Scaling

Scaling your applications in Railway is made easy with configurable options for horizontal scaling as well as vertical autoscaling out-of-the-box.

How it Works

Vertical Autoscaling

By default Railway will scale your service up to the specified vCPU and Memory limits of your plan.

Horizontal Scaling with Replicas

Scale horizontally by manually increasing the number of replicas for a service in the service settings. Increasing the number of replicas on a service will create multiple instances of the service deployment.

Each replica has access to the full resources allocated by your plan. For instance, with the Pro plan, each of your replicas can utilize up to 32 vCPU and 32GB of memory, for example, if you had 2 replicas, your service would be able to utilize up to 64 vCPU and 64GB of memory split between the 2 replicas.

Multi-region replicas

Multi-region replicas are exactly as advertised -- horizontally scaled replicas that are located in different geographic regions.

The service settings panel will reveal an interface for assigning replicas to different regions.

Creating, deleting, and re-assigning replicas will trigger a staged change which upon applying will trigger a redeploy.

Load Balancing Between Replicas

If you are using multi-region replicas, Railway will automatically route public traffic to the nearest region and then randomly distribute requests to the replicas within that region.

If you are using a single region with multiple replicas, Railway will randomly distribute public traffic to the replicas of that region.

We plan to add more advanced load balancing strategies in the future.

Metrics

For services with multiple replicas, the metrics from all replicas are summed up and displayed in the metrics tab, for example, if you have 2 replicas, each using 100 MB of memory, the memory usage displayed in the metrics tab will be 200 MB.

Sticky Sessions

For now Railway does not support sticky sessions nor report the usage of the replicas within the metrics view.

Support

For information on how to use horizontal scaling with replicas, refer to this guide.