High Availability

High Availability (HA) is a system design approach that ensures a high level of operational performance and uptime. It is usually achieved by running multiple instances of the system and ensuring that at least one instance is available at all times. With Dragonfly, you can achieve high availability by using Replication and have read replicas that can be used as a failover mechanism in case the master fails.

Attaining high availability with Dragonfly is dependent on the underlying deployment method. In this section, we will cover how to achieve high availability with Dragonfly in the following scenarios:

High Availability with Dragonfly Operator

Dragonfly Operator is a Kubernetes operator that manages Dragonfly instances. It's available on GitHub and can be installed on any Kubernetes cluster.

One of the main features of the Operator is out-of-the-box high availability. It allows you to run Dragonfly in a highly available configuration with minimal effort. By default, When you set the replicas field to a value greater than 1, the Operator will automatically configure Dragonfly to run in HA mode where one instance is the master and the rest are replicas. As these pods go down, the Operator will automatically reconfigure Replication to maintain the desired number of replicas with one master available at all times.

The application clients can continue using the same Dragonfly service, without any changes, and the Operator will automatically update the pod selectors to point to the right master.

Let's see how this works in practice.

Follow the installation instructions to install the Operator on your Kubernetes cluster.

Creating a Dragonfly instance

To create a sample Dragonfly instance, you can run the following command:

kubectl apply -f https://raw.githubusercontent.com/dragonflydb/dragonfly-operator/main/config/samples/v1alpha1_dragonfly.yaml

This will create a Dragonfly instance with 2 replicas. You can check the status of the instance by running

kubectl describe dragonflies.dragonflydb.io dragonfly-sample

A service of the form <dragonfly-name>.<namespace>.svc.cluster.local will be created, that selects the master instance. You can use this service to connect to the cluster. As pods are added/removed, the service will automatically update to point to the new master.

Connecting with `redis-cli`

To connect to the cluster using redis-cli, you can run:

kubectl run -it --rm --restart=Never redis-cli --image=redis:7.0.10 -- redis-cli -h dragonfly-sample.default

This will create a temporary pod that runs redis-cli and connects to the cluster. After pressing shift + R, You can then run Redis commands as usual. For example, to set a key and get it back, you can run

If you don't see a command prompt, try pressing enter.
dragonfly-sample.default:6379> GET 1
(nil)
dragonfly-sample.default:6379> SET 1 2
OK
dragonfly-sample.default:6379> GET 1
"2"
dragonfly-sample.default:6379> exit
pod "redis-cli" deleted

You can check which pod is the master by running

kubectl get pods -l role=master
NAME                 READY   STATUS    RESTARTS   AGE
dragonfly-sample-0   1/1     Running   0          2m

Let's see how the Operator maintains high availability in the face of failures.

Delete the master pod:

kubectl delete pod -l role=master

The Operator will automatically create a new master pod and update the service to point to the new master. You can check the status of the Dragonfly instance by running

kubectl get pods -l role=master
NAME                 READY   STATUS    RESTARTS   AGE
dragonfly-sample-1  1/1     Running   0          2m

The data should also be preserved. You can check this by running

kubectl run -it --rm --restart=Never redis-cli --image=redis:7.0.10 -- redis-cli -h dragonfly-sample.default
If you don't see a command prompt, try pressing enter.
dragonfly-sample.default:6379> GET 1
"2"
dragonfly-sample.default:6379> exit
pod "redis-cli" deleted

As we can see, the data is still available and the service is pointing to the new master. This way, the Operator ensures that the Dragonfly instance is always available with zero user intervention.

High Availability