Class: Aws::SageMaker::Types::ClusterDrainPolicy

Inherits:

Struct

Object
Struct
Aws::SageMaker::Types::ClusterDrainPolicy

show all

Defined in:: gems/aws-sdk-sagemaker/lib/aws-sdk-sagemaker/types.rb

Overview

Specifies how pods are evicted from an instance before it is patched with the DrainFirst strategy. By default, evictions respect Kubernetes Pod Disruption Budgets.

Constant Summary collapse

SENSITIVE =

[]

Instance Attribute Summary collapse

#drain_timeout_in_seconds ⇒ Integer
The maximum time, in seconds, to wait for the pods on an instance to drain.
#on_drain_timeout ⇒ String
The action to take if the pods on an instance don't drain within DrainTimeoutInSeconds.

Instance Attribute Details

#drain_timeout_in_seconds ⇒ `Integer`

The maximum time, in seconds, to wait for the pods on an instance to drain. Valid values are from 0 to 5400 (90 minutes). The default is 3600 (1 hour).

Returns:

(Integer)

# File 'gems/aws-sdk-sagemaker/lib/aws-sdk-sagemaker/types.rb', line 6559

class ClusterDrainPolicy < Struct.new(
  :drain_timeout_in_seconds,
  :on_drain_timeout)
  SENSITIVE = []
  include Aws::Structure
end

#on_drain_timeout ⇒ `String`

The action to take if the pods on an instance don't drain within DrainTimeoutInSeconds. The default is Skip.

Skip: Doesn't patch the instance. The instance is uncordoned and remains on the previous AMI version. This can leave the instance group partially patched. You can then use UpdateClusterSoftware with the desired ImageReleaseVersion to manually update the remaining instances.
Force: Deletes the remaining pods without respecting Pod Disruption Budgets, and then patches the instance.