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 SURGE, use maxSurge and maxUnavailable 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 blueGreenSettings to configure the blue-green upgrade related settings. 1. standardRolloutPolicy 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. batchSoakDuration is the soak time after each batch gets drained. 2. nodePoolSoakDuration is the soak time after all blue nodes are drained. After this period, the blue pool nodes will be deleted.

JSON representation
{
  "maxSurge": integer,
  "maxUnavailable": integer,
  "strategy": enum (NodePoolUpdateStrategy),
  "blueGreenSettings": {
    object (BlueGreenSettings)
  }
}
Fields
maxSurge

integer

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

maxUnavailable

integer

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

strategy

enum (NodePoolUpdateStrategy)

Update strategy of the node pool.

blueGreenSettings

object (BlueGreenSettings)

Settings for blue-green upgrade strategy.

NodePoolUpdateStrategy

Strategy used for node pool update.

Enums
NODE_POOL_UPDATE_STRATEGY_UNSPECIFIED Default value if unset. GKE internally defaults the update strategy to SURGE for unspecified strategies.
BLUE_GREEN blue-green upgrade.
SURGE SURGE is the traditional way of upgrading a node pool. maxSurge and maxUnavailable determines the level of upgrade parallelism.

BlueGreenSettings

Settings for blue-green upgrade.

JSON representation
{

  // Union field rollout_policy can be only one of the following:
  "standardRolloutPolicy": {
    object (StandardRolloutPolicy)
  },
  "autoscaledRolloutPolicy": {
    object (AutoscaledRolloutPolicy)
  }
  // End of list of possible types for union field rollout_policy.
  "nodePoolSoakDuration": string
}
Fields
Union field rollout_policy. The rollout policy controls the general rollout progress of blue-green. rollout_policy can be only one of the following:
standardRolloutPolicy

object (StandardRolloutPolicy)

Standard policy for the blue-green upgrade.

autoscaledRolloutPolicy

object (AutoscaledRolloutPolicy)

Autoscaled policy for cluster autoscaler enabled blue-green upgrade.

nodePoolSoakDuration

string (Duration format)

Time needed after draining entire blue pool. After this period, blue pool will be cleaned up.

A duration in seconds with up to nine fractional digits, ending with 's'. Example: "3.5s".

StandardRolloutPolicy

Standard rollout policy is the default policy for blue-green.

JSON representation
{

  // Union field update_batch_size can be only one of the following:
  "batchPercentage": number,
  "batchNodeCount": integer
  // End of list of possible types for union field update_batch_size.
  "batchSoakDuration": string
}
Fields
Union field update_batch_size. Blue pool size to drain in a batch. update_batch_size can be only one of the following:
batchPercentage

number

Percentage of the blue pool nodes to drain in a batch. The range of this field should be (0.0, 1.0].

batchNodeCount

integer

Number of blue nodes to drain in a batch.

batchSoakDuration

string (Duration format)

Soak time after each batch gets drained. Default to zero.

A duration in seconds with up to nine fractional digits, ending with 's'. Example: "3.5s".

AutoscaledRolloutPolicy

This type has no fields.

Autoscaled rollout policy uses cluster autoscaler during blue-green upgrades to scale both the green and blue pools.