Container Service for Kubernetes (ACK) provides container storage services based on the Container Storage Interface (CSI) plug-in. CSI enables the integration with Elastic Block Storage (EBS), Apsara File Storage NAS (NAS), Cloud Paralleled File System (CPFS), Object Storage Service (OSS), and local disks of Alibaba Cloud. CSI also enables the compatibility with native Kubernetes storage services, such as emptyDir, HostPath, Secret, and ConfigMap. This topic describes the overview, features, and limits of the CSI plug-in, and the permissions that are required to use the CSI plug-in.
ACK container storage architecture
|Alibaba Cloud storage service||Statically provisioned volume||Dynamically provisioned volume||Deployed by default||Feature||Scenario|
|Disk||Supported||Supported||Yes||Non-shared storage. A disk can be mounted only to one node.||
For more information, see Disk volume overview.
|NAS||Supported||Supported||Yes||Shared storage that provides high performance and high throughput.||
For more information, see NAS volume overview.
|OSS||Supported||Not supported||Yes||Shared storage that supports file systems in user space.||
Note OSS buckets are mounted by using ossfs, which is implemented as a file system in user space (FUSE). The write performance is limited when you use OSS volumes. We recommend that you use other storage volumes in scenarios that require high write performance.
For more information, see OSS volume overview.
|CPFS||Supported||Supported||No||Shared storage that features high performance and high bandwidth.||
- A PV is a piece of storage in the cluster. A PV has a lifecycle that is independent of the pod that uses the PV. Different types of PV can be created based on different types of StorageClass.
- A PVC is a request for storage in the cluster. PVs are node resources consumed by pods. PVCs are claims that consume PVs. When PVs are insufficient, PVCs can dynamically provision PVs.
ACK container storage features
The following table describes the storage features supported by different types of ACK cluster.
|Storage type||Feature||ACK cluster (Linux-based)||ASK cluster||Registered cluster (hybrid cloud or multicloud)||ACK@Edge||ACK cluster (Windows-based)||Container Service cluster||ACK cluster that supports sandboxed containers|
|Block storage||Mount and unmount disks||✔️||✔️||❌||❌||✔️||✔️||✔️|
|Container I/O monitoring||✔️||❌||❌||❌||❌||✔️||❌|
|File systems||XFS, ext4, and dBFS are supported.||XFS and ext4 are supported.||❌||❌||NTFS is supported.||XFS and ext4 are supported.||XFS and ext4 are supported.|
|Block devices and raw devices||✔️||❌||❌||❌||❌||✔️||❌|
|Data restoration from snapshots||✔️||✔️||❌||❌||❌||✔️||✔️|
|Disk queue settings||✔️||❌||❌||❌||❌||✔️||❌|
|Customer managed key (CMK)-based encryption and Bring Your Own Key (BYOK) encryption||✔️||✔️||❌||❌||✔️||✔️||✔️|
|File storage||Create, mount, and unmount NAS file systems||✔️||✔️||✔️||✔️||❌||✔️||✔️|
|Mount and unmount Samba file systems||❌||❌||✔️||❌||✔️||❌||❌|
|Recycle bin (CNFS)||✔️||❌||❌||✔️||❌||❌||❌|
|Subdirectories or shared directories of dynamically provisioned volumes (CNFS)||✔️||❌||✔️||✔️||✔️||✔️||✔️|
|CMK-based encryption (CNFS and Extreme NAS file systems )||✔️||❌||❌||✔️||❌||✔️||✔️|
|Quota limits (CNFS)||Supported only by managed Kubernetes clusters||❌||❌||✔️||❌||❌||❌|
|Capacity and I/O monitoring (CNFS)||✔️||❌||✔️||✔️||❌||❌||❌|
|Online resizing (CNFS)||✔️||❌||✔️||✔️||❌||❌||❌|
|Object storage||Mount and unmount OSS buckets||✔️||✔️||✔️||✔️||❌||✔️||✔️|
|Local storage||Linux Volume Manager (LVM)-managed block storage||✔️||❌||✔️||✔️||❌||✔️||❌|
|Automated volume groups||✔️||❌||✔️||✔️||❌||✔️||❌|
|LVM-managed capacity-aware scheduling||✔️||❌||✔️||✔️||❌||✔️||❌|
|PMEM Direct Mem||✔️||❌||❌||❌||❌||❌||❌|
|LVM-managed persistent memory (PMEM)||✔️||❌||❌||❌||❌||❌||❌|
CSI deployment architectures
|Managed Kubernetes cluster||Dedicated Kubernetes cluster|
In managed Kubernetes clusters, CSI-Provisioner and CSI-Plugin are deployed on worker nodes.
In dedicated Kubernetes clusters, CSI-Provisioner is deployed on master nodes. CSI-Plugin is automatically deployed as DaemonSets on master and worker nodes.
Permissions required to use CSI
Before you can use the CSI plug-in to mount, unmount, create, and delete volumes, you must grant the plug-in the permissions to access other cloud resources. You can use an AccessKey pair or a Resource Access Management (RAM) role to grant permissions to the CSI plug-in. The default method is to use a RAM role. The following table describes the two authorization methods.
|Use an AccessKey pair||Use a RAM role|
The CSI plug-in uses the RAM role AliyunCSManagedCsiRole to access resources of other cloud services. For more information, see AliyunCSManagedCsiRole. For more information about how to grant permissions to a RAM role, see Grant permissions to a RAM role.
Limits on CSI
When you use the CSI plug-in in ACK clusters, take note of the limits on the CSI plug-in and Alibaba Cloud storage services.
- Limits on Alibaba Cloud storage services
Alibaba Cloud storage service Limit Disk
- You can mount at most 15 disk volumes to a node.
- You can provision a disk as a volume only for one pod.
- You cannot mount disks of all types to ECS instances of all types. For more information, see Instance family.
- You cannot mount or unmount subscription disks as volumes.
- You can mount a disk only to an ECS instance in the same zone as the disk.
- We recommend that you create StatefulSets instead of Deployments to use disk volumes.
Note Deployments are used to create stateless applications. When a pod is restarted, the start time of the new pod may overlap with the end time of the old pod. If multiple pods are created for a Deployment, no dedicated volume is provisioned for each pod.
- The minimum capacity of a disk volume is 20 GiB.
- You can mount a NAS volume only to ECS instances in the same virtual private cloud (VPC) as the NAS file system.
- The number of NAS file systems that you can create is subject to a quota limit. To request the quota increase, Submit a ticket to the NAS team.
OSS We recommend that you do not perform data write operations on OSS volumes. Use other storage media for data write operations.Note OSS volumes are mounted by using ossfs, which is implemented as a file system in user space (FUSE). The write performance is limited when you use OSS volumes. We recommend that you use other storage volumes in scenarios that require high write performance. CPFS
- The CPFS driver is highly dependent on the OS kernel. After you deploy the CPFS environment, do not upgrade the OS kernel.
- You can install the CPFS driver but cannot upgrade the CPFS driver.
The CSI-CPFS plug-in is a Kubernetes CSI component that mounts CPFS volumes to application pods.
The CPFS driver is a client driver that implements the CPFS protocol at the kernel layer.Relationship between the CSI-CPFS plug-in and the CPFS driver:
- If the CPFS driver is not installed on a node, the CPFS driver is automatically installed when you deploy the CSI-CPFS plug-in.
- If the CPFS driver is installed on a node, the CPFS driver is not installed or upgraded when you deploy the CSI-CPFS plug-in.
- Limits on the CSI plug-in
The CSI plug-in is an open source plug-in for ACK clusters. In other types of clusters, such as clusters deployed in third-party clouds and self-managed clusters on Alibaba Cloud, you cannot directly use the CSI plug-in for reasons such as cluster configurations, permission management, and network differences. If you want to use the CSI plug-in in these types of clusters, you must modify the cluster configurations based on the source code. For more information, see alibaba-cloud-csi-driver.
Requirements on Kubernetes versions
To use the CSI plug-in in an ACK cluster, the Kubernetes version of the cluster must
be 1.14 or later. Besides, the kubelet parameter
--enable-controller-attach-detach must be set to
Installation and upgrade of the CSI plug-in
For more information about how to install and upgrade the CSI plug-in, see Install and upgrade the CSI plug-in.
Differences between the CSI and FlexVolume plug-ins
|Flexvolume||FlexVolume is a traditional mechanism to extend the storage systems developed by the
Kubernetes community. ACK supports FlexVolume. FlexVolume consists of the following
For more information about FlexVolume, see Overview.
For more information about how to upgrade FlexVolume, see Manage system components.
|CSI||The Kubernetes community recommends the CSI plug-in. The CSI plug-in provided by ACK
is compatible with the features of the community version. CSI consists of the following
- You must select a plug-in when you create an ACK cluster.
- You cannot use CSI and FlexVolume in the same cluster.
- You cannot change the plug-in from FlexVolume to CSI for a cluster.
- For newly created ACK clusters, we recommend that you use CSI. The ACK technical team will continuously upgrade CSI to support more features of the CSI community version.
- For existing clusters, we recommend that you use the plug-in that is already installed. The ACK technical team will continue its support for FlexVolume.
How to check the storage plug-in used in a cluster
- Method 1: Check node annotations by using the console
- Log on to the ACK console.
- In the left-side navigation pane, click Clusters.
- On the Clusters page, find the cluster to which you want to add a node. Then, click the name of the cluster or click Details in the Actions column.
- In the left-side navigation pane of the details page, choose .
- Select a node and click Actions column. in the
- In the Overview section, check Annotations.
volumes.kubernetes.io/controller-managed-attach-detach: trueis displayed, the cluster uses the CSI plug-in. If
volumes.kubernetes.io/controller-managed-attach-detach: trueis not displayed, the cluster uses the FlexVolume plug-in.
- Method 2: Check kubelet parameters
Run the following command to check kubelet parameters:
ps -ef | grep kubelet
If the value of
true, the cluster uses the CSI plug-in. If the value of
false, the cluster uses the FlexVolume plug-in.