Database Query Best Practices - Prevent Connection Pool Exhaustion

CRITICAL SKILL - READ THIS BEFORE QUERYING RAILWAY DATABASE

The Problem

When querying Railway production database from local development environment, you may encounter:

psycopg2.OperationalError: sorry, too many clients already

This happens when:

• Multiple local dev servers are running (npm run dev, npm run develop)
• Each dev server holds database connections open
• Connection pool limit is reached (300 total connections)
• No new connections can be established

ALWAYS Use This Approach

1. Check Connection Pool Status FIRST

Before ANY database query, check active connections:

python3 << 'EOF'
import psycopg2

DATABASE_URL = "postgresql://postgres:password@host:port/railway"

try:
    conn = psycopg2.connect(DATABASE_URL, connect_timeout=10)
    cur = conn.cursor()

    # Check active connections
    cur.execute("""
        SELECT count(*)
        FROM pg_stat_activity
        WHERE datname = 'railway'
    """)
    active = cur.fetchone()[0]

    # Check pool limit
    cur.execute("SHOW max_connections")
    max_conn = cur.fetchone()[0]

    print(f"Active connections: {active}/{max_conn}")

    if active > int(max_conn) * 0.9:
        print(f"WARNING: Connection pool at {(active/int(max_conn))*100:.1f}% capacity")
        print("Consider closing dev servers before querying")
    else:
        print("Connection pool healthy - safe to query")

    cur.close()
    conn.close()

except Exception as e:
    print(f"Cannot connect: {e}")
    print("\nSOLUTION:")
    print("1. Kill local dev servers: pkill -9 -f 'npm run dev'")
    print("2. Wait 30 seconds for connections to close")
    print("3. Try again")
EOF

2. Always Use Context Managers

NEVER do this:

conn = psycopg2.connect(DATABASE_URL)
cur = conn.cursor()
cur.execute("SELECT * FROM users")
# Forgot to close! Connection leak!

ALWAYS do this:

import psycopg2

DATABASE_URL = "postgresql://..."

try:
    conn = psycopg2.connect(DATABASE_URL, connect_timeout=30)
    cur = conn.cursor()

    # Do your queries
    cur.execute("SELECT * FROM users")
    results = cur.fetchall()

    # Process results...

finally:
    # ALWAYS close in finally block
    if cur:
        cur.close()
    if conn:
        conn.close()

3. Use Short-Lived Connections

For one-off queries, open connection, query, close immediately:

def get_user_count():
    """Get user count - connection opened and closed in function"""
    conn = None
    try:
        conn = psycopg2.connect(DATABASE_URL, connect_timeout=30)
        cur = conn.cursor()

        cur.execute("SELECT COUNT(*) FROM users")
        count = cur.fetchone()[0]

        cur.close()
        return count

    finally:
        if conn:
            conn.close()  # Connection immediately released

# Good: Connection closed after function returns
user_count = get_user_count()

4. Kill Dev Servers Before Queries

If you need to run database queries, stop dev servers first:

# Stop all dev servers
pkill -9 -f "npm run dev"
pkill -9 -f "npm run develop"

# Wait for connections to close
sleep 10

# Now safe to query
python3 scripts/your_query_script.py

# Restart dev servers after
cd src/backend && npm run dev &
cd AINative-website && npm run dev &

5. Use Railway CLI for Quick Queries (RECOMMENDED)

Instead of Python scripts, use Railway CLI when possible:

# Login to Railway
railway login

# Link to project
railway link

# Run SQL query directly
railway run psql -c "SELECT COUNT(*) FROM users WHERE created_at >= NOW() - INTERVAL '30 days'"

This uses Railway's managed connections and doesn't consume local pool.

Emergency: Connection Pool Full

If you encounter "too many clients already":

Option 1: Kill Local Dev Servers

pkill -9 -f "npm run dev"
pkill -9 -f "npm run develop"
pkill -9 node
sleep 30  # Wait for DB connections to close

Option 2: Check Active Connections on Railway

来源于7w4.net。

railway run psql -c "
SELECT
    pid,
    usename,
    application_name,
    client_addr,
    state,
    query_start
FROM pg_stat_activity
WHERE datname = 'railway'
ORDER BY query_start DESC
LIMIT 20"

Option 3: Terminate Idle Connections (LAST RESORT)

railway run psql -c "
SELECT pg_terminate_backend(pid)
FROM pg_stat_activity
WHERE datname = 'railway'
  AND state = 'idle'
  AND query_start < NOW() - INTERVAL '10 minutes'"

Pre-Query Checklist

Before running ANY database query, verify:

• [ ] Are local dev servers running? → Stop them first
• [ ] Is connection pool healthy? → Run connection check
• [ ] Using try/finally to close connections? → Yes
• [ ] Using short-lived connections? → Yes
• [ ] Can I use Railway CLI instead? → Prefer this

Summary

The Golden Rule:

ALWAYS close database connections immediately after use. NEVER leave connections open in dev servers during queries.

Best Approach:

1. Stop dev servers
2. Query database
3. Close connection
4. Restart dev servers

Even Better: Use Railway CLI for ad-hoc queries instead of Python scripts.

这个 Skill 质量中等偏上。优点是问题定位准确，提供的解决方法和代码示例都很实用，预查询检查清单能有效帮助避免问题。不足是内容覆盖范围有限，只针对特定数据库环境，缺乏更全面的使用指南和改进建议。对于需要处理数据库连接管理的开发者来说有一定帮助，但实用性和泛用性还有提升空间。