[Bugfix] Broken Heartbeat system in Row Streamer

[siddharth16396]

This PR implements the fix for the bug mentioned #18391

TL;DR
VStreamRows uses a broken heartbeat mechanism that only sends one heartbeat and then exits. In contrast, VStream (used in normal binlog replication) sends heartbeats continuously. This PR fixes the issue by introducing a proper loop in the rowstreamer.go heartbeat goroutine to ensure heartbeats are sent throughout the copy phase.

We found this out because Some of our long-running schema changes failed due to missing heartbeats during the copy phase of VReplication.

Description

This is already patched in our internal fork of vitess and has been tested to work correctly for all the failing cases we saw.

We've identified a critical architectural inconsistency in Vitess VReplication that explains this seemingly contradictory behavior. The reason some long-running schema changes don't get heartbeats while others do is because different phases of the operation use different heartbeat mechanisms.

The Two Different Heartbeat Systems

1. Working Heartbeat System (vstreamer.go)

Used by normal VReplication (binlog streaming phase):

// vstreamer.go lines 297-360
hbTimer := time.NewTimer(HeartbeatTime) // 900ms
defer hbTimer.Stop()

for {
    hbTimer.Reset(HeartbeatTime)  // ← Proper timer reset in loop
    select {
    case ev, ok := <-throttledEvents:
        // Process events
    case <-hbTimer.C:
        // Send heartbeat every 900ms
        if err := injectHeartbeat(false, ""); err != nil {
            return err
        }
    }
}

This system works correctly - it continuously sends heartbeats every 900ms.

2. Broken Heartbeat System (rowstreamer.go)

Used by table copying phase (VStreamRows):

// rowstreamer.go lines 372-385
go func() {
    select {  // ← NO FOR LOOP! Only executes once!
    case <-rs.ctx.Done():
        return
    case <-heartbeatTicker.C:
        safeSend(&binlogdatapb.VStreamRowsResponse{Heartbeat: true})
    }  // ← Goroutine exits after first heartbeat
}()

This system is broken - it only sends one heartbeat and then the goroutine dies.

Which Operations Use Which System

Operations That DO Get Heartbeats:

• Normal VReplication workflows (MoveTables, Reshard after copy phase)
• Catchup phase during onlineDDL
• Binlog streaming operations
• Short onlineDDL operations (complete within 10 seconds)

These use vplayer.go → VStream() → vstreamer.go (working heartbeats)

Operations That DON'T Get Heartbeats:

• OnlineDDL copy phase on large tables
• Long-running VStreamRows operations
• Table copying during MoveTables/Reshard

These use vcopier.go → VStreamRows() → rowstreamer.go (broken heartbeats)

The Code Flow

Copy Phase (Broken Heartbeats):

vcopier.copyTable() 
  → sourceVStreamer.VStreamRows() 
  → tabletConnector.VStreamRows() 
  → gRPC call to tablet 
  → tabletserver.VStreamRows() 
  → vstreamer.StreamRows() 
  → rowstreamer.streamQuery() 
  → BROKEN heartbeat goroutine

Normal Replication (Working Heartbeats):

vplayer.play() 
  → sourceVStreamer.VStream() 
  → tabletConnector.VStream() 
  → gRPC call to tablet 
  → tabletserver.VStream() 
  → vstreamer.Stream() 
  → WORKING heartbeat timer

Why You See This Pattern

1. T=0: OnlineDDL starts, enters copy phase using VStreamRows
2. T=10s: First (and only) heartbeat sent from broken rowstreamer
3. T=10s+: Heartbeat goroutine exits, no more heartbeats sent
4. T=180min: Operation fails due to "no heartbeat activity"

The Fix

The immediate fix is to add a for loop to the rowstreamer heartbeat goroutine:

// In rowstreamer.go lines 372-385
go func() {
    for {  // ← Add this loop!
        select {
        case <-rs.ctx.Done():
            return
        case <-heartbeatTicker.C:
            safeSend(&binlogdatapb.VStreamRowsResponse{Heartbeat: true})
        }
    }
}()

This explains the inconsistent behavior we observed - it's not random, it's deterministic based on which code path the operation takes. The bug only affects the table copying component, not the binlog streaming component.

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

[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.

[arthurschreiber]

Nice, and good catch! Is this an issue in older versions of vitess as well? Do you think there's a test case you could write that fails before / passes after?

[siddharth16396]

@arthurschreiber : Yes this seems to be an issue on atleast v20,21,22 (so i'm guessing its on all)

WRT the test cases, i'll try to take a stab at it, since it was just a very minor change i didn't bother diving into it.

[timvaillancourt]

@siddharth16396 thanks for the fix. Is there a way we can ensure a stream gets continuous heartbeats in a unit test? I'm hoping there is an existing test we can add more validation to

[siddharth16396]

@timvaillancourt @arthurschreiber : I've added a UT which should work.

[timvaillancourt]

@siddharth16396 thanks for the new test. LGTM

[rohit-nayak-ps]

Nice catch!