Clearer memory limits for application containers

As part of our ongoing work to improve platform stability, we are updating how memory is managed for application containers on Scalingo, making memory behavior more predictable and reliable.

Starting 1 January 2027, application containers will no longer be able to rely on swap memory.

Most applications should not require immediate changes, but the transition period is a good opportunity to review memory usage and confirm that each application is correctly sized.

Why we are making this change

Until now, swap has acted as an additional buffer when an application exceeds its allocated memory. In practice, we have observed that frequent or heavy swap usage can hide under-provisioned applications, memory leaks, or workloads that require more memory than their current container size provides.

On a shared platform, intensive swap usage can also have an impact beyond the application using it. It can degrade node performance and affect the quality of service experienced by other customers.

For this reason, we are moving to a clearer memory model, where applications are expected to operate within their configured memory limit. This makes memory-related issues easier to identify, understand, and act on.

The goal is not to push applications toward larger containers by default, but to make memory behavior explicit so teams can better distinguish healthy usage from under-provisioning or memory issues that need investigation.

Rollout timeline

The change will be rolled out progressively:

• 1 June 2026: swap will no longer be enabled for new customers

• 1 September 2026: swap will no longer be enabled for new applications created by existing customers

• 1 January 2027: swap will be disabled for all application containers

Until 1 January 2027, existing applications will keep their current behavior, unless our team contacts you directly about a specific case.

What happens when an application runs out of memory?

After swap is disabled, an application that exceeds its available memory will no longer be able to temporarily rely on swap.

If this happens, the affected process, or in some cases the whole container, may be stopped with an out-of-memory error. The application will then restart according to the usual restart policy.

How to prepare for the change

We recommend reviewing memory usage over time, especially during traffic peaks, background jobs, deployments, scheduled tasks, and other workload-intensive periods. This will help you identify applications that still rely on swap as a hidden buffer and reduce the risk of unexpected OOM crashes when memory pressure occurs.

A good next step is to configure memory alerts before the change takes effect. As a starting point, you can set alerts around 80% to 90% of memory usage, depending on your application’s behavior and how quickly memory usage can increase.

You should also make sure that relevant notifiers are configured, so your team can react before an application reaches its memory limit.

Applications with significant swap usage

During the transition period, we will review significant swap usage patterns to identify applications that may require special attention.

When relevant, our team may contact affected customers directly to help them assess the situation and prepare for the change.

Need help?

If you have questions or would like help assessing the impact for your applications, please contact our support team at support@scalingo.com.