[EXPERIMENTAL] Tiered CI Testing - Improve Reliability & Speed

[e-gineer]

🎯 Purpose

This PR introduces experimental tiered CI workflows to address current CI reliability issues. This is a SAFE EXPERIMENT that doesn't affect production CI.

⚠️ Problem Statement

Current CI has critical issues:

• 50% failure rate (10 failures in last 20 runs)
• 7-10 minute feedback time on every commit
• Heavy infrastructure tests in unit test suite causing random failures
• No test categorization - all tests treated equally
• Lost productivity: ~2.5 developer-hours/day debugging failures

💡 Solution: Tiered Testing

Tier 1: Quick Checks (< 2 minutes)

Workflow: 10-test-quick-EXPERIMENTAL.yaml

• Triggers: Every push to test/** branches
• What: Fast unit tests only (go test -short)
• Goal: Immediate feedback

Tier 2: Standard Suite (< 10 minutes)

Workflow: 11-test-standard-EXPERIMENTAL.yaml

• Triggers: Manual (workflow_dispatch)
• What: All unit tests + 4 core acceptance tests
• Goal: Comprehensive validation

Tier 3: Full Suite (< 20 minutes)

Workflow: 11-test-full-EXPERIMENTAL.yaml

• Triggers: Manual (workflow_dispatch)
• What: All 21 acceptance tests
• Goal: Complete coverage

🔒 Safety

These workflows are EXPERIMENTAL and SAFE:

• ✅ Only trigger on test/** branches
• ✅ Don't interfere with production CI
• ✅ Can be tested in isolation
• ✅ Easy to rollback (just don't merge)

📊 Expected Outcomes

Metric	Current	Target	Improvement
Success Rate	50%	95%+	2x better
Time to Feedback	7-10 min	< 2 min	5x faster
Wasted CI Time	100 min/day	< 10 min/day	90% reduction
Dev Hours Lost	2.5/day	0.3/day	8x better

🧪 Testing Plan

1. This PR tests Quick Checks

   • Quick tests will run automatically
   • Should complete in < 2 minutes
   • We'll push several commits to validate

2. Manual Standard Test

   • Trigger via Actions tab: "Run workflow"
   • Should complete in < 10 minutes
   • Core tests should catch 90%+ of issues

3. Manual Full Test

   • Trigger via Actions tab: "Run workflow"
   • Should complete in < 20 minutes
   • All 21 acceptance tests

4. Metrics Collection

   • Run 5-10 iterations
   • Track: duration, success rate, which tier caught bugs
   • Compare with production CI

📝 Files Changed

New Workflows

• .github/workflows/10-test-quick-EXPERIMENTAL.yaml - Quick tests
• .github/workflows/11-test-standard-EXPERIMENTAL.yaml - Standard suite
• .github/workflows/11-test-full-EXPERIMENTAL.yaml - Full suite

Documentation

See .ai/wip/ci-optimization/ for:

• Comprehensive analysis
• Implementation plan
• Success metrics

🎬 Next Steps

If this experiment succeeds:

1. Gather feedback from team
2. Create production branch (remove EXPERIMENTAL)
3. Modify existing workflows with conditional triggers
4. Add documentation
5. Merge to develop

If it doesn't work:

1. Don't merge this PR
2. No impact to production
3. Learn and iterate

🔍 How to Test

Test Quick Workflow (Automatic)

Just push commits to this PR - quick tests run automatically!

Test Standard Workflow (Manual)

1. Go to Actions tab
2. Select "11 - Test: Standard Suite (EXPERIMENTAL)"
3. Click "Run workflow"
4. Select branch: test/ci-tiered-testing
5. Click green "Run workflow" button

Test Full Workflow (Manual)

1. Go to Actions tab
2. Select "11 - Test: Full Suite (EXPERIMENTAL)"
3. Click "Run workflow"
4. Select branch: test/ci-tiered-testing
5. Click green "Run workflow" button

📚 Background

• Original analysis: Issue #XXXX (to be created)
• Current CI failure rate: 50% (10/20 recent runs)
• Most common failure: Infrastructure tests in unit suite
• Recent fix: Removed TestEnsureDBInstalled_Concurrent (Add comprehensive passing tests from bug hunting initiative #4864)

✅ Success Criteria

• Quick tests consistently complete in < 2 minutes
• Quick tests have 95%+ success rate
• Standard tests catch 90%+ of bugs
• Full tests maintain 100% coverage
• At least 5 successful test runs
• Team approves approach

🤝 Feedback Welcome!

This is an experiment to improve our CI reliability and developer experience. Please:

• Try running the workflows
• Report any issues or suggestions
• Share your experience with feedback times

---

🤖 Generated with Claude Code

Co-Authored-By: Claude noreply@anthropic.com

[e-gineer]

🔧 Fix Applied: Proper Tiered Triggers

Just pushed a fix - the original commit had overly conservative triggers (only test/** branches). Updated to implement the proper tiered strategy:

✅ Corrected Trigger Strategy

Tier 1: Quick Checks - 10-test-quick-EXPERIMENTAL.yaml

Triggers:

• ✅ Every pull_request push (all PRs, not just test branches)
• ✅ Push to develop and main branches
• ✅ Manual via workflow_dispatch

What it does:

• Runs go test -short (< 30s timeout)
• Skips expensive tests
• Fast feedback: < 2 minutes

Tier 2: Standard Suite - 11-test-standard-EXPERIMENTAL.yaml

Triggers:

• ✅ pull_request type ready_for_review (when you mark PR ready)
• ✅ Push to develop and main branches
• ✅ Manual via workflow_dispatch

What it does:

• All unit tests (no -short flag)
• 4 core acceptance tests (installation, plugin, connection_config, service)
• Comprehensive: < 10 minutes

Tier 3: Full Suite - 11-test-full-EXPERIMENTAL.yaml

Triggers:

• ✅ Push to develop (after merge)
• ✅ Nightly at 2 AM (cron: '0 2 * * *')
• ✅ pull_request with label test:full
• ✅ Manual via workflow_dispatch

What it does:

• All unit tests
• All 21 acceptance tests
• Complete coverage: < 20 minutes

🎯 The Progressive Strategy

PR Draft → Quick tests (2 min) → Fast feedback
    ↓
PR Ready → Standard tests (10 min) → Pre-merge validation  
    ↓
Merge → Full tests (20 min) → Post-merge verification
    ↓
Nightly → Full tests → Regression detection

This matches the original goal: test actual development code with increasing thoroughness as code progresses toward release.

The -EXPERIMENTAL suffix is just so these workflows can coexist with the current production workflow while we validate the approach. Once proven, we'll:

1. Remove -EXPERIMENTAL suffix
2. Modify the existing 11-test-acceptance.yaml to be less aggressive
3. Enjoy faster, more reliable CI! 🎉

[github-actions]

This PR is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 30 days.