Redis Patterns

by affaan-mef648e01899bNo license275K starsListed Oct 8, 2026Updated Oct 8, 2026Repository updated 3 days ago

Redisデータ構造パターン、キャッシング戦略、分散ロック、レート制限、Pub/Sub、本番アプリケーション用コネクション管理。

AI-generated overview

Reference guide for Redis data structure patterns, caching, locking, rate limiting, Pub/Sub and connection management.

What it does
This skill is a reference document of Redis best practices for common backend use cases. It covers data structure selection, cache-aside and write-through caching, cache invalidation, session storage, fixed and sliding window rate limiting, distributed locks, Pub/Sub and Streams, key naming and TTL strategy, connection pooling, cluster and Sentinel setups, eviction policies and anti-patterns. It produces guidance and code snippets rather than running anything itself.
When to use it
Use it when adding caching to an application, implementing rate limiting or throttling, building distributed locks or coordination, configuring session or token storage, or using Pub/Sub or Redis Streams for messaging. It also fits production Redis configuration such as pooling, eviction and clustering.
Requirements
No scripts or runtime are required; it is instructions and reference material only. The code examples assume a Redis server and a Python Redis client, and some examples reference optional libraries such as redlock-py.

Redis Patterns

一般的なバックエンド使用例に対するRedisベストプラクティスの参考資料。

How It Works

Redisはメモリ内データ構造ストアで、文字列、ハッシュ、リスト、セット、ソート済みセット、ストリームなどをサポートします。単一インスタンスでは個々のRedisコマンドは原子的ですが、マルチステップワークフローはLuaスクリプト、MULTI/EXECトランザクション、または明示的な同期化が必要です。RDBスナップショットまたはAOFログを通じてデータをオプションで永続化します。クライアントはRESPプロトコルを使用してTCP経由で通信します。接続プール不可欠でリクエストごとのハンドシェイクオーバーヘッドを回避します。

When to Activate

  • アプリケーションにキャッシング追加
  • レート制限またはスロットリング実装
  • 分散ロックまたはコーディネーション構築
  • セッションまたはトークンストレージ設定
  • Pub/SubまたはRedis Streams for messaging使用
  • 本番環境でRedis設定(プール、削除、クラスタリング)

Data Structure Cheat Sheet

Use CaseStructureExample Key
Simple cacheStringproduct:123
User sessionHashsession:abc
LeaderboardSorted Setscores:weekly
Unique visitorsSetvisitors:2024-01-01
Activity feedListfeed:user:456
Event streamStreamevents:orders
Counters / rate limitsString (INCR)ratelimit:user:123
Bloom filter / HLLHyperLogLoghll:pageviews

Core Patterns

Cache-Aside (Lazy Loading)

python
import redisimport json
r = redis.Redis(host='localhost', port=6379, decode_responses=True)
def get_product(product_id: int):    cache_key = f"product:{product_id}"    cached = r.get(cache_key)
    if cached:        return json.loads(cached)
    product = db.query("SELECT * FROM products WHERE id = %s", product_id)    r.setex(cache_key, 3600, json.dumps(product))  # TTL: 1 hour    return product

Write-Through Cache

python
def update_product(product_id: int, data: dict):    # DB書き込み先    db.execute("UPDATE products SET ... WHERE id = %s", product_id)
    # キャッシュを即座に更新    cache_key = f"product:{product_id}"    r.setex(cache_key, 3600, json.dumps(data))

Cache Invalidation

python
# タグベース削除 — セット内で関連キーをグループ化def cache_product(product_id: int, category_id: int, data: dict):    key = f"product:{product_id}"    tag = f"tag:category:{category_id}"    pipe = r.pipeline(transaction=True)    pipe.setex(key, 3600, json.dumps(data))    pipe.sadd(tag, key)    pipe.expire(tag, 3600)    pipe.execute()
def invalidate_category(category_id: int):    tag = f"tag:category:{category_id}"    keys = r.smembers(tag)    if keys:        r.delete(*keys)    r.delete(tag)

Session Storage

python
import timeimport uuid
def create_session(user_id: int, ttl: int = 86400) -> str:    session_id = str(uuid.uuid4())    key = f"session:{session_id}"    pipe = r.pipeline(transaction=True)    pipe.hset(key, mapping={        "user_id": user_id,        "created_at": int(time.time()),    })    pipe.expire(key, ttl)    pipe.execute()    return session_id
def get_session(session_id: str) -> dict | None:    data = r.hgetall(f"session:{session_id}")    return data if data else None
def delete_session(session_id: str):    r.delete(f"session:{session_id}")

Rate Limiting

Fixed Window (Simple)

python
def is_rate_limited(user_id: int, limit: int = 100, window: int = 60) -> bool:    key = f"ratelimit:{user_id}:{int(time.time()) // window}"    pipe = r.pipeline(transaction=True)    pipe.incr(key)    pipe.expire(key, window)    count, _ = pipe.execute()    return count > limit

Sliding Window (Lua — Atomic)

lua
-- sliding_window.lualocal key = KEYS[1]local now = tonumber(ARGV[1])local window = tonumber(ARGV[2])local limit = tonumber(ARGV[3])
redis.call('ZREMRANGEBYSCORE', key, 0, now - window)local count = redis.call('ZCARD', key)
if count < limit then    -- Use unique member (now + sequence) to avoid collisions within the same millisecond    local seq_key = key .. ':seq'    local seq = redis.call('INCR', seq_key)    redis.call('EXPIRE', seq_key, math.ceil(window / 1000))    redis.call('ZADD', key, now, now .. '-' .. seq)    redis.call('EXPIRE', key, math.ceil(window / 1000))    return 1endreturn 0
python
sliding_window = r.register_script(open('sliding_window.lua').read())
def allow_request(user_id: int) -> bool:    key = f"ratelimit:sliding:{user_id}"    now = int(time.time() * 1000)    return bool(sliding_window(keys=[key], args=[now, 60000, 100]))

Distributed Locks

Distributed Lock (Single Node — SET NX PX)

python
import uuid
def acquire_lock(resource: str, ttl_ms: int = 5000) -> str | None:    lock_key = f"lock:{resource}"    token = str(uuid.uuid4())    acquired = r.set(lock_key, token, px=ttl_ms, nx=True)    return token if acquired else None
def release_lock(resource: str, token: str) -> bool:    release_script = """    if redis.call('get', KEYS[1]) == ARGV[1] then        return redis.call('del', KEYS[1])    else        return 0    end    """    result = r.eval(release_script, 1, f"lock:{resource}", token)    return bool(result)
# Usagetoken = acquire_lock("order:payment:123")if token:    try:        process_payment()    finally:        release_lock("order:payment:123", token)

マルチノード設定の場合、フルRedlockアルゴリズムを実装する redlock-py ライブラリを使用してください。

Pub/Sub & Streams

Pub/Sub (Fire-and-Forget)

python
# Publisherdef publish_event(channel: str, payload: dict):    r.publish(channel, json.dumps(payload))
# Subscriber (blocking — run in separate thread/process)def subscribe_events(channel: str):    pubsub = r.pubsub()    pubsub.subscribe(channel)    for message in pubsub.listen():        if message['type'] == 'message':            handle(json.loads(message['data']))

Redis Streams (Durable Queue)

python
# Producerdef emit(stream: str, event: dict):    r.xadd(stream, event, maxlen=10000)  # Cap stream length
# Consumer group — guarantees at-least-once deliverytry:    r.xgroup_create('events:orders', 'processor', id='0', mkstream=True)except Exception:    pass  # Group already exists
def consume(stream: str, group: str, consumer: str):    while True:        messages = r.xreadgroup(group, consumer, {stream: '>'}, count=10, block=2000)        for _, entries in (messages or []):            for msg_id, data in entries:                process(data)                r.xack(stream, group, msg_id)

配信保証、コンシューマーグループ、または再生が必要な場合、Pub/Sub代わりにStreamsを優先してください。

Key Design

Naming Conventions

# Pattern: resource:id:fielduser:123:profileorder:456:statuscache:product:789
# Pattern: namespace:resource:idmyapp:session:abc123myapp:ratelimit:user:123
# Pattern: resource:date (time-bound keys)stats:pageviews:2024-01-01

TTL Strategy

Data TypeSuggested TTL
User session24h (86400)
API response cache5–15 min
Rate limit windowMatch window size
Short-lived tokens5–10 min
Leaderboard1h–24h
Static/reference data1h–1 week

常にTTLを設定してください。TTLなしのキーは無限に蓄積してメモリ圧力を引き起こします。

Connection Management

Connection Pooling

python
from redis import ConnectionPool, Redis
pool = ConnectionPool(    host='localhost',    port=6379,    db=0,    max_connections=20,    decode_responses=True,    socket_connect_timeout=2,    socket_timeout=2,)
r = Redis(connection_pool=pool)

Cluster Mode

python
from redis.cluster import RedisCluster
r = RedisCluster(    startup_nodes=[{"host": "redis-1", "port": 6379}],    decode_responses=True,    skip_full_coverage_check=True,)

Sentinel (High Availability)

python
from redis.sentinel import Sentinel
sentinel = Sentinel(    [('sentinel-1', 26379), ('sentinel-2', 26379)],    socket_timeout=0.5,)master = sentinel.master_for('mymaster', decode_responses=True)replica = sentinel.slave_for('mymaster', decode_responses=True)

Eviction Policies

PolicyBehaviorBest For
noevictionError on write when fullQueues / critical data
allkeys-lruEvict least recently usedGeneral cache
volatile-lruLRU only among keys with TTLMixed data store
allkeys-lfuEvict least frequently usedSkewed access patterns
volatile-ttlEvict soonest-to-expirePrioritize long-lived data

redis.confを通じて設定:maxmemory-policy allkeys-lru

Anti-Patterns

Anti-PatternProblemFix
Keys with no TTLMemory grows unboundedAlways set TTL
KEYS * in productionBlocks the server (O(N))Use SCAN cursor
Storing large blobs (>100KB)Slow serialization, memory pressureStore reference + fetch from object store
Single Redis for everythingNo isolation between cache & queueUse separate DBs or instances
Ignoring connection pool limitsConnection exhaustion under loadSize pool to workload
Not handling cache miss stampedeThundering herd on cold startUse locks or probabilistic early expiry
FLUSHALL without thoughtWipes entire instanceScope deletes by key pattern

Cache Miss Stampede Prevention

python
import threading
_locks: dict[str, threading.Lock] = {}_locks_mutex = threading.Lock()
def get_with_lock(key: str, fetch_fn, ttl: int = 300):    cached = r.get(key)    if cached:        return json.loads(cached)
    with _locks_mutex:        if key not in _locks:            _locks[key] = threading.Lock()        lock = _locks[key]    with lock:        cached = r.get(key)  # Re-check after acquiring lock        if cached:            return json.loads(cached)        value = fetch_fn()        r.setex(key, ttl, json.dumps(value))        return value

マルチプロセスデプロイメント:インプロセスロックを上記の分散ロックセクション から acquire_lock/release_lock に置き換えてください。

Examples

Django/Flask APIエンドポイントにキャッシング追加: レスポンスに5分TTLでCache-asideを使用。リクエストパラメータでキーを指定。

ユーザーごとにAPIレート制限: 低トラフィックエンドポイントに固定ウィンドウを pipeline(transaction=True) で使用;正確なユーザーごと制限にはsliding-windowの Lua使用。

ワーカー間のバックグラウンドジョブ調整: 予想ジョブ期間を超えるTTLで acquire_lock を使用。常に finally ブロックでリリース。

複数購読者への通知のファンアウト: ファイアアンドフォーゲットにPub/Subを使用。保証配信または再生が必要な場合、Streamsに切り替え。

Quick Reference

PatternWhen to Use
Cache-asideRead-heavy, tolerate slight staleness
Write-throughStrong consistency required
Distributed lockPrevent concurrent access to a resource
Sliding window rate limitAccurate per-user throttling
Redis StreamsDurable event queue with consumer groups
Pub/SubBroadcast with no delivery guarantees needed
Sorted Set leaderboardRanked scoring, pagination
HyperLogLogApproximate unique count at low memory

Related

  • Skill: postgres-patterns — リレーショナルデータパターン
  • Skill: backend-patterns — APIおよびサービスレイヤーパターン
  • Skill: database-migrations — スキーマバージョニング
  • Skill: django-patterns — Djangoキャッシュフレームワーク統合
  • Agent: database-reviewer — 全データベースレビューワークフロー

Source and attribution

Source:affaan-m/eccindocs/ja-JP/skills/redis-patternsat commitef648e0

License: No license

Content belongs to its original authors. SourceWeft indexes it from a public repository.

Report or request removal