Copy page

Features

Autokill

Autokill automatically deletes instances, service groups, templates, and checkpoints when a configured condition triggers. It stops workloads from piling up after they finish and are no longer needed.

You can configure autokill on:

• Instances: killed after a time, or after reaching a request limit.
• Instance templates: killed if nobody clones them within a specified time.
• Service groups: killed after the group has been empty for a specified time.
• Checkpoints: killed if nobody restores them within a specified time.

Set autokill with the --autokill flag on the create and edit commands of each resource.

Instance autokill

Kill after stopped duration

Automatically kill an instance after it stops and remains stopped for a given time:

Code
unikraft instance create --metro fra \
  --autokill time=1h ...

This example kills the instance 1 hour after it stops.

Kill after a request count

Kill an instance after it serves a specified request count:

Code
unikraft instance create --metro fra \
  --autokill num-requests=1000 ...

Updating autokill on a stopped instance

Code
unikraft instance edit <instance-name> --autokill time=10s

The new value replaces the whole policy. To keep a request limit, pass it again, for example --autokill time=10s,num-requests=1000.

Service group autokill

Automatically kill a service group after it stays empty (no instances) for a given time:

Code
unikraft service create --metro fra \
  --autokill time=5m ...

This example kills the service group 5 minutes after its last instance leaves.

Instance template autokill

Automatically kill an instance template if nobody clones it within a given time:

Code
unikraft instance template create --metro fra \
  --autokill time=24h ...

This example removes the template after 24 hours without a clone.

Checkpoint autokill

Automatically kill a checkpoint if nobody restores it within a given time:

Code
unikraft instance checkpoint create my-instance \
  --autokill time=1h

This example removes the checkpoint after 1 hour without a restore.

Changing autokill

Use the edit command of the resource to set or change its autokill policy:

Code
unikraft service edit <service-name> --autokill time=10m
unikraft instance template edit <template-name> --autokill time=48h
unikraft instance checkpoint edit <checkpoint-name> --autokill time=2h

Listing and filtering

The autokill.time and autokill.num-requests fields work like any other field. You can list, filter, sort, and inspect them:

Code
unikraft instance list --filter 'autokill.time>0s' --sort autokill.time

Learn more

• Unikraft Cloud's REST API reference, in particular the sections on instances and services.