Feature Request: Support `max_unavailable` in `upgrade_settings` for AKS Node Pools

[aydosman]

Is there an existing issue for this?

• I have searched the existing issues

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• Please do not leave comments along the lines of "+1", "me too" or "any updates", they generate extra noise for issue followers and do not help prioritize the request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment and review the contribution guide to help.

Description

AKS supports a maxUnavailable setting that limits how many nodes (number or percentage) can be unavailable during a node-pool upgrade. When used the managed service cordons and drains nodes up to this threshold without creating surge nodes. This is valuable for clusters with constrained IP capacity.

Today the provider exposes upgrade_settings.max_surge (and timers like drain/node soak) but does not expose max_unavailable. Please add max_unavailable to the upgrade_settings.

New or Affected Resource(s)/Data Source(s)

• azurerm_kubernetes_cluster
• azurerm_kubernetes_cluster_node_pool

Potential Terraform Configuration

resource "azurerm_kubernetes_cluster_node_pool" "example" {
  name                  = "np00001"
  kubernetes_cluster_id = azurerm_kubernetes_cluster.example.id
  vm_size               = "Standard_DS2_v2"
  node_count            = 3

  upgrade_settings {
    max_surge       = "0"
    max_unavailable = "30%"
    
    drain_timeout_in_minutes      = 30
    node_soak_duration_in_minutes = 5
  }
}

References

• AKS docs – maxUnavailable

[sojournerdev]

I don’t mind supporting this, but my main limitation right now is around bumping the go-azure-sdk version.

The business logic for wiring these fields into the provider looks straightforward, but the SDK upgrade would be a bigger change since a lot of services still reference the existing 2025-02-01 packages.

From the SDK definitions:

Current:
2025-02-01:

type AgentPoolUpgradeSettings struct {
	DrainTimeoutInMinutes     *int64  `json:"drainTimeoutInMinutes,omitempty"`
	MaxSurge                  *string `json:"maxSurge,omitempty"`
	NodeSoakDurationInMinutes *int64  `json:"nodeSoakDurationInMinutes,omitempty"`
}

New:
2025-05-01

type AgentPoolUpgradeSettings struct {
	DrainTimeoutInMinutes     *int64                   `json:"drainTimeoutInMinutes,omitempty"`
	MaxSurge                  *string                  `json:"maxSurge,omitempty"`
	MaxUnavailable            *string                  `json:"maxUnavailable,omitempty"`
	NodeSoakDurationInMinutes *int64                   `json:"nodeSoakDurationInMinutes,omitempty"`
	UndrainableNodeBehavior   *UndrainableNodeBehavior `json:"undrainableNodeBehavior,omitempty"`
}

[alexp-openai]

Any updates on this change? looks like the dependent one has been merged.

[sojournerdev]

Hey @alexp-openai

Just saw the PR get merged for the dependency. I’ve been sick this whole week, but can get started on it next week. How does that sound?

[sojournerdev]

@alexp-openai @aydosman

I was about to sit down tonight and code this, but it looks like someone beat me to it. If that works for you, I believe this should close it out.

#30515

[aydosman]

Seems like it was merged then reverted #30554 but I don't see anything in the latest release v4.45.0

[aydosman]

@3mbe - Any update on this feature? we are eagerly waiting.

[aydosman]

Looks like this was implemented in #30563 - closing.

[Hazmi35]

The default_node_pool.upgrade_settings does not have these properties?

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.