Configure
Sizes and limits
The sizes#
| Size | CPU | Memory | Temporary disk |
|---|---|---|---|
| Starter | 0.5 | 512 MB | 2 GB |
| Starter Plus | 0.5 | 1 GB | 2 GB |
| Standard | 1 | 2 GB | 5 GB |
| Pro | 2 | 4 GB | 10 GB |
| Pro Plus | 4 | 8 GB | 20 GB |
| Pro Max | 4 | 16 GB | 40 GB |
- New services start at Standard; static sites at Starter. You choose on the New screen.
- Copies multiply it. A Standard service with 3 copies uses 3 × 2 GB. Each copy runs on a server of its own.
- A Container App has one size for the whole app, shared by its containers. A container's own limit in your Compose file is kept if they all fit; the rest is shared by how much each container actually uses. If the file's limits add up to more than the size, the deploy stops before anything changes and says so.
- Your workspace has a total: the memory all its services may use together. Sizes that would go past it are greyed out. Ask us to raise it.
What happens at the limit#
| When the app tries to use more… | What happens |
|---|---|
| Memory than its size | The process is stopped and the app restarted. Only this app is affected. You get an email: “… ran out of memory”. |
| CPU than its size | It's slowed down, never stopped. |
| Temporary disk than its size | Nothing is stopped. You get an email so you can clean up. |
Temporary disk is the space inside the container: files your app writes outside a volume. It's cleared on every deploy. Keep logs on the console, and anything that must last in a volume.
“Reserved now; the limit is switched on later”
Seeing what it uses#
The service's Metrics page shows its size, its peak memory against it, how many times it ran out of memory in the last 24 hours, and its temporary disk against its allowance. If peak memory keeps getting close to the size, choose a bigger one.
Changing the size#
Open Compute in the workspace sidebar to see every project and its services, with total reserved memory and your workspace limit. Each service has its own size control. Owners and Admins can use custom sliders where enabled; fixed tiers keep the same permissions as service Settings. Viewers see sizes read-only. GPUs are managed separately. Confirm the effect before saving. Save only keeps the running app unchanged; Save and redeploy saves the size and runs the current version again with the latest settings. Redeploying restarts the app, so it can be briefly unavailable. Container App containers may restart. A smaller size can run out of memory if the app uses more than the new limit. Redeploy also needs deploy permission. Stopped services and services never deployed offer Save only. If redeploy cannot start after saving, the row says the size is saved and gives the reason. A queued redeploy links to its progress. For older services with limits off, the size is reserved but redeploy does not enable tier limits; existing explicit limits remain.
Open the service's Settings, then Size. Choose another size and press Change size. It applies from the service's next deploy: press Redeploy to apply it now. A smaller size is always allowed; a bigger one must fit your workspace's total and a server.
Custom sizes (where switched on)
In workspaces where we've switched on custom sizes, the Size card has sliders instead: drag to any CPU, memory and storage that fits, then press Apply. The standard sizes are marked along each bar, and each bar stops at what its server has room for.
| Slider | Steps | What it changes |
|---|---|---|
| CPU | 0.25, up to 8 | The CPU the app may use. |
| Memory | 256 MB, from 512 MB to 16 GB | The memory the app may use. Can't go below what it's using right now. |
| Temporary disk | 1 GB | The allowance you're emailed about. Not a hard limit. |
| Storage | 1 GB | The allowance for its volumes together, with emails at 80% and 90% full. Only grows, so data is never cut off. |
- Nothing changes until you press Apply, and Apply shows you a summary first.
- CPU and memory change on the running app, without a restart, once its limits are switched on. For an older service whose limits were off, the size is saved for its next deploy. If its server cannot apply a live change, the result says that it is saved for the next deploy instead. A deploy is a separate action and may restart containers; saving here never starts one.
- Temporary disk and storage are watched, not capped: nothing inside the container changes; you get an email when it's over.
- Choosing a standard size again replaces a custom one.
- Only Owners and Admins can use the sliders.
Data that must last#
The size doesn't include storage for your data. Data that must survive a deploy belongs in a volume: a named volume in a Container App's Compose file (for example a database's /var/lib/postgresql/data). Volumes are backed up every night.