Skip to content
aki

Quick start

Start the aki server, connect with redis-cli, write some keys, and watch them survive a restart.

This walks you from an empty directory to a durable Redis-compatible store in a few commands. You need the aki binary from the installation page and any Redis client. The examples use redis-cli, which ships with Redis.

Start the server

aki server --dbfile data.aki

aki creates data.aki (plus the data.aki-wal and data.aki-shm sidecars) and listens on 127.0.0.1:6379, the default Redis port. Leave it running and open a second terminal.

To listen somewhere else, pass --addr:

aki server --dbfile data.aki --addr 0.0.0.0:6379

Connect and run commands

redis-cli

127.0.0.1:6379> set greeting "hello from aki"
OK
127.0.0.1:6379> get greeting
"hello from aki"
127.0.0.1:6379> rpush tasks build test ship
(integer) 3
127.0.0.1:6379> lrange tasks 0 -1
1) "build"
2) "test"
3) "ship"
127.0.0.1:6379> hset user:1 name ada role admin
(integer) 2
127.0.0.1:6379> hgetall user:1
1) "name"
2) "ada"
3) "role"
4) "admin"

Everything you know from Redis works the same way. Strings, lists, sets, hashes, sorted sets, streams, expirations, and the rest all behave as they do on a real Redis server.

Use RESP3

aki speaks both RESP2 and RESP3. Ask for RESP3 with the -3 flag and you get the richer reply types, such as maps for HGETALL:

redis-cli -3 hgetall user:1

1# "name" => "ada"
2# "role" => "admin"

You can also switch a connection at runtime with HELLO 3.

Prove it is durable

This is the part Redis makes you work for and aki does by default. Write a key, stop the server with a hard kill, start it again, and the key is still there.

redis-cli set survives yes
# stop the server with Ctrl-C, or even kill -9 its PID
aki server --dbfile data.aki
redis-cli get survives

"yes"

The write was committed to the write-ahead log and fsynced before the client got its OK, so a crash cannot lose it. On the next open aki replayed the log and recovered to the last committed transaction.

Bring data in from Redis

If you have an existing Redis dump, import it when you first create the file:

aki server --dbfile data.aki --load-rdb /path/to/dump.rdb

The import only runs when data.aki does not exist yet, so it never clobbers a database you already have. You can also convert between formats offline with aki import and aki dump, covered in the CLI reference.

Inspect the file

Without a running server, you can validate the on-disk structure:

aki check data.aki

Where to go next

3 min read