Limits and resources

Limits and resources

Resources allocated per service

These caps cover the vast majority of use cases. If your application needs more (heavy processing, large build), contact support for an adjustment.

Storage

• Application services: the filesystem is ephemeral. Anything written to disk disappears when the pod restarts or redeploys. Never store data there that needs to persist.
• Databases: persistent storage via a Kubernetes volume (PVC). Data survives restarts, updates, and migrations.

Databases

A database service gets a larger allocation than an application service, and its storage is persistent:

Managed databases and your plan quota

There are two ways to get a database, with two counting rules:

• Managed database (MySQL or PostgreSQL created from the Databases page): it runs on the platform's shared database servers and is not counted against your plan's vCPU or memory. It has its own limits: with a paid subscription, up to 10 databases of 10 GB with 50 concurrent connections each; without a subscription, one 500 MB database with 10 connections.
• Database run as a service in a project (MySQL, MariaDB, PostgreSQL, MongoDB or Redis container): it is a service like any other. It uses your plan's vCPU, memory and storage, within the limits of the table above.

A plan's resources (vCPU, memory, storage, traffic) are one envelope shared by the services it covers: a plan with unlimited services lets you create as many as you need, within that same envelope.

Build

• maximum build time: 1 hour. Past that, the build is stopped and the deployment marked as failed. A build that reaches this limit is almost always stuck (a development server started instead of the build command, a script waiting for input);
• recommended image size: under 2 GB, for fast deployments;
• builds do not consume your services' resources: they run in a separate environment with their own allocation.

Network

• each service is exposed over HTTPS on its domain(s);
• no artificial limit on the number of simultaneous connections;
• available bandwidth depends on the subscribed plan.

Availability

The infrastructure targets 99.5% availability, with no contractual commitment. Scheduled maintenance is announced in advance on the status page.