Fix race in replicationLagModule of go/vt/throttle

[timvaillancourt]

Description

This PR addresses a race discovered in replicationLagModule of go/vt/throttle. The TL;DR is there are concurrent update/delete()s and concurrent reads/iterations of a map without locking

This race is causing vttablets with --enable-tx-throttler to crash after 40-100 minutes of running Vitess v15 w/txthrottler backports from v16-v19:

fatal error: concurrent map iteration and map write
goroutine 455 [running]:
vitess.io/vitess/go/vt/throttler.(*Throttler).MaxLag(0xc000bb4728?, 0xb94630?)
        vitess.io/vitess/go/vt/throttler/throttler.go:236 +0x9c
vitess.io/vitess/go/vt/vttablet/tabletserver/txthrottler.(*txThrottlerStateImpl).updateMaxLag(0xc0009f85a0)
        vitess.io/vitess/go/vt/vttablet/tabletserver/txthrottler/tx_throttler.go:399 +0x1b4
created by vitess.io/vitess/go/vt/vttablet/tabletserver/txthrottler.newTxThrottlerState in goroutine 99
        vitess.io/vitess/go/vt/vttablet/tabletserver/txthrottler/tx_throttler.go:321 +0x4f6

This output includes the TestThrottlerMaxLag test I added in my first commit to reproduce the race: b83f057

Related Issue(s)

Resolves #16102

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 88.46154% with 3 lines in your changes missing coverage. Please review.

Project coverage is 69.42%. Comparing base (978c59d) to head (1da308b).
Report is 11 commits behind head on main.

Files with missing lines	Patch %	Lines
go/vt/throttler/replication_lag_cache.go	91.30%	2 Missing ⚠️
go/vt/throttler/throttler.go	66.66%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #16078      +/-   ##
==========================================
- Coverage   69.43%   69.42%   -0.01%     
==========================================
  Files        1571     1571              
  Lines      203195   203321     +126     
==========================================
+ Hits       141079   141157      +78     
- Misses      62116    62164      +48     

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

[timvaillancourt]

Moved this mostly-unchanged from throttler.go (.MaxLag(...)) instead of holding the mu sync.Mutex lock from a different struct

.MaxLag() now calls this func

[timvaillancourt]

Moved into replicationLagCache

[timvaillancourt]

I think I figured out why we only see this on v15 + v16-v19 txthrottler backports:

This txthrottler regression is increasing the rate of healthcheck streams, thus .add() calls to replicationLagCache

[timvaillancourt]

Squashed commits to make this cherry-pick-able

[mattlord]

Squashed commits to make this cherry-pick-able

FWIW, we do squash and merge by default so you can always cherry-pick the squashed commit that is made to the upstream branch.

[timvaillancourt]

Squashed commits to make this cherry-pick-able

FWIW, we do squash and merge by default so you can always cherry-pick the squashed commit that is made to the upstream branch.

Right 👍. Here my goal is to cherry-pick this to slackhq pre-review/merge, so a squash makes that just 1 commit

[github-actions]

This PR is being marked as stale because it has been open for 30 days with no activity. To rectify, you may do any of the following:

• Push additional commits to the associated branch.
• Remove the stale label.
• Add a comment indicating why it is not stale.

If no action is taken within 7 days, this PR will be closed.

[timvaillancourt]

@timvaillancourt have you been able to confirm that the performance is acceptable in your environment?

@mattlord yes, it turns out the mutex is only used:

1. Every --healthcheck_interval / 2 (for our production 500ms w/--healthcheck_interval 1s) to calculate + cache the max lag value
2. Every --healthcheck_interval multiplied by number of non-PRIMARY tablets in the shard

In our environment this means the mutex is held about 5-7~ times per second

Sorry for the late review. What really concerns me is whether the mutex gets used on the data path. As long as we only use it in ProcessRecords() or other background goroutines, I think that's fine. But if a query/transaction needs to pass through the mutex, I'd go as far as saying this is a no-go.

@shlomi-noach agreed, I think we're lucky the max lag value is cached 👍

As for the use of atomic.Pointer, I gave this a shot but ran into a few concerns. Because the maps are read/written concurrently in several methods where the map-var falls out of scope, this means a .Load() is required before a .Set() or .Swap() in several places. I'm new to atomic.Pointer, but to me there is a risk of data inconsistency because there is a "gap" between the load and write in more than one place. I think if the map was always in scope to all methods that read it (don't need a .Load()) this wouldn't be a problem however, but this is all new to me again

I also tried sync.Map, which avoided that concern but I encountered another problem: no support for calling len() on sync.Map. So then I stored the len of the map in a separate atomic value and started to wonder if it's really worth it if the mutex isn't under that much pressure. Any thoughts/ideas appreciated on this!

For now I'll address the other PR comments

[timvaillancourt]

Adding backport tags as discussed offline with @deepthi as this addresses a crash

[mattlord]

LGTM! I only had some minor nits. Thanks, @timvaillancourt