Kubernetes is an open source orchestration platform that is commonly used to manage
containerized applications and services. This topic describes the features, types,
and limits of ACK clusters.
Background information
Alibaba Cloud provides different types of ACK clusters to meet the requirements of
diverse scenarios.
ACK clusters are most commonly used and suitable for most scenarios.
Serverless Kubernetes (ASK) clusters are suitable for handling agile workloads that
require quick scaling, and processing individual tasks or multiple parallel tasks.
For more information, see ASK overview.
ACK edge clusters are the most suitable option when you want to handle edge computing
services such as Internet of Things (IoT) and Content Delivery Network (CDN). For
more information, see ACK@Edge overview.
ACK also provides highly integrated solutions for sectors such as genomics computing
and AI-empowered big data computing. ACK optimizes container performance based on
the high-performance computing and networking capabilities of Infrastructure-as-a-Service
(IaaS). ACK allows you to centrally manage clusters that are deployed in multi-cloud
or hybrid cloud environments. You can log on to the ACK console to manage your Kubernetes
clusters deployed in data centers or third-party clouds.
Limits
For more information about the limits of ACK clusters, see Limits.
Cluster lifecycle
The following table describes the states of a cluster and the following figure shows
the transitions between the states.
Table 1. Cluster states
State
Description
Initializing
Creating the cloud resources that are used by the cluster.
Creation Failed
Failed to create the cloud resources that are used by the cluster.
Running
The cloud resources used by the cluster are created.
Updating
Updating the metadata of the cluster.
Scaling
Adding nodes to the cluster.
Removing
Removing nodes from the cluster.
Upgrading
Upgrading the cluster.
Draining
Evicting pods from a node to other nodes. After all pods are evicted from the node,
the node becomes unschudulable.
Deleting
Deleting the cluster.
Deletion Failed
Failed to delete the cluster.
Deleted (invisible to users)
The cluster is deleted.
Figure 1. State transitions
Cluster types
ACK clusters are classified into ACK Pro, ACK standard, and ACK dedicated.
Item
ACK Pro
ACK standard
ACK dedicated
Feature
You need only to create worker nodes. ACK creates and manages master nodes.
You must create master nodes and worker nodes.
ACK standard clusters are easy to use, cost-effective, and highly available. You do
not need to manage master nodes.
For more information about the differences between ACK standard clusters and ACK Pro
clusters, see Comparison.
ACK standard clusters are easy to use, cost-effective, and highly available. You do
not need to manage master nodes.
ACK dedicated clusters allow you to manage the cluster infrastructure in a more fine-grained
manner. You must design, maintain, and upgrade the clusters on your own.
Billing method
You are charged for cluster management based on the number of clusters. In addition,
you are also charged for worker nodes and infrastructure resources. For more information,
see Billing.
Cluster management is free of charge. However, you are charged for worker nodes and
infrastructure resources. For more information, see Billing.
Cluster management is free of charge. However, you are charged for master nodes, worker
nodes, and infrastructure resources. For more information, see Billing.
Use scenario
Applicable to the production and testing environments of enterprise users.
Applicable to the learning and testing needs of individual users.
Applicable to the studies and customization of Kubernetes.
User Personas
Cluster creation procedure
Features
The following table describes the features of ACK clusters.
Cluster upgrade: You can upgrade Kubernetes with a few clicks and manage the upgrade
of system components in a unified manner. For more information, see Upgrade the Kubernetes version of an ACK cluster.
Elastic scaling: You can scale up or scale down resources in the console with a few clicks based
on your requirements. You can also use service-level affinity rules and perform horizontal
scaling.
Multi-cluster management: You can manage applications in data centers and clusters in multiple clouds and
regions in a unified manner.
Permission management: You can grant permissions to users in the Resource Access Management
(RAM) console or by using role-based access control (RBAC) policies.
Node pool management
You can manage the lifecycle of node pools. You can configure different specifications
for node pools in a cluster, such as vSwitches, runtimes, operating systems, and security
groups. For more information, see Node pool overview.
Application management
Application creation: You can create multiple types of application from images and
templates. You can configure environment variables, application health checks, data
disks, and logging.
Application lifecycle management: You can view, update, and delete applications, roll
back application versions, view application events, perform rolling updates of applications,
use new application versions to replace earlier application versions, and use triggers
to redeploy applications.
Application pod scheduling: You can schedule application pods based on the following
three policies: pod affinity, node affinity, and pod anti-affinity.
Application pod scaling: You can scale the number of application pods manually or
by using the Horizontal Pod Autoscaler (HPA).
Application release: Phased release and blue-green release are supported.
App Catalog: You can use App Catalog to simplify the integration of Alibaba Cloud
services.
Application Center: After an application is deployed, the application center displays the topology of
the application on one page. You can also manage and roll back the application version
in scenarios such as continuous deployment.
Application backup and recovery: You can back up Kubernetes applications and restore
applications from backup data. For more information, see Back up and restore applications.
You can create Block Storage volumes, Apsara File Storage NAS (NAS) volumes, Object
Storage Service (OSS) volumes, and Cloud Paralleled File System (CPFS) volumes.
You can bind a volume to a PVC.
You can dynamically create and migrate volumes.
You can view and update volumes and PVCs by running scripts.
Network
You can set up container networks based on the Flannel or Terway plug-in. For more
information, see Overview.
You can specify CIDR blocks for Services and pods.
You can use the NetworkPolicy feature. For more information, see Use network policies.
You can use Ingresses to route requests.
You can use DNS-based service discovery. For more information, see Overview.
Monitoring: You can monitor clusters, nodes, applications, and pods. You can use the
Prometheus plug-in.
Logging: You can view cluster logs, pod logs, and application logs.
Alerting: You can configure alerts to manage exceptions in the cluster based on various
metrics for different scenarios. For more information, see Alert management.
Cost analysis: provides visualized analysis on resource usage and cost distribution to help improve
resource utilization.
Runtime Security: allows you to manage security policies of the container runtime, configure routine
inspections of application security, and configure security monitoring and alerting
on the runtime. This enhances the overall security capabilities of containers.
Sandboxed-Container: allows you to run an application in a sandboxed and lightweight virtual machine.
This virtual machine has a dedicated kernel, isolates applications from each other,
and provides enhanced security. Sandboxed-Container is suitable in scenarios such
as untrusted application isolation, fault isolation, performance isolation, and load
isolation among multiple users.
TEE-based confidential computing: provides a cloud-native and all-in-one solution for developing, managing, and delivering
trusted, confidential computing applications based on Intel Software Guard Extensions
(SGX). This solution ensures data security, integrity, and confidentiality. Confidential
computing allows you to isolate sensitive data and code by using a trusted execution
environment.
Heterogeneous computing
GPU and Neural Processing Unit (NPU) computing: allows you to create clusters that
use GPU-accelerated or NPU-accelerated instances as worker nodes. Supports scheduling,
monitoring, auto scaling, and O&M management of GPU resources. For more information,
see Create an ACK managed cluster with GPU-accelerated nodes.
GPU sharing: allows you to implement a GPU sharing framework in your cluster deployed
in the cloud or in a data center to run multiple containers on a GPU-accelerated node.
For more information, see cGPU overview.
Cloud-native AI: provides cloud-native AI computing capabilities and supports orchestration
and management of data computing tasks. For more information, see Overview.
