Recovery policy

Written by

Updated at July 9, 2026

When creating an instance group, you can choose how instances that fail a health check or an application health check will be recovered.

You can set the recovery policy in the auto_healing_policy key in the YAML file.

You can manage recovery policies using the CLI.

Here is how a YAML file entry may look like:

...
auto_healing_policy:
  auto_healing_action: RESTART
...

Where:

Key Value

Key	Value
`auto_healing_action`	Instance recovery action. There are three types of actions: `RESTART`: On failure, the instance will be restarted. `RECREATE`: On failure, the original instance will be deleted with a new one deployed in its place. Recreation may be helpful if the data is irrelevant, and the software needs to be redeployed. `NONE`: Disables recovery for the group's instances. Use if you need to recover the instance manually or retrieve its data for diagnostics. The default value is `RESTART`.

auto_healing_action

Instance recovery action. There are three types of actions:

RESTART: On failure, the instance will be restarted.
RECREATE: On failure, the original instance will be deleted with a new one deployed in its place.
Recreation may be helpful if the data is irrelevant, and the software needs to be redeployed.
NONE: Disables recovery for the group's instances.
Use if you need to recover the instance manually or retrieve its data for diagnostics.

The default value is RESTART.

If you do not set the recovery policy, the VM instances will be auto-healed depending on the deployment policy settings.