ACK clusters move through a defined set of states across three phases: creation, operation, and deletion. This topic explains what each state means, what triggers abnormal states, and how to resolve them.
Cluster lifecycle
The following diagram shows the states an ACK cluster moves through and how transitions happen between them.
ACK regularly checks the running status of clusters. If a cluster meets specific criteria for an abnormal state, its status automatically changes to inactive or unavailable and ACK sends you a notification by email and internal message.
The table below lists all cluster states, their meanings, and whether cluster management fees apply for ACK Managed Cluster Pro Edition.
| Phase | Cluster state | Description | Cluster management fee (Pro Edition) |
|---|---|---|---|
| Creation and deployment | Initializing (initial) |
The cluster is being created. | Not charged |
| Creation and deployment | Creation failed (failed) |
The cluster failed to be created. | Not charged |
| Operation and maintenance | Running (running) |
The cluster is running. | Charged |
| Operation and maintenance | Upgrading (upgrading) |
The cluster is being upgraded. | Charged |
| Operation and maintenance | Draining node (draining) |
Pods on the node are being evicted and rebuilt on other nodes. After the process completes, the node becomes unschedulable. | Charged |
| Operation and maintenance | Removing node (removing) |
Nodes are being removed from the cluster. | Charged |
| Operation and maintenance | Updating configuration (updating) |
The cluster metadata is being updated. | Charged |
| Operation and maintenance | Inactive (inactive) |
The cluster is temporarily unavailable due to abnormal conditions. See Inactive (inactive). | Not charged |
| Operation and maintenance | Unavailable (unavailable) |
The cluster is no longer usable because its underlying cloud resources are abnormal. See Unavailable (unavailable). | Not charged |
| Deletion and release | Deleting (deleting) |
The cluster is being deleted. | Not charged |
| Deletion and release | Deletion failed (delete_failed) |
The cluster failed to be deleted. | Not charged |
| Deletion and release | Deleted (deleted) |
The cluster has been deleted and is no longer visible. | Not charged |
For billing details, see Billing overview. Billing policies for associated cloud resources take precedence. See Cloud resource fees.
Abnormal cluster states
Inactive (inactive)
A cluster enters the inactive state when specific abnormal conditions are detected. The status code identifies the cause and whether you can resolve it yourself.
| Status code | Cause | Solution |
|---|---|---|
KMSUnhealthy |
Encryption at rest for Secrets is enabled using Key Management Service (KMS). The KMS service is suspended due to an overdue payment or other reasons, causing the control plane to stop running. | 1. Log on to the Key Management Service console. 2. Identify the reason for the suspension and resume the service. 3. Submit a ticket to restore the cluster status. |
NoNodeForLongTime |
No nodes have existed in an ACK Managed Cluster Basic Edition for 14 consecutive days. | Submit a ticket to restore the cluster status. After restoration, upgrade to ACK Managed Cluster Pro Edition. |
AssumeRoleNotFound |
The system cannot find the service role for Container Service for Kubernetes (ACK), causing the control plane to become abnormal. | 1. Check the roles required for ACK in Service roles of ACK. 2. Submit a ticket to restore the cluster status. |
AssumeUserNotFound |
The system cannot find the Resource Access Management (RAM) user associated with ACK, causing the control plane to become abnormal. | Submit a ticket for technical support. |
SecurityGroupNotFound |
The system cannot find the security group for ACK, causing the control plane to become abnormal. | Submit a ticket for technical support. |
UnderMaintenance |
The control plane is under background maintenance. | Submit a ticket for technical support. |
ServiceInDebt |
Your available credit (account balance plus cash coupons) is less than the outstanding bill amount. Your ACK Managed Cluster Pro Edition enters the inactive state, and all operations that require API server access fail. If the payment remains overdue for more than 15 days, ACK suspends the service and deletes the cluster's control plane resources. Associated cloud resources — NAT Gateways, SLB instances, ECS instances, and ESS scaling groups — are not automatically released but may behave unexpectedly. | Top up your account and pay the overdue bill. The cluster automatically returns to a normal state after payment. |
Unavailable (unavailable)
A cluster enters the unavailable state when its underlying cloud resources are released. Clusters in this state cannot be recovered.
| Cause | Affected resource | Recoverable | Solution |
|---|---|---|---|
| The Classic Load Balancer (CLB) instance for the cluster's API server is released. This occurs when the instance is manually released, a subscription instance expires and is automatically released, or a pay-as-you-go instance is released due to an overdue payment. Note
Starting December 1, 2024, newly created CLB instances incur instance fees. See Adjustment announcement for Classic Load Balancer CLB billing items. |
CLB instance | No | Delete the cluster and create a new one. See Delete a cluster and Create an ACK managed cluster. |
Impact of abnormal states
When a cluster is in the inactive or unavailable state, the following impacts apply.
Billing
The control plane is scaled in when a cluster enters an abnormal state. After the scale-in, cluster management fees are no longer charged. Fees for associated cloud resources continue to be charged.
Operation restrictions
Only the following operations are available on a cluster in the inactive or unavailable state:
-
Change the deletion protection status of the cluster
-
Delete the cluster
Scaling group behavior
ACK disables the associated scaling groups to prevent new ECS instances from being created and incurring unexpected fees. After the cluster returns to a normal state, if the scaling groups are still disabled, manually enable them in the Auto Scaling console.