Error Handling

Build resilient automations that handle failures gracefully

Automations run unattended, so they need to handle errors gracefully. Here’s how to build resilient workflows.

Retry Logic

API calls can fail temporarily. Build in retries for transient failures.

Try up to 3 times:
  - Fetch data from API
  - If successful, proceed
  - If rate limited, wait 60 seconds and retry
  - If all attempts fail, log error and notify team

Fallback Behavior

When the primary approach fails, have a backup plan.

Try to post to #team-updates
If channel doesn't exist:
  - Post to #general instead
  - Notify admin about missing channel

Partial Success

Don’t let one failure stop the whole automation. Process items independently.

For each PR in the list:
  - Try to process the PR
  - If processing fails:
    - Log the error with PR number
    - Continue with remaining PRs
  - Track success/failure counts

At the end:
  - Report: "Processed 8/10 PRs successfully"
  - List failed items for manual review

Dry Run Mode

Add a flag to test automations without making real changes.

If DRY_RUN mode:
  - Log what would be done
  - Show the Slack message that would be posted
  - List the PRs that would be created
  - Don't make actual changes
Otherwise:
  - Execute normally

Testing Strategies

Use Test Markers

Isolate test runs from production:

Only process items with label "automation-test"
This keeps testing separate from real data

Incremental Testing

Build up complexity gradually:

Phase 1: Just fetch and log data
Phase 2: Add filtering logic
Phase 3: Add action logic (comments, labels)
Phase 4: Enable posting to real channels

Test Channels

Use dedicated test channels before going live:

Development: Post to #automation-test
Staging: Post to #team-updates-test
Production: Post to #team-updates

Debugging Tips

• Log decision points: Record counts at each step (“Found 15 PRs → After filtering: 8 PRs → Posted successfully”)
• Include timestamps: Correlate logs with events for easier debugging
• Save raw responses: In DEBUG mode, preserve API responses and status codes

Security Considerations

Never Log Secrets

Don’t include tokens or sensitive data in logs:

Log: "Using API key ending in ...x7k9"  ✓
Log: "Using API key sk-abc123..."       ✗

Validate Input

Check data before using it:

Before posting message:
- Check message doesn't contain API keys or tokens
- Validate URLs are from trusted domains
- Ensure content length is reasonable

Limit Scope

Request only the permissions your automation needs. Don’t use admin tokens for read-only operations.