Restate 1.2 Announcement Demos: Distributed Restate

This repository contains two demos showcasing distributed deployments with Restate. The first one demonstrates high availability, and the second one shows how to scale up from a single node to a cluster, and reconfigure clusters on the fly.

Prerequisites

You need the following tools installed:

• Node.js and npm to run the TypeScript services used in this demo.
• Docker and Docker Compose to run the Restate cluster.

Demo 1: High Availability Demo

This demo shows how Restate can handle node failures and still be available for writes. It also shows how the cluster recovers when the nodes are brought back up, maintaining consistency at all times. Finally, it shows how the cluster stalls when there are not enough nodes available to maintain the replication factor.

Pre-demo setup

Start the service:

npm run app-dev

Go to the folder with the docker compose file for this demo and start the Restate cluster:

cd availability-demo
docker compose up

Register the service

restate deployments register http://host.docker.internal:9080

You can also do this via the UI at http://localhost:9070, if you don't have the CLI installed.

Start the three counter clients in three different terminal windows, to start sending requests:

npm run client

This client will try sending requests to node1 by default. If this node is down, it tries node2, and then node3.

npm run client2

This client will try sending requests to node2 by default. If this node is down, it tries node3, and then node1.

npm run client3

This client will try sending requests to node3 by default. If this node is down, it tries node1, and then node2.

Note: in a real-world scenario, you would have a load balancer in front of the nodes, and the clients would be configured to send requests to the load balancer. For the sake of the demo, to show clearly which node is down, we are using the clients to send requests to specific nodes.

Demo

Check the partitioning and leaders per partition:

docker compose exec node1 restatectl status

You can use restatectl status if you have restatectl installed locally.

Have a look at a partition that has node3 as the leader. We will now kill this node and see how the leader changes. The cluster maintains write availability, since the replication factor is 2, and there are still two nodes available.

docker compose kill node3

Check how the leaders changed for the partitions that before had node3 as leader:

docker compose exec node1 restatectl status

Restart the node:

docker compose restart node3

You will see that the node becomes available again and gets added to the cluster.

Now kill two nodes and to see how the cluster stalls when we lose write availability (# nodes < replication factor):

docker compose kill node2
docker compose kill node3

Bring back the nodes, and see how the cluster recovers and starts processing again:

docker compose restart node2
docker compose restart node3

Demo 2: Scale up from one node to a cluster

This demo shows how to scale up a Restate deployment from a single node to a cluster, and how to reconfigure the cluster on the fly.

Pre-demo setup

Start the service:

npm run app-dev

Go to the folder with the docker compose file for this demo and start the first node of the Restate cluster:

cd scale-demo
docker compose up -d node1

Register the service

restate deployments register http://host.docker.internal:9080

You can also do this via the UI at http://localhost:9070, if you don't have the CLI installed.

Start the counter client in a different terminal window, to start sending requests:

npm run client

Demo

Show the status:

docker compose exec node1 restatectl status

You can use restatectl status if you have restatectl installed locally.

Spin up the other nodes:

docker compose up -d node2

docker compose up -d node3

Reconfigure the cluster to have log replication set to 2:

docker compose exec node1 restatectl cluster config set --log-replication 2

Show that the metadata cluster includes these nodes now:

docker compose exec node1 restatectl meta status

Show the partitioning and leaders per partition:

docker compose exec node1 restatectl status

Rerun until applied-lsn is the same, or 1 less, for all nodes.

Then kill node1:

docker compose kill node1

Show that the cluster is still available and that the leaders have changed:

docker compose exec node2 restatectl status

Note that this command is pointed at node2, since node1 is down.