Imported from jeremylongshore/tons-of-skills-marketplace (plugins/saas-packs/replit-pack/skills/replit-reliability-patterns/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill replit-reliability-patterns. Copyright stays with the author (MIT).

Replit Reliability Patterns

Overview

Production reliability patterns for Replit's container-based hosting. Replit containers restart on deploy, sleep on inactivity (Autoscale), and have ephemeral filesystems. These patterns ensure your app survives container lifecycle events gracefully.

Prerequisites

• Replit Deployment configured
• External storage for persistent state (PostgreSQL or Object Storage)
• Understanding of Replit container lifecycle

Container Lifecycle

Container starts → App boots → Handles requests → [Sleep or Restart]
                                                         │
                    ┌────────────────────────────────────┘
                    │
            ┌───────┴──────┐
            │ Sleep trigger │  Autoscale: no traffic for ~5 min
            │ Restart trigger│  Deploy, config change, or crash
            └───────┬──────┘
                    │
        State lost: filesystem, in-memory data, caches
        State kept: PostgreSQL, KV Database, Object Storage, Secrets

Instructions

Step 1: Graceful Startup

// Handle cold starts — prioritize accepting requests over initialization
import express from 'express';

const app = express();
let ready = false;

// Accept requests immediately
app.listen(parseInt(process.env.PORT || '3000'), '0.0.0.0', () => {
  console.log(`Server started in ${process.uptime().toFixed(1)}s`);
  // Initialize in background
  initialize().catch(console.error);
});

// Health endpoint reflects readiness
app.get('/health', (req, res) => {
  res.status(ready ? 200 : 503).json({
    status: ready ? 'ready' : 'initializing',
    uptime: process.uptime(),
  });
});

async function initialize() {
  const start = Date.now();
  // Pre-connect database
  await pool.query('SELECT 1');
  // Warm caches
  await warmCache();
  ready = true;
  console.log(`Initialization complete in ${Date.now() - start}ms`);
}

Step 2: Graceful Shutdown

Replit sends SIGTERM before stopping containers. Save state during shutdown.

// Graceful shutdown handler
let shutdownInProgress = false;

async function shutdown(signal: string) {
  if (shutdownInProgress) return;
  shutdownInProgress = true;

  console.log(`${signal} received. Shutting down gracefully...`);

  // 1. Stop accepting new requests
  server.close();

  // 2. Finish in-flight requests (give them 10 seconds)
  await new Promise(resolve => setTimeout(resolve, 10000));

  // 3. Save critical state
  try {
    await saveAppState();
  } catch (err: any) {
    console.error('Failed to save state:', err.message);
  }

  // 4. Close database connections
  await pool.end();

  // 5. Close KV database
  // @replit/database: call close() to terminate cleanly
  // replit.db (Python): call replit.db.close()

  console.log('Shutdown complete');
  process.exit(0);
}

process.on('SIGTERM', () => shutdown('SIGTERM'));
process.on('SIGINT', () => shutdown('SIGINT'));

const server = app.listen(PORT, '0.0.0.0');

Step 3: Persistent State (Survive Restarts)

Never rely on the local filesystem for data that must persist:

// BAD: Local filesystem is ephemeral
import fs from 'fs';
fs.writeFileSync('state.json', JSON.stringify(appState));
// GONE after container restart

// GOOD: Use Replit KV Database for small state
import Database from '@replit/database';
const kv = new Database();

async function saveAppState() {
  await kv.set('app:state', {
    lastActive: Date.now(),
    version: process.env.npm_package_version,
    counters: appCounters,
  });
}

async function loadAppState() {
  return (await kv.get('app:state')) || { lastActive: 0, counters: {} };
}

// GOOD: Use Object Storage for larger data / files
import { Client } from '@replit/object-storage';
const storage = new Client();

async function saveReport(data: any) {
  await storage.uploadFromText(
    `reports/${new Date().toISOString()}.json`,
    JSON.stringify(data)
  );
}

Python equivalent:

from replit import db
from replit.object_storage import Client as Storage
import json, time, signal, sys

# Persistent state via KV
def save_state(data):
    db["app:state"] = {
        "data": data,
        "saved_at": time.time()
    }

def load_state():
    return db.get("app:state", {"data": {}, "saved_at": 0})

# Persistent files via Object Storage
storage = Storage()

def save_backup(filename, content):
    storage.upload_from_text(f"backups/{filename}", content)

# Graceful shutdown
def shutdown(signum, frame):
    print("Shutting down...")
    save_state(app_data)
    db.close()
    sys.exit(0)

signal.signal(signal.SIGTERM, shutdown)

Step 4: Keep-Alive for Non-Deployment Repls

For Repls not using Deployments, prevent sleep with external pinging:

Option 1: External cron service (recommended)
- UptimeRobot (free: 50 monitors, 5-min intervals)
- cron-job.org (free: 1-min intervals)
- URL: https://your-repl.replit.app/ping
- Interval: 4 minutes (Replit sleeps after ~5 min)

Option 2: Self-ping (less reliable — sleeps if service itself sleeps)

// Lightweight ping endpoint
app.get('/ping', (req, res) => res.send('pong'));

Best option: Use Replit Deployments instead
- Autoscale: scales to zero but wakes on request
- Reserved VM: always-on, no sleeping
- Both are more reliable than keep-alive hacks

Step 5: Database Connection Resilience

// Auto-reconnect on database failures
import { Pool } from 'pg';

function createResilientPool(): Pool {
  const pool = new Pool({
    connectionString: process.env.DATABASE_URL,
    ssl: { rejectUnauthorized: false },
    max: 5,
    idleTimeoutMillis: 30000,
    connectionTimeoutMillis: 5000,
  });

  pool.on('error', (err) => {
    console.error('Pool error (will auto-reconnect):', err.message);
    // Pool auto-replaces failed connections on next query
  });

  return pool;
}

// Retry wrapper for database queries
async function queryWithRetry(
  pool: Pool,
  sql: string,
  params?: any[],
  retries = 3
): Promise<any> {
  for (let attempt = 1; attempt <= retries; attempt++) {
    try {
      return await pool.query(sql, params);
    } catch (err: any) {
      if (attempt === retries) throw err;
      console.warn(`DB query failed (attempt ${attempt}): ${err.message}`);
      await new Promise(r => setTimeout(r, 1000 * attempt));
    }
  }
}

Step 6: Deployment Health Monitor

// Self-monitoring deployment health
const healthMetrics = {
  startTime: Date.now(),
  requestCount: 0,
  errorCount: 0,
  lastError: null as string | null,
};

app.use((req, res, next) => {
  healthMetrics.requestCount++;
  res.on('finish', () => {
    if (res.statusCode >= 500) {
      healthMetrics.errorCount++;
      healthMetrics.lastError = `${res.statusCode} on ${req.method} ${req.path}`;
    }
  });
  next();
});

app.get('/health', (req, res) => {
  const uptime = (Date.now() - healthMetrics.startTime) / 1000;
  const errorRate = healthMetrics.requestCount > 0
    ? (healthMetrics.errorCount / healthMetrics.requestCount * 100).toFixed(2)
    : '0';

  res.json({
    status: parseFloat(errorRate) > 5 ? 'degraded' : 'healthy',
    uptime: `${uptime.toFixed(0)}s`,
    requests: healthMetrics.requestCount,
    errors: healthMetrics.errorCount,
    errorRate: `${errorRate}%`,
    lastError: healthMetrics.lastError,
    memory: Math.round(process.memoryUsage().heapUsed / 1024 / 1024) + 'MB',
  });
});

Error Handling

Resources

• Replit Deployments
• Replit KV Database
• Object Storage
• Deployment Rollbacks

Next Steps

For policy enforcement, see replit-policy-guardrails.