Rebase yolo-rebase-22.03 onto v22.0.3

[sbaker617]

What's this?

This PR rebases the yolo-rebase-22.03 branch onto upstream v22.0.3 (from v22.0.2), achieving a clean linear history by removing the merge commit and replaying all Slack-specific commits on top of the latest upstream release.

Summary

• Base: v22.0.3 (commit 53de5fdf41)
• Commits rebased: 30 Slack-specific commits
  • 23 commits from before the v22.0.2 merge (CI setup, VTOrc improvements, query logging, durability policy, etc.)
  • 7 commits from after the merge (PRs [slack-22.0] set the previous release to slack-19.0 #774, onlineddl v22 backport #782, [slack-22.0]: forward port: consul conn override flags (Slack) #783, [slack-22.0] V22 cdc backports round two #784, [slack-22.0] Improve PlannedReparentShard timeout error messages #786, [slack-22.0] forward-fit: Allow setting grpc send and recv message sizes independe… #787, [slack-22.0] Forward port structured logging #788)
• History: Linear (no merge commits)
• Build status: ✅ Successful

Rebase Strategy

The rebase used git rebase -i ef39c169b7 --onto release-22.0 to:

1. Start from v22.0.2 base (ef39c16)
2. Replay all Slack commits onto v22.0.3
3. Remove the merge commit (79f97fb) automatically
4. Result in clean linear history: v22.0.3 → 30 Slack commits

Conflicts Encountered & Resolution

CI Workflow Conflicts (Multiple commits)

Issue: Extensive conflicts in .github/workflows/ files due to Slack-specific CI configuration vs upstream changes.

Resolution: Systematically accepted Slack's workflow versions using git checkout --theirs for all workflow files. Handled file deletions (MySQL 5.7 workflows, vdiff2 workflow) appropriately.

smartconnpool Race Condition Fix (Commit 11: #727)

Issue: Conflict between Slack's channel-based waitlist implementation and upstream's semaphore-based approach.

Resolution:

• Kept Slack's channel-based implementation in waitlist.go
• Removed upstream's semaphore files (sema_race.go, sema.s, sema_norace.go)
• This aligns with Slack's architectural choice for connection pool management

Version/Build Configuration Conflicts (Commit 22: v22.0.2 merge)

Issue: Version numbers, Go version, bootstrap version conflicts.

Resolution: Accepted upstream versions:

• Go version: 1.24.12 (from 1.24.4)
• Bootstrap version: 43.6 (from 43.2)
• Vitess version: 22.0.3

Dependency Conflicts (Commits 7, 26)

Issue: go.mod/go.sum conflicts with dependency version mismatches.

Resolution: Accepted upstream dependency versions as they're more recent and tested with v22.0.3.

Post-Rebase Fixes (Additional commit 78a025d)

After the rebase completed, several build issues required fixes:

1. Missing Dependencies

Problem: Build failed with missing imports for cgroup metrics:

go/vt/servenv/metrics_host.go:22:2: no required module provides package github.com/shirou/gopsutil/v4/cpu
go/vt/servenv/metrics_host.go:23:2: no required module provides package github.com/shirou/gopsutil/v4/mem

Fix: Added missing dependencies:

• github.com/shirou/gopsutil/v4 for CPU/memory metrics
• github.com/containerd/cgroups/v3 for cgroup support

2. Proto File Mismatch

Problem: Generated proto files had field mismatches causing build errors.

Fix: Ran make proto to regenerate all proto files from source, ensuring consistency with v22.0.3 proto definitions.

3. SlackCrossCell Durability Policy Interface

Problem: Upstream added HasSemiSync() method to Durabler interface, but vitess-addons package's SlackCrossCell type doesn't implement it yet:

cannot use &durability.SlackCrossCell{} as Durabler value: missing method HasSemiSync

Fix: Created slackCrossCellWrapper type in slack_cross_cell_shim.go that:

• Wraps the vitess-addons SlackCrossCell type
• Delegates all existing methods to the wrapped type
• Implements HasSemiSync() returning true (SlackCrossCell uses semi-sync)
• This is a temporary shim until vitess-addons is updated

Slack Features Verified

All critical Slack-specific functionality has been preserved:

✅ SmartConnPool: Channel-based waitlist implementation (not semaphore)
✅ Structured Logging: --structured-logging and --structured-log-level flags present
✅ gRPC Message Sizes: Independent --grpc_max_message_send_size and --grpc_max_message_recv_size flags
✅ VStream Enhancements: Transaction chunking, liveness monitoring, OOM prevention
✅ Query Logging: --querylog-time-threshold and --querylog-emit-on-any-condition-met flags
✅ VTOrc Improvements: Recovery controls, debug endpoints, health checks
✅ Durability Policy: slack_cross_cell policy registered
✅ v19.0 Compatibility: tools/get_previous_release.sh points to slack-19.0
✅ CDC/VReplication: Cgroup metrics, VStream improvements, TablesToCopy usage

Testing Recommendations

CI Verification

• All unit tests pass
• End-to-end tests pass (especially VStream, VReplication, VTOrc)
• Upgrade/downgrade tests with v19.0 compatibility
• SmartConnPool tests under load
• Query logging with custom flags

Manual Verification

• VStream transaction chunking prevents OOM under load
• Structured logging output format is correct
• Slack durability policy (slack_cross_cell) works for reparenting
• gRPC message size limits work independently
• Query logging thresholds and conditions work as expected

Notes

• Build verified locally on macOS (arm64)
• All 30 Slack commits preserved in order
• No functional changes to Slack-specific code besides compatibility fixes
• The slackCrossCellWrapper should be removed once vitess-addons is updated to implement HasSemiSync()

---

Rebased by Claude Code

[salesforce-cla]

Thanks for the contribution! Before we can merge this, we need @frouioui @vitess-bot @systay @harshit-gangal @dbussink to sign the Salesforce Inc. Contributor License Agreement.

[salesforce-cla]

Thanks for the contribution! Before we can merge this, we need @vitess-bot @GuptaManan100 to sign the Salesforce Inc. Contributor License Agreement.