Google Kubernetes Engine v1 API - Class NodePool.Types.UpgradeSettings (3.10.0)

public sealed class NodePool.Types.UpgradeSettings : IMessage<NodePool.Types.UpgradeSettings>, IEquatable<NodePool.Types.UpgradeSettings>, IDeepCloneable<NodePool.Types.UpgradeSettings>, IBufferMessage, IMessage

Reference documentation and code samples for the Google Kubernetes Engine v1 API class NodePool.Types.UpgradeSettings.

These upgrade settings control the level of parallelism and the level of disruption caused by an upgrade.

maxUnavailable controls the number of nodes that can be simultaneously unavailable.

maxSurge controls the number of additional nodes that can be added to the node pool temporarily for the time of the upgrade to increase the number of available nodes.

(maxUnavailable + maxSurge) determines the level of parallelism (how many nodes are being upgraded at the same time).

Note: upgrades inevitably introduce some disruption since workloads need to be moved from old nodes to new, upgraded ones. Even if maxUnavailable=0, this holds true. (Disruption stays within the limits of PodDisruptionBudget, if it is configured.)

Consider a hypothetical node pool with 5 nodes having maxSurge=2, maxUnavailable=1. This means the upgrade process upgrades 3 nodes simultaneously. It creates 2 additional (upgraded) nodes, then it brings down 3 old (not yet upgraded) nodes at the same time. This ensures that there are always at least 4 nodes available.

These upgrade settings configure the upgrade strategy for the node pool. Use strategy to switch between the strategies applied to the node pool.

If the strategy is ROLLING, use max_surge and max_unavailable to control the level of parallelism and the level of disruption caused by upgrade.

  1. maxSurge controls the number of additional nodes that can be added to the node pool temporarily for the time of the upgrade to increase the number of available nodes.
  2. maxUnavailable controls the number of nodes that can be simultaneously unavailable.
  3. (maxUnavailable + maxSurge) determines the level of parallelism (how many nodes are being upgraded at the same time).

If the strategy is BLUE_GREEN, use blue_green_settings to configure the blue-green upgrade related settings.

  1. standard_rollout_policy is the default policy. The policy is used to control the way blue pool gets drained. The draining is executed in the batch mode. The batch size could be specified as either percentage of the node pool size or the number of nodes. batch_soak_duration is the soak time after each batch gets drained.
  2. node_pool_soak_duration is the soak time after all blue nodes are drained. After this period, the blue pool nodes will be deleted.

Inheritance

object > NodePool.Types.UpgradeSettings

Namespace

GoogleGoogle.CloudGoogle.Cloud.ContainerV1

Assembly

Google.Cloud.Container.V1.dll

Constructors

UpgradeSettings()

public UpgradeSettings()

UpgradeSettings(UpgradeSettings)

public UpgradeSettings(NodePool.Types.UpgradeSettings other)
Parameter
NameDescription
otherNodePoolTypesUpgradeSettings

Properties

BlueGreenSettings

public BlueGreenSettings BlueGreenSettings { get; set; }

Settings for blue-green upgrade strategy.

Property Value
TypeDescription
BlueGreenSettings

HasStrategy

public bool HasStrategy { get; }

Gets whether the "strategy" field is set

Property Value
TypeDescription
bool

MaxSurge

public int MaxSurge { get; set; }

The maximum number of nodes that can be created beyond the current size of the node pool during the upgrade process.

Property Value
TypeDescription
int

MaxUnavailable

public int MaxUnavailable { get; set; }

The maximum number of nodes that can be simultaneously unavailable during the upgrade process. A node is considered available if its status is Ready.

Property Value
TypeDescription
int

Strategy

public NodePoolUpdateStrategy Strategy { get; set; }

Update strategy of the node pool.

Property Value
TypeDescription
NodePoolUpdateStrategy