Home / Databases / Redis

DatabasesHow-toBeginner

Redis basics: data types, TTLs and safe redis-cli habits for your first day

Learn Redis by trying it - connect with redis-cli, use strings, hashes, lists, sets and sorted sets, set TTLs, find keys safely with SCAN, check memory use and understand the errors you are most likely to see.

Deepak Bhardwaj6 min read

Applies toRedis 6 and later, Valkey, redis-cli, Azure Cache for Redis, Amazon ElastiCache

The goal

Your app uses Redis, and you've been asked to look at it. Redis is an in-memory data store. It's usually a cache, sometimes also a queue or a session store. This post covers the commands you'll use often and the few that can cause an outage.

Step 1: connect

Practice on a throwaway server:

docker run -d --name redis-dev -p 6379:6379 redis:7
docker exec -it redis-dev redis-cli

For a real server:

redis-cli -h redis01.contoso.com -p 6379 --user appuser --askpass

Managed services usually need TLS:

redis-cli -h myredis.redis.cache.windows.net -p 6380 --tls --askpass

--askpass prompts for the password, which keeps it out of your shell history. Test your connection:

PING
INFO server
SELECT 0
DBSIZE

PING should answer PONG. DBSIZE shows the number of keys in the current database.

Step 2: strings and TTLs

A key points to a value. Strings are the simplest kind:

SET session:1001 "asha" EX 3600
GET session:1001
TTL session:1001
EXPIRE session:1001 600
PERSIST session:1001
DEL session:1001
EXISTS session:1001

Reading a TTL result:

TTL returns Meaning
A positive number Seconds until it expires
-1 The key exists and never expires
-2 The key doesn't exist

Counters are also strings, and updates are atomic:

INCR page:views
INCRBY page:views 10
GET page:views

Tip

Give cache keys a TTL unless you have a good reason not to. Keys with no expiry build up until memory runs out. Use a clear naming pattern too, such as app:type:id.

Step 3: other data types

Each key has one type. Use the right one, and you avoid a lot of application code.

Type Good for Example
String Cached values, counters SET, GET, INCR
Hash An object with fields HSET, HGET, HGETALL
List Simple queues, recent items LPUSH, RPOP, LRANGE
Set Unique members SADD, SISMEMBER, SMEMBERS
Sorted set Rankings, time-ordered items ZADD, ZRANGE, ZSCORE

Try them:

HSET user:1001 name "Asha" city "Pune" plan "pro"
HGET user:1001 city
HGETALL user:1001

LPUSH jobs "job-1" "job-2"
LRANGE jobs 0 -1
RPOP jobs

SADD online:users 1001 1002 1003
SISMEMBER online:users 1002
SCARD online:users

ZADD leaderboard 150 "asha" 90 "ravi" 220 "meera"
ZRANGE leaderboard 0 -1 REV WITHSCORES

TYPE user:1001 tells you what kind of key you have.

Step 4: find keys without hurting the server

Redis runs commands one at a time on a main thread, so a slow command makes everyone wait.

# Safe: walks the keys in small steps
redis-cli -h redis01 --scan --pattern 'session:*' | head -20

# Rough view of the biggest keys
redis-cli -h redis01 --bigkeys

Inside the shell:

SCAN 0 MATCH session:* COUNT 100

SCAN returns a cursor and some keys. Call it again with the new cursor until it returns 0.

Caution

Avoid these on production:

  • KEYS * blocks the server while it reads every key.
  • FLUSHALL and FLUSHDB delete all data. Check the server name twice, and keep them restricted with ACLs.
  • HGETALL, SMEMBERS or LRANGE 0 -1 on very large keys also block.
  • DEL on a huge key can stall the server. Use UNLINK, which frees the memory in the background.

Step 5: check health and memory

INFO memory
INFO stats
INFO replication
CONFIG GET maxmemory
CONFIG GET maxmemory-policy
MEMORY USAGE user:1001
CLIENT LIST
SLOWLOG GET 10

The fields to look for:

Field Section Meaning
used_memory_human memory Memory used by data
maxmemory_human memory The limit. 0 means none
mem_fragmentation_ratio memory Much higher than 1.5 suggests fragmentation
evicted_keys stats Keys removed because of the memory limit
keyspace_hits, keyspace_misses stats Whether the cache is working
connected_clients clients Compare with maxclients
role replication master or replica

Work out the cache hit rate:

hit rate = keyspace_hits / (keyspace_hits + keyspace_misses)

A sudden drop in hit rate can overload the database behind the cache. Some managed services block or rename a few of these commands.

Common errors

Error Meaning Fix
NOAUTH Authentication required You haven't logged in Reconnect with credentials, or run AUTH
WRONGPASS Bad username or password Check the secret, and the user
NOPERM Your ACL doesn't allow the command or key Ask for the right permission
WRONGTYPE Wrong command for the key's type Check with TYPE key
OOM command not allowed Memory limit reached, and policy won't evict See memory full
READONLY You're writing to a replica Connect to the primary
MOVED Redis Cluster: the key is on another node Use redis-cli -c or a cluster client
MISCONF Disk snapshot failed, so writes stop See the MISCONF fix

Is the data safe?

By default, think of Redis as fast and not necessarily durable. Depending on the configuration, a restart can lose recent writes, or even all of them.

CONFIG GET save
CONFIG GET appendonly
Setting Effect
save has values Periodic snapshots (RDB)
appendonly yes Write log (AOF), which loses less on a crash
Both off All data is lost on restart

If the data matters, read about persistence. If it's only a cache, plan for the app to refill it.

Prevent it next time

  • Set maxmemory and an eviction policy that suits the use.
  • Require a password or ACL, use TLS and never expose port 6379 to the internet.
  • Monitor memory, evictions, hit rate and connected clients.
  • Keep the redis-cli cheat sheet handy, and read about caching patterns next.

Related fixes