m1, m2 map[string]struct{}) bool {
	if len(m1) != len(m2) {
		return false
	}
	for k := range m1 {
		if _, ok := m2[k]; !ok {
			return false
		}
	}
	return true
}

▶ go test -run '^$' -bench .
goos: darwin
goarch: amd64
pkg: github.com/crazycs520/test
Benchmark_cmp1-12        2272569               521 ns/op
Benchmark_cmp2-12       21689436                54.1 ns/op
PASS
ok      github.com/crazycs520/test      2.955s

[codecov]

Codecov Report

Merging #16293 into master will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff             @@
##             master     #16293   +/-   ##
===========================================
  Coverage   79.8503%   79.8503%           
===========================================
  Files           520        520           
  Lines        139888     139888           
===========================================
  Hits         111701     111701           
  Misses        19231      19231           
  Partials       8956       8956           

[danmay319]

LGTM

[sre-bot]

/run-all-tests

[zz-jason]

clear prepared plan when isolation read changes.

I prefer to add isolation read engines to the key of the prepared plan cache. It's different from schema change. schema version change is unpreventable, but isolation read engine change is. Once the user changed back to the original value of the isolation read engine, the cache can be used again at that time.

@eurekaka What's your opinion?

[eurekaka]

clear prepared plan when isolation read changes.

I prefer to add isolation read engines to the key of the prepared plan cache. It's different from schema change. schema version change is unpreventable, but isolation read engine change is. Once the user changed back to the original value of the isolation read engine, the cache can be used again at that time.

@eurekaka What's your opinion?

+1

[lzmhhh123]

clear prepared plan when isolation read changes.

I prefer to add isolation read engines to the key of the prepared plan cache. It's different from schema change. schema version change is unpreventable, but isolation read engine change is. Once the user changed back to the original value of the isolation read engine, the cache can be used again at that time.
@eurekaka What's your opinion?

+1

OK, I'll change it to that way.

[lzmhhh123]

/rebuild

[zz-jason]

LGTM

[eurekaka]

LGTM

[sre-bot]

Your auto merge job has been accepted, waiting for:

• 17541

[sre-bot]

/run-all-tests

[sre-bot]

@lzmhhh123 merge failed.

[lzmhhh123]

/merge

[lzmhhh123]

/merge

[lzmhhh123]

/merge

[sre-bot]

/run-all-tests

[sre-bot]

@lzmhhh123 merge failed.

[sre-bot]

/run-all-tests

[sre-bot]

@lzmhhh123 merge failed.

[sre-bot]

/run-all-tests

[sre-bot]

@lzmhhh123 merge failed.

[lzmhhh123]

/run-cherry-pick

[zz-jason]

/run-cherry-picker

[sre-bot]

cherry pick to release-4.0 in PR #17570