---
description: Test event-driven systems by replaying production events
triggers:
  - manual
  - test:replay
agents:
  - tester
  - data-engineer
---

# Replay Testing Workflow

Test system behavior by replaying historical events.

## Prerequisites
- [ ] Event store with historical data
- [ ] Isolated test environment
- [ ] Anonymization requirements defined

## Phase 1: Event Preparation

### Step 1.1: Select Events for Replay
```yaml
agent: data-engineer
action: query
criteria:
  - Time range
  - Event types
  - Specific scenarios
  - Edge cases
sources:
  - Event store
  - Kafka topics (with retention)
  - Dead letter queues
```

### Step 1.2: Anonymize Events
```yaml
agent: data-engineer
action: transform
anonymization:
  - PII fields: hash or fake
  - Financial data: obfuscate
  - IDs: maintain referential integrity
  - Timestamps: optional offset
```

### Step 1.3: Create Event Dataset
```yaml
agent: data-engineer
action: export
format:
  - JSON Lines
  - Avro files
  - Parquet
outputs:
  - event_dataset
  - metadata (counts, types, time range)
```

## Phase 2: Environment Setup

### Step 2.1: Provision Test Environment
```yaml
agent: tester
action: setup
components:
  - Service under test
  - Dependencies (mocked or real)
  - Database (clean state)
  - Message broker
isolation:
  - Network isolation
  - No production access
  - Clean state between runs
```

### Step 2.2: Configure Replay Infrastructure
```yaml
agent: tester
action: configure
replay_tools:
  - Kafka replay consumer
  - Custom replay script
  - Event sourcing replay
settings:
  - Replay speed (1x, 10x, max)
  - Ordering guarantees
  - Pause/resume capability
```

## Phase 3: Replay Execution

### Step 3.1: Baseline Replay
```yaml
agent: tester
action: replay
mode: baseline
purpose: Establish expected behavior
capture:
  - Final state
  - Processing metrics
  - Error counts
```

### Step 3.2: Modified Replay
```yaml
agent: tester
action: replay
mode: comparison
with_changes:
  - New code version
  - Modified configuration
  - Different event order
capture:
  - State differences
  - Performance differences
  - New errors
```

## Phase 4: Analysis

### Step 4.1: State Comparison
```yaml
agent: tester
action: compare
aspects:
  - Database state diff
  - Emitted events diff
  - Side effects diff
  - Error handling diff
tools:
  - Database diff tools
  - Event comparison scripts
  - State snapshot comparison
```

### Step 4.2: Performance Analysis
```yaml
agent: tester
action: analyze
metrics:
  - Processing throughput
  - Latency distribution
  - Resource usage
  - Error rates
```

### Step 4.3: Issue Identification
```yaml
agent: tester
action: identify
issues:
  - State inconsistencies
  - Missing events
  - Processing errors
  - Performance regressions
```

## Phase 5: Reporting

### Step 5.1: Generate Report
```yaml
agent: tester
action: report
contents:
  - Replay summary
  - State comparison results
  - Performance metrics
  - Identified issues
  - Recommendations
```

## Outputs
- [ ] Anonymized event dataset
- [ ] Replay configuration
- [ ] Baseline results
- [ ] Comparison report
- [ ] Issue list

## Quality Gates
- Events properly anonymized
- Replay environment isolated
- Baseline established
- Differences explained
- No data leakage