Skip to main content

Create an Auto Scaling group

Create an Auto Scaling group​

An Auto Scaling group dynamically manages a group of instances based on the instance specifications defined in a launch template. It controls and manages the group according to the workload demands of your application.
To create an Auto Scaling group:

  1. In the KakaoCloud console, go to Beyond Compute Service > Auto Scaling.

  2. On the Auto Scaling group menu, click [Create Auto Scaling group].

  3. Enter the required information on the Create Auto Scaling group page, and click [Create].

    ItemDescription
    Basic informationName: Name of the Auto Scaling group
    - Must be unique within the project

    Auto Scaling group description: Description of the Auto Scaling group

    Select launch template and version: Launch template and version to associate with the Auto Scaling group
    Network settingsSubnets in which the Auto Scaling group creates instances
    - If you select subnets in multiple availability zones, instances are distributed according to the availability zone distribution policy
    Load balancer connectionConnection between the Auto Scaling group and load balancer target groups
    - Do not connect a load balancer: Does not connect a target group to the Auto Scaling group
    - Connect a load balancer: Connects a target group located in a subnet selected in the network settings of the Auto Scaling group
    Availability zone distribution policyPolicy for distributing instances when subnets in multiple availability zones are selected in the network settings
    - Balanced distribution: Maintains an even number of instances across availability zones
    - Best-effort distribution: Retries instance creation in another available zone if creation fails (planned)
    Health checkHealth check settings for instances managed by the Auto Scaling group
    - Grace period: Time to defer health checks while an instance completes initialization after creation starts
    Group sizeDesired capacity: Number of instances to maintain in the Auto Scaling group
    - Enter a value from 0 to 50

    Minimum capacity: Minimum number of instances to maintain even when a scale-in operation deletes instances
    - Enter a value from 0 to 50

    Maximum capacity: Maximum number of instances allowed when a scale-out operation creates instances
    - Enter a value from 0 to 50
    Advanced settingsInstance protection from scale-in: Settings that exclude instances from deletion during a scale-in operation
    - When set to Enabled, scale-in protection is enabled for all instances created after this setting is applied
info
  • The launch template version associated with the Auto Scaling group must contain valid values for the following resources:
    • Image, instance type, volume, key pair, and network settings (security group)
  • When connecting a target group to an Auto Scaling group, you can configure a common traffic port for all target instances in that target group.
  • You can connect up to 30 target groups to one Auto Scaling group.
  • Balanced distribution is currently available as an availability zone distribution policy. Best-effort distribution will be supported in the future.
  • Instances that run complex applications might require additional time to stabilize after booting. If a health check runs before an instance has fully stabilized, the instance might be marked as Unhealthy. Set a sufficiently long health check grace period for your environment.
  • If Instance protection from scale-in is enabled, all newly created instances are excluded from deletion. When a scale-in operation is required, the Auto Scaling group repeatedly retries instance deletion because it cannot delete protected instances. Disable scale-in protection for enough instances to allow the group to scale in to the appropriate size.
caution
  • When connecting load balancer target groups, you can select only one subnet, and all target groups must be in that subnet.

Create a scaling policy​

KakaoCloud Auto Scaling supports scheduled scaling policies that run a scaling operation once or repeatedly at specified times for managed instances.

Create a scheduled scaling policy​

A scheduled scaling policy runs a scaling operation once or repeatedly at a specified time.
It is suitable for applications with predictable patterns in which load temporarily increases and then decreases at regular intervals, helping optimize cost and performance for your environment.
To create a scheduled scaling policy:

  1. In the KakaoCloud console, go to Beyond Compute Service > Auto Scaling.

  2. On the Auto Scaling group menu, click [Configure scheduled scaling policy]. Alternatively, on the Details > Scaling policy tab, select the Scheduled scaling policy tab and click [Create scheduled scaling policy].

  3. Enter the required information on the Create scheduled scaling policy page, and click [Create].

    ItemDescription
    Policy nameName of the scheduled scaling policy
    Desired capacityNumber of instances to maintain when the scheduled scaling policy runs
    Minimum capacityMinimum size of the Auto Scaling group when the scheduled scaling policy runs
    Maximum capacityMaximum size of the Auto Scaling group when the scheduled scaling policy runs
    RecurrenceWhether to run the scheduled scaling policy once or repeatedly
    - Recurring: Runs the policy at the selected interval, beginning at the specified start date and time
    - Once: Runs the policy once at the specified start date and time, and then deletes the policy
    Start date and timeDate and time when the scheduled scaling policy runs
    End date and timeAvailable when Recurring is selected. Specify when recurring operations from the scheduled scaling policy must stop
info
  • The recurrence information for a scheduled scaling policy with Recurring selected is stored as a Cron expression.
  • Each time a recurring scheduled scaling policy runs, its start date and time are updated by adding the recurrence interval.
  • A scheduled scaling policy with Recurring selected runs first at the specified start date and time, and then at each configured interval.