getting-started.md

deploy with docker

You can use docker-compose.yaml to deploy all needed services. Make sure to uncomment volumes if you need persistence.

docker compose up etcd -d 

It deploys etcd to store state of the cluster. After running the etcd, we need to do some inital setup in it. First create root user and another user with limited privileges for MySQL Manager.

docker compose exec etcd etcdctl user add root --new-user-password="password"
docker compose exec etcd etcdctl user grant-role root root
docker compose exec etcd etcdctl user add mm --new-user-password="password"
docker compose exec etcd etcdctl role add mm 
docker compose exec etcd etcdctl role grant-permission mm --prefix=true readwrite mm/cluster1/
docker compose exec etcd etcdctl user grant-role mm mm
docker compose exec etcd etcdctl auth enable

We created mm user with password password. In a production environment you need to set a much stronger password. User mm is granted access to mm/cluster1/ prefix in etcd. Note that the environment variables in mm container must match the values given above.

Then run all other containers:

docker compose up -d --build

It deploys two MySQL servers, their exporters and a ProxySQL. MySQL Manager is set up by building an image from Dockerfile at the root of the project. Now start cluster with this command:

docker compose exec mm python cli/mysql-cli.py init -f /etc/mm/cluster-spec.yaml

Running python cli/mysql-cli.py init -f /etc/mm/cluster-spec.yaml in mm makes MySQL Manager to start cluster setup, watch for their state and failover if needed. After a minute check mm logs:

docker compose logs mm -f 

You should see these lines in its logs:

Source is mysql-s1
Replica is mysql-s2

You can check server states by running this:

docker compose exec mm python cli/mysql-cli.py mysql get-cluster-status

The output is like this:

source=up
replica=up

destroy

to destroy the containers run:

docker compose down