|
| 1 | +# Datadog Error Debugging Workflow |
| 2 | + |
| 3 | +This example demonstrates how to use OpenHands agents to automatically debug errors from Datadog in a GitHub Actions workflow. |
| 4 | + |
| 5 | +## Overview |
| 6 | + |
| 7 | +The workflow: |
| 8 | +1. Fetches errors from Datadog based on configurable queries |
| 9 | +2. Searches for or creates GitHub issues to track errors |
| 10 | +3. Clones relevant repositories for comprehensive analysis |
| 11 | +4. Uses OpenHands AI agents to analyze code and identify root causes |
| 12 | +5. Posts debugging insights as comments on GitHub issues |
| 13 | + |
| 14 | +## Files |
| 15 | + |
| 16 | +- `workflow.yml` - GitHub Actions workflow with manual trigger |
| 17 | +- `datadog_debugging.py` - Main debugging script |
| 18 | +- `debug_prompt.jinja` - Template for AI debugging prompts |
| 19 | + |
| 20 | +## Features |
| 21 | + |
| 22 | +### Manual Trigger |
| 23 | +Run on-demand via GitHub Actions UI with configurable inputs: |
| 24 | +- **Query Type**: Choose between `log-query` (search) or `log-error-id` (specific error ID) |
| 25 | +- **Datadog Query**: |
| 26 | + - For `log-query`: Search query like `service:deploy ClientDisconnect` |
| 27 | + - For `log-error-id`: Specific error tracking ID like `2adba034-ab5a-11f0-b04e-da7ad0900000` |
| 28 | +- Repository list to analyze |
| 29 | +- Issue repository for tracking |
| 30 | +- Parent issue for organization |
| 31 | +- LLM model selection |
| 32 | + |
| 33 | +### Smart Issue Management |
| 34 | +- Searches for existing issues before creating duplicates |
| 35 | +- Uses URL encoding for proper GitHub API queries |
| 36 | +- Selects oldest matching issue when duplicates exist |
| 37 | +- Links to parent tracking issue |
| 38 | + |
| 39 | +### Multi-Repository Analysis |
| 40 | +- Clone multiple repositories for comprehensive context |
| 41 | +- Agent has full view of all relevant codebases |
| 42 | +- Identifies root causes across repository boundaries |
| 43 | + |
| 44 | +### AI-Powered Debugging |
| 45 | +- Automatic code analysis using OpenHands agents |
| 46 | +- Identifies error locations and root causes |
| 47 | +- Provides actionable fix recommendations |
| 48 | +- Posts detailed findings as GitHub comments |
| 49 | + |
| 50 | +## Setup |
| 51 | + |
| 52 | +### Required Secrets |
| 53 | + |
| 54 | +Configure these in your repository Settings → Secrets and variables → Actions: |
| 55 | + |
| 56 | +```yaml |
| 57 | +DD_API_KEY: Your Datadog API key |
| 58 | +DD_APP_KEY: Your Datadog Application key |
| 59 | +DD_SITE: Your Datadog site (e.g., us5.datadoghq.com) |
| 60 | +LLM_API_KEY: API key for LLM service |
| 61 | +LLM_BASE_URL: Base URL for LLM service (optional) |
| 62 | +``` |
| 63 | +
|
| 64 | +**Note**: `GITHUB_TOKEN` is automatically provided by GitHub Actions. |
| 65 | + |
| 66 | +### Installation |
| 67 | + |
| 68 | +1. Copy `workflow.yml` to your repository's `.github/workflows/` directory (e.g., `.github/workflows/datadog-debugging.yml`) |
| 69 | +2. Configure the required secrets in repository Settings → Secrets and variables → Actions |
| 70 | +3. Optionally, customize the workflow inputs and defaults in the YAML file |
| 71 | + |
| 72 | +**Note**: The workflow automatically downloads the latest version of `datadog_debugging.py` and `debug_prompt.jinja` from the SDK repository at runtime. No need to copy these files to your repository unless you want to customize them. |
| 73 | + |
| 74 | +## Usage |
| 75 | + |
| 76 | +### Via GitHub Actions UI |
| 77 | + |
| 78 | +1. Go to the **Actions** tab in your repository |
| 79 | +2. Select **Datadog Error Debugging** workflow |
| 80 | +3. Click **Run workflow** |
| 81 | +4. Configure inputs: |
| 82 | + - **Query Type**: Choose `log-query` or `log-error-id` (default: `log-query`) |
| 83 | + - **Datadog Query**: |
| 84 | + - For `log-query`: Search query (default: `service:deploy ClientDisconnect`) |
| 85 | + - For `log-error-id`: Error tracking ID (e.g., `2adba034-ab5a-11f0-b04e-da7ad0900000`) |
| 86 | + - **Repository List**: Comma-separated repos to analyze (default: `OpenHands/OpenHands,All-Hands-AI/infra`) |
| 87 | + - **Issue Repository**: Where to create issues (default: `All-Hands-AI/infra`) |
| 88 | + - **Parent Issue**: Optional parent issue URL for tracking |
| 89 | + - **Issue Prefix**: Prefix for issue titles (default: `DataDog Error: `) |
| 90 | + - **LLM Model**: Model to use (default: `openhands/claude-sonnet-4-5-20250929`) |
| 91 | +5. Click **Run workflow** |
| 92 | + |
| 93 | +### Via GitHub CLI |
| 94 | + |
| 95 | +**Search for errors matching a query:** |
| 96 | +```bash |
| 97 | +gh workflow run datadog-debugging.yml \ |
| 98 | + -f query_type="log-query" \ |
| 99 | + -f datadog_query="service:deploy ClientDisconnect" \ |
| 100 | + -f repo_list="OpenHands/OpenHands,All-Hands-AI/infra" \ |
| 101 | + -f issue_repo="All-Hands-AI/infra" |
| 102 | +``` |
| 103 | + |
| 104 | +**Debug a specific error by ID:** |
| 105 | +```bash |
| 106 | +gh workflow run datadog-debugging.yml \ |
| 107 | + -f query_type="log-error-id" \ |
| 108 | + -f datadog_query="2adba034-ab5a-11f0-b04e-da7ad0900000" \ |
| 109 | + -f repo_list="OpenHands/OpenHands,All-Hands-AI/infra,All-Hands-AI/deploy" \ |
| 110 | + -f issue_repo="All-Hands-AI/infra" |
| 111 | +``` |
| 112 | + |
| 113 | +## Example |
| 114 | + |
| 115 | +### Input (Search Query) |
| 116 | +```yaml |
| 117 | +query_type: "log-query" |
| 118 | +datadog_query: "service:deploy ClientDisconnect" |
| 119 | +repo_list: "OpenHands/OpenHands,All-Hands-AI/infra,All-Hands-AI/deploy" |
| 120 | +issue_repo: "All-Hands-AI/infra" |
| 121 | +issue_parent: "https://github.com/All-Hands-AI/infra/issues/672" |
| 122 | +``` |
| 123 | + |
| 124 | +### Input (Specific Error ID) |
| 125 | +```yaml |
| 126 | +query_type: "log-error-id" |
| 127 | +datadog_query: "2adba034-ab5a-11f0-b04e-da7ad0900000" |
| 128 | +repo_list: "OpenHands/OpenHands,All-Hands-AI/infra,All-Hands-AI/deploy" |
| 129 | +issue_repo: "All-Hands-AI/infra" |
| 130 | +issue_parent: "https://github.com/All-Hands-AI/infra/issues/672" |
| 131 | +``` |
| 132 | + |
| 133 | +### Output |
| 134 | +- **Console**: Progress logs showing error fetching, repository cloning, and agent analysis |
| 135 | +- **GitHub Issue**: Created or updated with error details |
| 136 | +- **GitHub Comment**: AI-generated analysis with root cause and recommendations |
| 137 | +- **Artifacts**: Debugging data and logs saved for 7 days |
| 138 | + |
| 139 | +### Real Example |
| 140 | + |
| 141 | +See a real run with production data: |
| 142 | +- Error: `starlette.requests.ClientDisconnect` (1,526 occurrences) |
| 143 | +- Issue: https://github.com/All-Hands-AI/infra/issues/703 |
| 144 | +- AI Analysis: https://github.com/All-Hands-AI/infra/issues/703#issuecomment-3480707049 |
| 145 | + |
| 146 | +The agent identified: |
| 147 | +- Error locations in `github.py` and `gitlab.py` |
| 148 | +- Root cause: Unhandled `ClientDisconnect` exceptions |
| 149 | +- Recommendations: Add proper error handling for client disconnections |
| 150 | + |
| 151 | +## Configuration |
| 152 | + |
| 153 | +### Datadog Query Examples |
| 154 | + |
| 155 | +```yaml |
| 156 | +# ClientDisconnect errors |
| 157 | +service:deploy ClientDisconnect |
| 158 | +
|
| 159 | +# Server errors (5xx) |
| 160 | +service:deploy http.status_code:5* |
| 161 | +
|
| 162 | +# Database errors |
| 163 | +service:deploy (database OR postgresql) status:error |
| 164 | +
|
| 165 | +# Authentication errors |
| 166 | +service:deploy (authentication OR authorization) status:error |
| 167 | +
|
| 168 | +# Rate limit errors |
| 169 | +service:deploy rate_limit status:error |
| 170 | +``` |
| 171 | + |
| 172 | +### Repository List Format |
| 173 | + |
| 174 | +Comma-separated list of `owner/repo`: |
| 175 | +``` |
| 176 | +OpenHands/OpenHands,All-Hands-AI/infra,All-Hands-AI/deploy |
| 177 | +``` |
| 178 | + |
| 179 | +### LLM Model Options |
| 180 | + |
| 181 | +- `openhands/claude-sonnet-4-5-20250929` - Best quality (default) |
| 182 | +- `openhands/claude-haiku-4-5-20251001` - Faster, cheaper |
| 183 | +- `anthropic/claude-3-5-sonnet-20241022` - Alternative |
| 184 | + |
| 185 | +## Workflow Details |
| 186 | + |
| 187 | +### Inputs |
| 188 | + |
| 189 | +| Input | Type | Required | Default | Description | |
| 190 | +|-------|------|----------|---------|-------------| |
| 191 | +| `datadog_query` | string | Yes | `service:deploy ClientDisconnect` | Datadog query to search for errors | |
| 192 | +| `repo_list` | string | Yes | `OpenHands/OpenHands,All-Hands-AI/infra` | Comma-separated list of repositories | |
| 193 | +| `issue_repo` | string | Yes | `All-Hands-AI/infra` | Repository to create/update issues in | |
| 194 | +| `issue_parent` | string | No | - | Parent GitHub issue URL for tracking | |
| 195 | +| `issue_prefix` | string | No | `DataDog Error: ` | Prefix for issue titles | |
| 196 | +| `max_errors` | string | No | `5` | Maximum number of errors to fetch | |
| 197 | +| `llm_model` | string | No | `openhands/claude-sonnet-4-5-20250929` | LLM model to use | |
| 198 | + |
| 199 | +### Outputs |
| 200 | + |
| 201 | +- **GitHub Issues**: Created or updated with error details |
| 202 | +- **GitHub Comments**: AI analysis posted to issues |
| 203 | +- **Artifacts**: Debugging data and logs (retained for 7 days) |
| 204 | + |
| 205 | +### Permissions |
| 206 | + |
| 207 | +```yaml |
| 208 | +permissions: |
| 209 | + contents: read # Clone repositories |
| 210 | + issues: write # Create/update issues and comments |
| 211 | +``` |
| 212 | + |
| 213 | +## Customization |
| 214 | + |
| 215 | +### For Production Use |
| 216 | + |
| 217 | +Consider creating a separate configuration repository with: |
| 218 | +- Scheduled runs (daily for critical, weekly for comprehensive) |
| 219 | +- Predefined error query categories |
| 220 | +- Repository group definitions |
| 221 | +- Environment-specific settings |
| 222 | + |
| 223 | +See the All-Hands-AI/infra example for a production-ready implementation. |
| 224 | + |
| 225 | +### Adding Scheduled Runs |
| 226 | + |
| 227 | +Add to the workflow's `on:` section: |
| 228 | + |
| 229 | +```yaml |
| 230 | +on: |
| 231 | + workflow_dispatch: |
| 232 | + # ... existing inputs ... |
| 233 | + |
| 234 | + schedule: |
| 235 | + # Daily at 09:00 UTC for critical errors |
| 236 | + - cron: '0 9 * * *' |
| 237 | + # Weekly on Monday at 09:00 UTC for full scan |
| 238 | + - cron: '0 9 * * 1' |
| 239 | +``` |
| 240 | + |
| 241 | +### Matrix Strategy |
| 242 | + |
| 243 | +Run multiple queries in parallel: |
| 244 | + |
| 245 | +```yaml |
| 246 | +jobs: |
| 247 | + debug-errors: |
| 248 | + strategy: |
| 249 | + matrix: |
| 250 | + query: |
| 251 | + - "service:deploy ClientDisconnect" |
| 252 | + - "service:deploy http.status_code:5*" |
| 253 | + - "service:deploy database status:error" |
| 254 | + fail-fast: false |
| 255 | +``` |
| 256 | + |
| 257 | +## Troubleshooting |
| 258 | + |
| 259 | +### Workflow Fails to Start |
| 260 | +- Verify all required secrets are configured |
| 261 | +- Check `GITHUB_TOKEN` has necessary permissions |
| 262 | +- Review workflow syntax with `yamllint` |
| 263 | + |
| 264 | +### No Issues Created |
| 265 | +- Verify issue repository exists and is accessible |
| 266 | +- Check `GITHUB_TOKEN` has `issues: write` permission |
| 267 | +- Review workflow logs for API errors |
| 268 | + |
| 269 | +### Agent Analysis Incomplete |
| 270 | +- Increase workflow timeout if needed |
| 271 | +- Check `LLM_API_KEY` is valid and has quota |
| 272 | +- Try a different LLM model |
| 273 | +- Reduce number of repositories to analyze |
| 274 | + |
| 275 | +### Repository Clone Failures |
| 276 | +- Verify repository names use `owner/repo` format |
| 277 | +- Check `GITHUB_TOKEN` has access to private repos |
| 278 | +- Ensure repositories exist and are accessible |
| 279 | + |
| 280 | +## Related Examples |
| 281 | + |
| 282 | +- **Basic Action**: `examples/03_github_workflows/01_basic_action/` - Simple workflow example |
| 283 | +- **PR Review**: `examples/03_github_workflows/02_pr_review/` - PR automation example |
| 284 | +- **TODO Management**: `examples/03_github_workflows/03_todo_management/` - Automated TODO tracking |
| 285 | + |
| 286 | +## Benefits |
| 287 | + |
| 288 | +1. **Automated Debugging**: AI analyzes code without manual intervention |
| 289 | +2. **Reduced MTTR**: Faster root cause identification |
| 290 | +3. **Context-Aware**: Multi-repo analysis for complete picture |
| 291 | +4. **No Duplicates**: Smart issue tracking prevents clutter |
| 292 | +5. **Actionable Insights**: Clear recommendations for fixes |
| 293 | +6. **Scalable**: Easy to add new error categories |
| 294 | + |
| 295 | +## Learn More |
| 296 | + |
| 297 | +- [Datadog API Documentation](https://docs.datadoghq.com/api/) |
| 298 | +- [GitHub Actions Documentation](https://docs.github.com/en/actions) |
| 299 | +- [OpenHands SDK Documentation](https://github.com/OpenHands/software-agent-sdk) |
0 commit comments