Reliability Strategy Builder

Reliability Strategy Builder

Build resilient systems with proper failure handling and SLOs.

Reliability Patterns

1. Circuit Breaker

Prevent cascading failures by stopping requests to failing services.

class CircuitBreaker {
  private state: "closed" | "open" | "half-open" = "closed";
  private failureCount = 0;
  private lastFailureTime?: Date;

  async execute(operation: () => Promise): Promise {
    if (this.state === "open") {
      if (this.shouldAttemptReset()) {
        this.state = "half-open";
      } else {
        throw new Error("Circuit breaker is OPEN");
      }
    }

    try {
      const result = await operation();
      this.onSuccess();
      return result;
    } catch (error) {
      this.onFailure();
      throw error;
    }
  }

  private onSuccess() {
    this.failureCount = 0;
    this.state = "closed";
  }

  private onFailure() {
    this.failureCount++;
    this.lastFailureTime = new Date();

    if (this.failureCount >= 5) {
      this.state = "open";
    }
  }

  private shouldAttemptReset(): boolean {
    if (!this.lastFailureTime) return false;
    const now = Date.now();
    const elapsed = now - this.lastFailureTime.getTime();
    return elapsed > 60000; // 1 minute
  }
}

2. Retry with Backoff

Handle transient failures with exponential backoff.

async function retryWithBackoff(
  operation: () => Promise,
  maxRetries = 3,
  baseDelay = 1000
): Promise {
  for (let attempt = 0; attempt  {
  try {
    // Try primary database
    return await primaryDb.users.findById(userId);
  } catch (error) {
    logger.warn("Primary DB failed, using cache");

    // Fallback to cache
    const cached = await cache.get(`user:${userId}`);
    if (cached) return cached;

    // Final fallback: return minimal user object
    return {
      id: userId,
      name: "Unknown User",
      email: "unavailable",
    };
  }
}

4. Bulkhead Pattern

Isolate failures to prevent resource exhaustion.

class ThreadPool {
  private pools = new Map();

  constructor() {
    // Separate pools for different operations
    this.pools.set("critical", new Semaphore(100));
    this.pools.set("standard", new Semaphore(50));
    this.pools.set("background", new Semaphore(10));
  }

  async execute(priority: string, operation: () => Promise) {
    const pool = this.pools.get(priority);
    await pool.acquire();

    try {
      return await operation();
    } finally {
      pool.release();
    }
  }
}

SLO Definitions

SLO Template

service: user-api
slos:
  - name: Availability
    description: API should be available for successful requests
    target: 99.9%
    measurement:
      type: ratio
      success: status_code < 500
      total: all_requests
    window: 30 days

  - name: Latency
    description: 95% of requests complete within 500ms
    target: 95%
    measurement:
      type: percentile
      metric: request_duration_ms
      threshold: 500
      percentile: 95
    window: 7 days

  - name: Error Rate
    description: Less than 1% of requests result in errors
    target: 99%
    measurement:
      type: ratio
      success: status_code < 400 OR status_code IN [401, 403, 404]
      total: all_requests
    window: 24 hours

Error Budget

Error Budget = 100% - SLO

Example:
SLO: 99.9% availability
Error Budget: 0.1% = 43.2 minutes/month downtime allowed

Failure Mode Analysis

| Component   | Failure Mode | Impact | Probability | Detection               | Mitigation                     |
| ----------- | ------------ | ------ | ----------- | ----------------------- | ------------------------------ |
| Database    | Unresponsive | HIGH   | Medium      | Health checks every 10s | Circuit breaker, read replicas |
| API Gateway | Overload     | HIGH   | Low         | Request queue depth     | Rate limiting, auto-scaling    |
| Cache       | Eviction     | MEDIUM | High        | Cache hit rate          | Fallback to DB, larger cache   |
| Queue       | Backed up    | LOW    | Medium      | Queue depth metric      | Add workers, DLQ               |

Reliability Checklist

Infrastructure

• [ ] Load balancer with health checks
• [ ] Multiple availability zones
• [ ] Auto-scaling configured
• [ ] Database replication
• [ ] Regular backups (tested!)

Application

• [ ] Circuit breakers on external calls
• [ ] Retry logic with backoff
• [ ] Timeouts on all I/O
• [ ] Fallback mechanisms
• [ ] Graceful degradation

Monitoring

• [ ] SLO dashboard
• [ ] Error budgets tracked
• [ ] Alerting on SLO violations
• [ ] Latency percentiles (p50, p95, p99)
• [ ] Dependency health checks

Operations

• [ ] Incident response runbook
• [ ] On-call rotation
• [ ] Postmortem template
• [ ] Disaster recovery plan
• [ ] Chaos engineering tests

Incident Response Plan

Severity Levels

SEV1 (Critical): Complete service outage, data loss
  - Response time: <15 minutes
  - Page on-call immediately

SEV2 (High): Partial outage, degraded performance
  - Response time: <1 hour
  - Alert on-call

SEV3 (Medium): Minor issues, workarounds available
  - Response time: <4 hours
  - Create ticket

SEV4 (Low): Cosmetic issues, no user impact
  - Response time: Next business day
  - Backlog

Incident Response Steps

1. Acknowledge: Confirm receipt within SLA
2. Assess: Determine severity and impact
3. Communicate: Update status page
4. Mitigate: Stop the bleeding (rollback, scale, disable)
5. Resolve: Fix root cause
6. Document: Write postmortem

Best Practices

1. Design for failure: Assume components will fail
2. Fail fast: Don't let slow failures cascade
3. Isolate failures: Bulkhead pattern
4. Graceful degradation: Reduce functionality, don't crash
5. Monitor SLOs: Track error budgets
6. Test failure modes: Chaos engineering
7. Document runbooks: Clear incident response

Output Checklist

• [ ] Circuit breakers implemented
• [ ] Retry logic with backoff
• [ ] Fallback mechanisms
• [ ] Bulkhead isolation
• [ ] SLOs defined (availability, latency, errors)
• [ ] Error budgets calculated
• [ ] Failure mode analysis
• [ ] Monitoring dashboard
• [ ] Incident response plan
• [ ] Runbooks documented

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: patricio0312rev
• Source: patricio0312rev/skillset
• License: MIT

Install and usage instructions live in the source repository linked above.