ddl: make "add foreign key" step in multiple schema versions

[crazycs520]

Signed-off-by: crazycs520 crazycs520@gmail.com

What problem does this PR solve?

Issue Number: close #38678

At the same point in time, there may be two versions of the schema in the TiDB cluster, so we can't create foreign key in one schema version, since this may break foreign key constraint, such as delete reference table
without foreign key constrain check in child table.

-- In TiDB-1 and Schema Version is 1
insert into t_has_foreign_key values (1, 1);
-- In TiDB-0 and  Schema Version is 0
delete from t_reference where id = 1; --Since doesn't know foreign key information in old version, so doesn't do foreign key constrain check.

So The DDL owner handle add foreign key constrain step should be:

1. None -> Write Only: add foreign key constrain which state is write-only into table.
2. Write Only - Write Reorg: check all row in the table whether has related foreign key exists in reference table, we can use following SQL to check:

   select count(*) from t2 where t2.a not in (select id from t1);

   The expected result is 0, otherwise, an error is returned and cancel the ddl job.
3. Write Reorg -> Public: update the foreign key constraint state to public.

A problem is, How the DML treat the foreign key with on delete/update cascade behaviour which state is non-public?
Here is an example:

create table t1 (id int key,a int, index(a));
create table t2 (id int key,a int, index(a));
insert into t1 values (1,1);
insert into t2 values (1,1);
alter  table t2 add constraint fk_1 foreign key (a) references t1(id) ON DELETE CASCADE;

The schema change of foreign key fk_1 is from None -> Write-Only -> Write-Reorg -> Public。
When the foreign key fk_1 in Write-Only state, a DML request has come to be processed:

delete from t1 where id = 1;

Then, TiDB shouldn't do cascade delete for foreign key fk_1 in state Write-Only, since the Add Foreign Key DDL job maybe
failed in Write-Reorg state and rollback the DDL job. But it is hard to rollback the cascade deleted executed before.

So, when execute DML with non-publick foreign key, TiDB will do foreign key constraint check instead of foreign key cascade behaviour.

What is changed and how it works?

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

Side effects

• Performance regression: Consumes more CPU
• Performance regression: Consumes more Memory
• Breaking backward compatibility

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

Please refer to Release Notes Language Style Guide to write a quality release note.

None

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• lcwangchao
• tangenta

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

The full list of commands accepted by this bot can be found here.

Reviewer can indicate their review by submitting an approval review.
Reviewer can cancel approval by submitting a request changes review.

[crazycs520]

/merge

[ti-chi-bot]

This pull request has been accepted and is ready to merge.

Commit hash: 3c3d2c0

[ti-chi-bot]

@crazycs520: Your PR was out of date, I have automatically updated it for you.

At the same time I will also trigger all tests for you:

/run-all-tests

If the CI test fails, you just re-trigger the test that failed and the bot will merge the PR for you after the CI passes.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the ti-community-infra/tichi repository.

[crazycs520]

/merge

[crazycs520]

/run-all-tests

[crazycs520]

/merge