• September 30, 2026
• 6 min read

Redis — Overview¶

What Is Redis¶

Redis is an in-memory data structure server. Clients send commands over TCP (the RESP protocol), and the server keeps keys in RAM. A key holds a typed value: a string, hash, list, set, sorted set, stream, JSON document and more. Reads and writes usually take well under a millisecond, and every single command is atomic: the server runs commands one by one.

The data can be written to disk (RDB snapshots, AOF log), but Redis is designed around memory: the data set must fit in RAM, and the default setup (RDB snapshots only) can lose minutes of writes on a crash. Treat it as a fast shared state and messaging layer next to your main database, not as a replacement for it.

flowchart LR
    API["orders-api<br/>redis-py"] -- "GET / SET / INCR<br/>RESP over TCP :6379" --> R[("Redis<br/>keys in RAM")]
    W["worker<br/>redis-py"] -- "XREADGROUP / BLPOP" --> R
    API -- "cache miss" --> PG[("PostgreSQL<br/>source of truth")]
    R -. "RDB / AOF" .-> D[("disk")]

Redis, Valkey and Licensing¶

For a test engineer the practical effect is small:

• The data model, commands and protocol of Valkey and Redis are the same for everything in this guide. redis-py, redis-cli, fakeredis and testcontainers work with both.
• Features added after the fork can differ (for example Redis 8 includes JSON and the Query Engine by default). Test against the same server and major version as production — managed services may run either one.
• Other servers speak the Redis protocol too (Dragonfly, KeyDB); compatibility with less common commands varies.

Where QA Engineers Meet Redis¶

When to Use Redis¶

Section Map¶

Minimal Setup¶

docker run -d --name redis -p 127.0.0.1:6379:6379 redis:8   # pin a minor version in CI
docker exec -it redis redis-cli ping                         # PONG

# uv add redis
import redis

r = redis.Redis(host="localhost", port=6379, decode_responses=True)
r.set("myapp:greeting", "hello", ex=60)          # expires in 60 s
print(r.get("myapp:greeting"))                   # 'hello'
print(r.ttl("myapp:greeting"))                   # 60

Examples in this guide were run with Redis 8.10.2 (redis:8 image), Redis 7.4.11 and Valkey 8.1.10 for compatibility checks, redis-py 8.1.0, fakeredis 2.38.0, testcontainers 4.15.0, pytest 9.1.1 and Python 3.13.

Cheat Sheet¶

Keys & Expiry (02)¶

Data Types (02)¶

Server (01, 05)¶

Quick Rules¶

1. Set a TTL on every cache and session key — a key without TTL lives until someone deletes it.
2. Use SCAN, not KEYS — KEYS * blocks the single-threaded command loop on a large keyspace.
3. Prefix keys with the app and entity (myapp:user:42) — it makes scans, ACLs and cleanup safe.
4. Never expose Redis to the internet — bind to a private network, require a password or ACL user, and publish Docker ports on 127.0.0.1 only.
5. Always set client timeouts and know the retry policy — a hanging cache call can take the whole API down.
6. Make multi-step updates atomic — INCR, SET NX, MULTI / EXEC or a Lua script, not "read, change, write" from Python.
7. Choose a maxmemory-policy on purpose — noeviction for queues and locks, allkeys-lru / allkeys-lfu for pure caches.
8. In tests use FLUSHDB on a dedicated database, never FLUSHALL on a shared server.

See also¶

• Digital Garden: Knowledge Base
• Databases — Types, Differences & Selection Guide
• PostgreSQL — Overview
• Queues vs Streams: Message Delivery, Ordering & Reliability
• Pytest — Python Testing Framework
• Docker & Docker Compose — Overview