Copy page

Tutorials

Scale To Zero Triggers

Unikraft Cloud supports scaling instances to zero based on different triggers. This allows you to optimize resource usage and costs by automatically setting instances to standby when they're not needed. They will automatically resume when required with minimal delay. The platform can scale down instances automatically, or you can scale them down manually. You can read more about Scale To Zero here.

Automatic triggers

Automatic triggers are the most common and easiest way to use scale-to-zero with your instances on Unikraft Cloud. By default, instances scale down automatically when there is no traffic coming to them for a specified period. You can also configure scale-to-zero to consider idle active connections when determining whether to scale down an instance.

No traffic

When you enable this trigger, Unikraft Cloud will track incoming connections to your instance. If there is no traffic for the configured duration, the instance scales down to zero. You can use the CLI to enable this trigger:

Code
unikraft run \
  --scale-to-zero policy=on,cooldown-time=5000 \
  ...

Idle connections

When enabling this trigger, the platform will consider all active connections to your instance, so all traffic going to them. The instance will scale down if you make no new network connections and all existing connections have been idle for the configured duration.

The idle policy requires stateful scale-to-zero, so set stateful=true together with it. You can enable this trigger using the CLI as follows:

Code
unikraft run \
  --scale-to-zero policy=idle,stateful=true,cooldown-time=5000 \
  ...

Manual triggers

For specific use cases, you might want to manually scale your instances to zero, or to manually disable scale-to-zero. Unikraft Cloud supports two ways to manually trigger scale-to-zero.

API

You can use the Unikraft Cloud API to manually scale down an instance. You will have to use the /instances/suspend endpoint to scale down an instance. This happens instantly, regardless of the configured automatic triggers.

First, set the required environment variables:

Code
# Set Unikraft Cloud access token
export UKC_TOKEN=token
# Set metro to Frankfurt, DE
export UKC_METRO=fra

Then invoke the following curl command to scale down an instance named my-instance:

PUT /instances/suspend
curl -X PUT https://api.fra.unikraft.cloud/v1/instances/suspend \
  -H "Authorization: Bearer ${UKC_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-instance",
    "drain_timeout_ms": 10000
  }'

You can do the same using the unikraft CLI:

Code
unikraft instance suspend my-instance --drain-timeout 10s

The drain timeout allows you to specify a grace period for existing connections to finish before the instance goes to standby. If set to 0, no grace period exists and the instance goes to standby immediately, dropping all existing connections. If set to -1 or unset, the platform will wait indefinitely for existing connections to finish before suspending the instance.

Device file

You can also manually scale down an instance by writing '+'/'-' to a special device file available inside the instance. Every instance has this file at /uk/libukp/scale_to_zero_disable. By writing '+' to this file, you disable automatic scale-to-zero for the instance until the number of '-' written matches the number of '+' written. For example, to disable automatic scale-to-zero, run the following command inside the instance:

Code
echo '+' > /uk/libukp/scale_to_zero_disable

Learn more

• The Scale To Zero feature page.
• The CLI reference.