Metrics Data Infrastructure Investigation - Accessibility Issue

[github-actions]

Problem

Meta-orchestrator workflows (agent-performance-analyzer, campaign-manager, workflow-health-manager) cannot access metrics data that should be stored in /tmp/gh-aw/repo-memory/default/metrics/.

The metrics-collector workflow is designed to store daily performance data, but this data is not accessible during meta-orchestrator runs, blocking data-driven decision making.

Expected Behavior

According to workflow specifications:

1. metrics-collector workflow runs daily
2. Stores metrics in /tmp/gh-aw/repo-memory/default/metrics/:
   • latest.json - Most recent snapshot
   • daily/YYYY-MM-DD.json - Historical daily metrics (30 days)
3. Meta-orchestrators read this data for analysis
4. Data persists via memory/meta-orchestrators git branch

Actual Behavior

During agent-performance-analyzer run #20687933156:

$ ls -la /tmp/gh-aw/repo-memory/default/metrics/
Permission denied and could not request permission from user

Files found in repo-memory:

• ✅ shared-alerts.md
• ✅ workflow-health-latest.md
• ❌ metrics/ directory (not accessible or not present)

Impact

Blocks critical functionality:

• ❌ Cannot perform historical trend analysis
• ❌ Cannot compare current vs past performance
• ❌ Cannot calculate 7-day and 30-day averages
• ❌ Cannot identify performance anomalies
• ❌ Cannot make data-driven recommendations

Affected workflows:

1. agent-performance-analyzer (this workflow)
2. campaign-manager (needs workflow metrics)
3. workflow-health-manager (needs execution stats)

Investigation Steps

1. Check metrics-collector Workflow

# Get recent metrics-collector runs
gh workflow view metrics-collector.md --repo githubnext/gh-aw
gh run list --workflow=metrics-collector.md --limit 5

# Check last run status and logs
gh run view (run-id) --log

Questions:

• Is metrics-collector running successfully?
• Are there any errors during data writing?
• Is the workflow using correct repo-memory configuration?

2. Verify Repo Memory Configuration

Check .github/workflows/metrics-collector.md:

tools:
  repo-memory:
    branch-name: memory/meta-orchestrators
    file-glob: "**"

Questions:

• Is branch-name correct and accessible?
• Does the branch exist in the repository?
• Are file permissions correct?

3. Check Branch and File Permissions

# List branches
git branch -r | grep memory

# Check branch contents
git ls-tree -r memory/meta-orchestrators

# Look for metrics files
git show memory/meta-orchestrators:metrics/latest.json

4. Test Repo Memory Access

Create a minimal test workflow:

---
description: Test repo-memory access
on: workflow_dispatch
tools:
  repo-memory:
    branch-name: memory/meta-orchestrators
    file-glob: "**"
---

1. Try to read: `/tmp/gh-aw/repo-memory/default/metrics/latest.json`
2. If missing, write a test file: `/tmp/gh-aw/repo-memory/default/test.txt`
3. Check if file persists to git branch

Possible Root Causes

1. Metrics collector not running: Workflow may be disabled or failing
2. Wrong branch name: memory/meta-orchestrators vs different branch
3. File path mismatch: Writing to different location than expected
4. Permission issue: Workflow lacks write access to branch
5. Timing issue: Metrics written after orchestrators run (need to check schedule)
6. Directory structure: metrics/ subdirectory not created properly

Success Criteria

• Identify root cause of metrics data inaccessibility
• Fix metrics-collector workflow if needed
• Verify data writes to correct branch and path
• Confirm meta-orchestrators can read metrics data
• Document proper repo-memory usage patterns
• Add monitoring to detect future access issues

Temporary Workaround

Until fixed, meta-orchestrators will:

• Rely on direct GitHub API queries where possible
• Use configuration-based analysis
• Document limitations in reports
• Skip historical trend analysis

Priority

HIGH - Blocks data-driven decision making for entire meta-orchestrator ecosystem.

Next Actions

1. Investigate metrics-collector recent runs
2. Check memory/meta-orchestrators branch status
3. Test repo-memory read/write with simple workflow
4. Fix identified issues
5. Verify meta-orchestrators can access data

---

Source: Agent Performance Analyzer (Run #20687933156)
Related: Meta-orchestrator coordination, data-driven analysis
Blocks: Historical trend analysis, performance benchmarking

AI generated by Agent Performance Analyzer - Meta-Orchestrator