This docs contains following sections:
Why This ProjectStart ClusterStop ClusterRun Redis CLIMake Cluster Discoverable From Other ContainerAcknowledgements
I'm total noob when it comes to docker & devops. Yet for some reason I need to be able to run redis cluster in docker for my dev environment.
So this project is the result of my learning experience towards installing redis cluster in docker. In the core, I use supervisord to spin up multiple redis instances in single container, then convert them to redis cluster using bash script running in CMD section at Dockerfile.
The reason why I choose to spin up multiple redis instances in single container rather than in the multiple one is simpy because I want to be able to access my redis cluster from host. If I spin up them from multiple container, I won't be able to do it since my redis client will always redirect me to another container using private ip. So the simplest solution I could think of is to spin up them from single instance thus I won't have trouble with redirection to private ip.
The client redirection itself is the inherent property of redis cluster. For more details please refer here.
To spin up the container, I prefer using docker-compose instead of docker stack. The reason is simpy because it offers me more feature & flexibility. But we could also spin up the container using docker stack. In this project, I already prepared helper for both commands using make. Please read following sections for more details.
To build this project, I'm using docker version 18.03.1-ce for Mac. The redis version I'm using is 4.0.10.
To start the cluster using docker-compose, we could use following helper command:
$> make upTo start the cluster using docker stack, we could use following helper command:
$> make start-swarmTo stop the cluster started using docker-compose, we could use following helper command:
$> make downTo stop the cluster started using docker stack, we could use following helper command:
$> make stop-swarmIf we want to run the internal redis-cli inside the container for some reason (for example redis-cli is not installed on host), we could use following helper command:
$> make run-cliBut this command only applies if we start our cluster using docker-compose.
If redis-cli is installed on host, we could also use it to connect with cluster. Simply use following command to do that:
$> redis-cli -c -p 7000Notice that in this stack, the assumption we use for accessing the cluster is via host.
So for example if we have web server which need to connect to redis cluster, we need to execute this server on host, not on container inside the stack. If we do the latter, the server won't be able to connect to the cluster since the cluster is not discoverable to other container.
But this kind of limitation is troublesome, right? Especially if we want to bundle our web server together with redis cluster on the same stack.
Luckily we could make this redis cluster discoverable to other container. However when we do this, the cluster won't be discoverable to host. So we need to make sure which one is required in our dev environment.
To make the redis cluster discoverable to other container, unset IP env variable on docker-compose.yml.
So from this:
environment:
- IP=0.0.0.0To this:
environment:
- IPIf we want to make it discoverable by host just reset the IP env variable to 0.0.0.0.
I would like to thank a lot to @thelinuxlich for giving me starting point to start this project via his awesome gist.
I would also like to thank @Grokzen for his awesome project which introduce me to the idea of using single container to spin up multiple redis instances so we could connect to our redis cluster from host.
In essence my project & his are doing the same thing (even the name is also coincidentally same 😂). But underneath, I personally think mine is much more easier to understand from the perspective of noob like myself 😅. But if you want richer features, please take a look at his project.