Review cluster quotas, capacity limits, and cloud service quotas for Container Service for Kubernetes.
Limitations
Product and cluster configuration limits
The following limits apply to Container Service for Kubernetes.
Limit Item | Description |
Account verification | Before creating an ACK cluster, complete identity verification and activate Container Service for Kubernetes. |
Account balance | When creating an ACK cluster, your account balance and cash coupon total must be at least 100 CNY to create pay-as-you-go Alibaba Cloud resources. If you plan to create an ACK managed cluster or an ACK dedicated cluster, and on the Cluster Configurations page, set the Billing Method to Subscription while specifying at least one ECS instance, ensure your account balance is sufficient to cover the subscription ECS instance fees and other product fees. If your balance is insufficient, the number of nodes created will be less than expected, and will be restored after renewal. |
Cluster configuration | After creating an ACK cluster, you cannot change these configurations.
|
ECS instance (node instance) configuration |
|
Cluster control plane component access traffic | Accessing control plane components (API Server, etcd) via API or CLI may trigger rate limiting when reading many events at once. We recommend using Event Center to query cluster events, or adding pagination parameters to reduce the size of a single request (e.g., If you frequently hit rate limits in an ACK Basic cluster, migrate to an ACK Pro cluster. See Hot migrate an ACK Basic cluster to an ACK Pro cluster. |
Single cluster capacity limits
ACK managed cluster Pro Edition and ACK managed cluster Basic Edition differ in scenarios and supported capacity.
ACK managed cluster Pro Edition: Recommended for enterprise production environments.
In the ACK managed cluster Pro Edition, when the Terway network plug-in is configured and IPvlan is enabled, the maximum number of pods is 50,000, and the total number of mappings between Services and pods is limited to 64,000 for a cluster with 5,000 nodes.
ACK managed cluster Basic Edition: For personal testing or learning only.
The following table lists the maximum capacity per resource type in a single cluster.
Type | ACK managed cluster Pro Edition | ACK managed cluster Basic Edition |
etcd storage capacity | 8 GB | 2 GB |
Total size of etcd objects for each resource type | 800 MB | 200 MB |
Node |
| 10 |
Pod | 500,000 Depends on the network plugin and available Pod CIDR block configuration | 300 |
Namespace | 10,000 | 100 |
ConfigMap | 30,000 | 300 |
Secret | 100,000 | 1,000 |
PVC | 100,000 | 1,000 |
PV | 100,000 | 1,000 |
Service | 10,000 | 100 |
Role | 50,000 | 500 |
RoleBinding | 50,000 | 500 |
CRD | 100,000 | 1,000 |
Number of CR resources per CRD type | 100,000 | 1,000 |
Quotas
Cluster quotas
The following table lists default quotas. To view adjustable quotas and request increases, go to Quota Center. Quota Center supports multiple cloud products. See Cloud products supported by Quota Center.
To increase the maximum node pools per cluster, also increase the total scaling groups quota for Auto Scaling (ESS). Log on to the Quota Platform to submit a request.
The maximum pods per node depends on the cluster network plugin.
Flannel: Determined by CIDR block planning at cluster creation. See the following table. Not adjustable.
Terway: Depends on the elastic network interfaces (ENIs) of the ECS instance type. Use higher-spec, newer instance types.
ACK managed cluster
Dimension | Basic Edition | Pro Edition |
Maximum number of clusters per Alibaba Cloud account | 2 | 100 |
Maximum number of node pools per cluster | 10 | 100 |
Maximum number of nodes per cluster | 10 |
|
Maximum number of serverless pods per cluster | 1,000 | 50,000 Important If many pods are associated with services, keep the number below 20,000. |
Maximum Pods per single node |
| |
Quota increase method | Cannot be requested | log on to the Quota Center console and submit an application |
ACK dedicated cluster
Maximum number of clusters per Alibaba Cloud account: 0 (Creation stopped)
Maximum node pools per cluster: 100
Maximum number of nodes per cluster:
Using Flannel container network plugin: 200 by default, 1,000 maximum
Using Terway container network plugin: 5,000 by default, 50,000 maximum
Maximum number of Pods per node:
Using Flannel container network plugin: 256
Using Terway: The pod limit per node depends on node specifications. See Calculate Pod Limit per Node.
Quota increase method:
log on to the Quota Center console and submit an application
ACK serverless cluster
Dimension | Basic Edition | Pro Edition |
Maximum number of clusters per Alibaba Cloud account | 2 | 100 |
Maximum number of pods per cluster | 1,000 | 50,000 Important If many pods are associated with services, keep the number below 20,000. |
Quota increase method | log on to the Quota Center console and submit an application | |
ACK Edge Cluster
Dimension | Basic Edition | Pro Edition |
Maximum number of clusters per Alibaba Cloud account | 2 | 100 |
Maximum number of node pools per cluster | 10 | 100 |
Maximum number of nodes per cluster | 10 | 1,000 |
Maximum Pods per node | 256 | 256 |
Quota increase method | Cannot be requested | log on to the Quota Center console and submit an application |
Registered clusters
Maximum number of clusters per Alibaba Cloud account: 5
Maximum node pools per cluster: 100
Maximum Pods per node: 256
Quota increase method:
log on to the Quota Center console and submit an application
Underlying Cloud Product Quotas
Product Type | Limit Item | Default Limit | Quota Increase Method |
Alibaba Cloud Resource Orchestration Service (ROS) quota | 100 | ||
On-demand instance vCPU limit | 500 cores | ||
Purchase high-specification on-demand instances (instances with more than 16 cores) | Instance types with fewer than 16 vCPU cores (excluding 16) | ||
Preemptible instance vCPU limit | 800 cores | ||
Convert pay-as-you-go to subscription | Unsupported instance types (families): t1, s1, s2, s3, c1, c2, m1, m2, n1, n2, e3 | ||
Maximum number of ECS instances in a single scaling group for ESS | 2,000 | log on to the Quota Center console and submit an application | |
Operating system | Nodes that run the following operating systems can be added to an ACK cluster:
| None | |
Applies only to clusters with the Flannel network plug-in The maximum number of custom route entries that you can create in a route table (excluding dynamically propagated route entries). | 200 | See General Quotas. | |
The maximum number of vSwitches that you can create in a VPC. | 150 | ||
The maximum number of VPCs that you can create in a region. | 10 | ||
Maximum private IP addresses per VPC security group per account per region |
| For basic security groups, apply on the Quota Center. | |
Maximum security groups per account per region | Quota ID: | ||
Maximum ENIs (secondary ENIs) per account per region | Quota ID: | ||
Pay-as-you-go EIPs per account (excludes subscription EIPs) | 20 | Increase the quota on the Quota Management page or Quota Center. | |
Maximum number of CLB instances that can be created by each Alibaba Cloud account | 60 | Use one of these methods:
| |
Maximum number of backend servers that can be added to a CLB instance | 200 | Use one of these methods:
| |
Maximum number of listeners that can be added to a CLB instance | 50 | Use one of these methods:
| |
Maximum number of times that a server can be added as a CLB backend server | 50 | None | |
Quota for pay-as-you-go cloud disks for an account across all regions | Number of instances across all regions under the account × 5. Each account can create a minimum of 10 pay-as-you-go cloud disks. | ||
Quota for pay-as-you-go cloud disk capacity used as data disks for an account | Depends on ECS usage, region, and disk type. View on the privilege quota page. See Elastic Block Storage. | ||
References
Each cluster type has specific precautions and risky operations. See Usage notes and important risky operations.
Follow Recommended Workload Configurations for stable and reliable cluster applications.
Adding nodes increases Pod capacity, but oversized clusters may degrade availability and performance. See Recommendations for large-scale clusters.