WIP: add optional flag to enable multi shard autocommit by default

[demmer]

Description

Note this is a RFC Draft PR -- soliciting feedback on the idea / proposed implementation.

Adds a new vtgate flag --default-multi-shard-autocommit which, as the name implies, opts the query engine into using multi-shard autocommit semantics by default, even if the plan does not contain the query directive MULTI_SHARD_AUTOCOMMIT.

The aim of this change is to help protect Slack or other large scale Vitess adopters that use autocommit from inadvertently expensive extra round trips on scatter DMLs.

As a general rule, Slack avoids scatter DMLs, but on occasion when they are needed, prefer the semantics of the MULTI_SHARD_AUTOCOMMIT=1 query directive (in which each shard executes its own autocommit DML) as opposed to the default behavior (in which each shard runs the DML in a transaction and vtgate issues a second round trip to commit).

We generally apply this as a directive in the application code, but that doesn't prevent mistakes where developers can inadvertently forget to apply the directive.

This PR adds an option to change the default behavior so that this is opted in by default. It should not affect any semantics on explicit transactions, nor should it affect any deployments that don't enable the feature.

Related Issue(s)

None.

Checklist

• "Backport to:" labels have been added if this change should be back-ported to release branches
• If this change is to be back-ported to previous releases, a justification is included in the PR description
• Tests were added or are not required
• Did the new or modified tests pass consistently locally and on CI?
• Documentation was added or is not required

Deployment Notes

[vitess-bot]

Review Checklist

Hello reviewers! 👋 Please follow this checklist when reviewing this Pull Request.

General

• Ensure that the Pull Request has a descriptive title.
• Ensure there is a link to an issue (except for internal cleanup and flaky test fixes), new features should have an RFC that documents use cases and test cases.

Tests

• Bug fixes should have at least one unit or end-to-end test, enhancement and new features should have a sufficient number of tests.

Documentation

• Apply the release notes (needs details) label if users need to know about this change.
• New features should be documented.
• There should be some code comments as to why things are implemented the way they are.
• There should be a comment at the top of each new or modified test to explain what the test does.

New flags

• Is this flag really necessary?
• Flag names must be clear and intuitive, use dashes (-), and have a clear help text.

If a workflow is added or modified:

• Each item in Jobs should be named in order to mark it as required.
• If the workflow needs to be marked as required, the maintainer team must be notified.

Backward compatibility

• Protobuf changes should be wire-compatible.
• Changes to _vt tables and RPCs need to be backward compatible.
• RPC changes should be compatible with vitess-operator
• If a flag is removed, then it should also be removed from vitess-operator and arewefastyet, if used there.
• vtctl command output order should be stable and awk-able.

[codecov]

Codecov Report

Attention: Patch coverage is 72.72727% with 3 lines in your changes missing coverage. Please review.

Project coverage is 67.69%. Comparing base (de33a39) to head (34774af).

Files with missing lines	Patch %	Lines
go/vt/vtgate/vtgate.go	0.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #17635      +/-   ##
==========================================
+ Coverage   67.68%   67.69%   +0.01%     
==========================================
  Files        1586     1586              
  Lines      255647   255654       +7     
==========================================
+ Hits       173034   173075      +41     
+ Misses      82613    82579      -34     

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

[harshit-gangal]

How about we use transaction_mode for this? Currently it can be single, multi, twopc.

[demmer]

Great question -- I considered that as well and if that's what we all think would be a better way to do this, I could try it out and see how it goes. That said, I opted for a separate option for a couple reasons: 1. Conceptually, a new "MULTI_AUTOCOMMIT_ALLOWED" transaction mode would be basically the same as MULTI just with this one tweak. So, the change seemed more like the `--no-scatter` option which was added a while back in that it affects a particular behavior of the vtgate rather than defining a whole new "mode". 2. Pragmatically, doing a new transaction mode would be a more involved patch to plumb down into all the right places and would involve a lot more boilerplate changes to end up in the 2-3 places that we actually have to change the executor.