SLO Designer

When this skill activates, you guide users through designing production-ready Service Level Objectives. Your role is to help identify critical user journeys, define measurable SLIs, set appropriate targets, and calculate error budgets.

Triggers

Activate when the user:

• "Design SLOs for my service"
• "Define reliability targets"
• "Calculate error budget"
• "Set up service level objectives"
• "Define SLIs for this system"
• "SLO design for API"
• "What should my availability target be?"

Core Concepts

Common SLI Categories

Availability

Percentage of successful requests.

availability_sli = (successful_requests / total_requests) * 100

Good for: APIs, web services, databases.

Latency

Response time percentiles (p50, p95, p99).

latency_sli = percentile(response_times, 99)

Good for: User-facing endpoints, real-time systems.

Throughput

Requests per second (RPS) or transactions.

throughput_sli = requests_per_second / expected_baseline

Good for: Batch processing, high-volume systems.

Error Rate

Percentage of 5xx responses.

error_rate_sli = (error_responses / total_responses) * 100

Good for: APIs, microservices.

Correctness

Percentage of correct results.

correctness_sli = (correct_results / total_results) * 100

Good for: Data pipelines, ML inference, calculations.

Process

1. DISCOVERY          Identify critical user journeys
        |             What matters to users?
        v
2. SLI DEFINITION     Select measurable indicators
        |             How do we measure success?
        v
3. SLO TARGETS        Set achievable targets
        |             What should we promise?
        v
4. ERROR BUDGET       Calculate allowed failures
        |             How much can we fail?
        v
5. ALERTING           Define burn rate alerts
        |             When do we intervene?
        v
6. DOCUMENTATION      Generate SLO document

Scripts

calculate_error_budget.py

Calculate error budget for a given SLO target:

python3 .claude/skills/slo-designer/scripts/calculate_error_budget.py \
  --target 99.9 \
  --period monthly

Arguments:

Exit Codes:

• 0: Success
• 1: Invalid arguments
• 2: Calculation error

generate_slo_document.py

Generate a complete SLO document from configuration:

python3 .claude/skills/slo-designer/scripts/generate_slo_document.py \
  --config path/to/slo-config.yaml \
  --output docs/slo-document.md

Question Framework

Use these questions to gather requirements:

1. Service Context

• What is the service name and purpose?
• Who are the primary users (internal, external, both)?
• What is the business criticality (revenue impact)?
• Are there existing SLAs or customer expectations?

2. User Journeys

• What are the 3-5 most critical user journeys?
• What actions must succeed for users to be satisfied?
• What is the acceptable response time for each journey?
• Which failures are most impactful?

3. Current State

• What metrics are already collected?
• What is the current availability (if known)?
• What are the current p50, p95, p99 latencies?
• What is the current error rate?

4. Infrastructure

• What is the deployment architecture?
• Are there external dependencies?
• What is the disaster recovery capability?
• What maintenance windows exist?

5. Targets

• What availability level is appropriate?
• What latency targets align with user expectations?
• How much error budget can the team responsibly manage?
• What alerting thresholds make sense?

SLO Target Guidelines

Choosing a target:

• Start conservative (lower targets are easier to meet)
• Consider external dependencies (chain reliability)
• Account for maintenance windows
• Leave margin from SLA (SLO should be tighter)

Error Budget Table

Burn Rate Alerting

Configure alerts based on budget consumption rate:

Multi-window alerting:

Alert if:
  burn_rate_1h > 14.4 AND burn_rate_6h > 6
  OR
  burn_rate_6h > 6 AND burn_rate_24h > 2

Output Template

Generate this structure:

# SLO Document: [Service Name]

## Service Overview
- **Name**: [Service name]
- **Owner**: [Team name]
- **Description**: [What the service does]
- **Business Criticality**: [Low/Medium/High/Critical]

## Critical User Journeys
1. [Journey 1]: [Description]
2. [Journey 2]: [Description]
3. [Journey 3]: [Description]

## Service Level Indicators

### SLI 1: Availability
- **Definition**: Percentage of successful HTTP requests
- **Measurement**: `sum(rate(http_requests_total{status!~"5.."}[5m])) / sum(rate(http_requests_total[5m]))`
- **Data Source**: Prometheus metrics

### SLI 2: Latency
- **Definition**: 99th percentile response time
- **Measurement**: `histogram_quantile(0.99, rate(http_request_duration_seconds_bucket[5m]))`
- **Data Source**: Prometheus metrics

## Service Level Objectives

| SLI | Target | Measurement Window | Rationale |
|-----|--------|-------------------|-----------|
| Availability | 99.9% | 30-day rolling | Industry standard for user-facing APIs |
| Latency (p99) | < 200ms | 30-day rolling | User research shows frustration above 200ms |

## Error Budgets

| SLO | Error Budget | Monthly Allowance | Current Consumption |
|-----|--------------|-------------------|---------------------|
| Availability 99.9% | 0.1% | 43 minutes | [Current] |
| Latency p99 < 200ms | 0.1% | 43 minutes | [Current] |

## Alerting Strategy

### Page-worthy Alerts (Critical)
- Burn rate > 14.4x for 1 hour AND > 6x for 6 hours
- Action: Immediate response required

### Ticket-worthy Alerts (Warning)
- Burn rate > 2x for 24 hours
- Action: Investigate within 1 business day

## Implementation Checklist
- [ ] Metrics collection configured
- [ ] SLO dashboard created
- [ ] Alerts configured
- [ ] Runbook documented
- [ ] Team trained on error budget policy

Related Concepts

SLO vs SLA

• SLO: Internal target. No contractual penalty.
• SLA: External contract. Financial consequences.
• Best Practice: SLO should be stricter than SLA to provide buffer.

Error Budget Policy

When error budget is exhausted:

1. Freeze non-critical feature work
2. Prioritize reliability improvements
3. Conduct incident reviews
4. Address technical debt

When error budget is healthy:

1. Invest in new features
2. Accept more risk
3. Run experiments

References

• Google SRE Book - SLOs
• The Art of SLOs
• Implementing SLOs