You can mount disk volumes to Container Service for Kubernetes (ACK) clusters. This topic describes the features, disk specifications, use scenarios, usage notes, billing rules, and StorageClasses that are related to disk volumes.

Introduction

Alibaba Cloud disks are block-level data storage resources for Elastic Compute Service (ECS). Alibaba Cloud disks provide low latency, high performance, high durability, and high reliability. Alibaba Cloud disks use a distributed triplicate mechanism to ensure data reliability for ECS instances. If service disruptions occur within a zone due to hardware errors, data in the zone is automatically replicated to an unaffected disk in another zone to ensure data availability.

  • Enhanced SSDs (ESSDs): ESSDs are based on the next-generation distributed block storage architecture and use the 25 Gigabit Ethernet and remote direct memory access (RDMA) technologies. ESSDs provide low-latency input and output. Each ESSD can provide up to 1,000,000 random read/write IOPS. For more information, see ESSDs.

    We recommend that you use ESSDs for scenarios such as online transactional processing (OLTP) databases, NoSQL databases, and Elasticsearch, Logstash, and Kibana (ELK) distributed logs.

  • Standard SSDs: Standard SSDs are high-performance disks that provide consistent high random IOPS and high data reliability.

    We recommend that you use standard SSDs for scenarios such as I/O-intensive applications, small and medium-sized relational databases, and NoSQL databases.

  • Ultra disks: Ultra disks are cost-effective and provide medium random IOPS and high data reliability.

    We recommend that you use ultra disks as system disks for scenarios such as development and testing.

  • Basic disks: Basic disks are the previous generation of disks and are unavailable for purchase.

Disk specifications

ItemESSD AutoPL diskESSDStandard SSDUltra diskBasic disk
PL3PL2PL1PL0
Capacity range per disk (GiB)40~32,7681261~32,768461~32,76820~32,76840~32,76820~32,76820~32,7685~2,000
Maximum IOPS1,000,0001,000,000100,00050,00010,00025,000 5,000Several hundreds
Maximum throughput (MB/s)4,0964,000750350180300 14030~40
Formula for calculating the IOPS per disk
  • Baseline performance: min{1,800 + 50 × Capacity, 50,000}
  • Provisioned performance: min{Provisioned IOPS, 50,000}
  • Burst performance: min{IOPS supported by the instance type, 1,000,000}
min{1,800 + 50 × Capacity, 1,000,000}min{1,800 + 50 × Capacity, 100,000}min{1,800 + 50 × Capacity, 50,000}min{1,800+12 × Capacity, 10,000}min{1,800 + 30 × Capacity, 25,000}min{1,800 + 8 × Capacity, 5,000}None
Formula for calculating the throughput per disk (MB/s)
  • Baseline performance: min{120 + 0.5 × Capacity, 350}
  • Provisioned performance: min{16 KB × Provisioned IOPS/1,024, Maximum throughput per disk}
  • Burst performance: min{Throughput supported by the instance type, 4 GB/s}
min{120 + 0.5 × Capacity, 4,000}min{120 + 0.5 × Capacity, 750}min{120 + 0.5 × Capacity, 350}min{100 + 0.25 × Capacity, 180}min{120 + 0.5 × Capacity, 300}min{100 + 0.15 × Capacity, 140}None
Average one-way random write latency in milliseconds (block size = 4 KB)0.20.20.20.20.3~0.50.5~21~35~10
API parameter valuecloud_autocloud_essdcloud_essdcloud_essdcloud_essdcloud_ssdcloud_efficiencycloud

For more information about disk performance, see EBS performance.

Use scenarios

The following table describes the operations that you can perform on disks to meet your business requirements.

Business requirementReferences
Store application data
Resize a system disk or a data disk

For more information about disk expansion, see Overview.

For more information about how to expand a disk, see the following topics:
Persistent storageFor more information, see the following topics:

Usage notes

  • Alibaba Cloud disks cannot be shared. A disk can be mounted only to one pod.
  • We recommend that you mount a disk by using a StatefulSet. If you use a Deployment to mount a disk, you must set the number of replicated pods to 1. If the number of replicated pods is not set to 1, multiple nodes may use the disk at the same time when replicated pods are scheduled to different nodes. In this case, you cannot prioritize the node where you want to mount or unmount a disk. When Deployment pods are restarted, disk mounting may fail due to the upgrade policy. Therefore, we recommend that you do not use a Deployment to mount a disk.
  • The type of disk must match the ECS instance types that are used in your cluster before you can mount a disk. For more information about the matching rules between disk types and ECS instance types, see Overview of instance families.
  • You can mount at most 16 disks to each node. The maximum capacity of each disk is 32 TiB.
  • When you mount a disk volume, if the securityContext.fsgroup parameter is set in the application template, kubelet performs the chmod and chown operations after the disk volume is mounted. This increases the mounting time. For more information about how to reduce the mounting time, see Why does it require a long time to mount a NAS volume? and Why does it require a long period of time to mount an OSS volume?.

Billing

  • Only pay-as-you-go disks can be mounted. If you change the billing method of an ECS instance in the cluster from pay-as-you-go to subscription, you cannot change the billing method of its disks to subscription. Otherwise, the disks cannot be mounted to the cluster.
  • For more information about the billing of disks, visit the ECS product page.
  • If the disk that you want to mount does not support the subscription billing method, you can purchase storage capacity units (SCUs) to reduce the costs. For more information about SCUs, see Overview.

For more information, see Billing.

StorageClasses

If you create a large number of PVCs, you must also create a large number of PVs. This increases the workloads of O&M engineers. To solve this issue, Kubernetes provides the StorageClass object that you can use to enable automatic PV creation.

StorageClass

ACK clusters support the following types of StorageClass:
  • alicloud-disk-efficiency: ultra disk.
  • alicloud-disk-ssd: standard SSD.
  • alicloud-disk-essd: ESSD.
  • alicloud-disk-available: a high-availability mode. In this mode, the system first attempts to create a standard SSD. If SSD resources are exhausted, the system attempts to create an ultra disk.
    Important
    • For alicloud-csi-provisioner versions earlier than 1.14.8.39-0d749258-aliyun, the system first attempts to create an ESSD.
    • If ESSDs are out of stock, the system attempts to create a standard SSD.
    • If standard SSDs are out of stock, the system attempts to create an ultra disk.
  • alicloud-disk-topology: creates a disk in WaitForFirstConsumer mode.

The first four types of StorageClass are suitable for clusters that are created in single zones. The last type of StorageClass is suitable for clusters that are created across zones.

When you configure a StorageClass, the following rules determine the zone where the disk is created:

  • If you set volumeBindingMode: WaitForFirstConsumer in the StorageClass configuration, a disk is created in the zone where the pod is deployed.
  • If you set volumeBindingMode: Immediate and specify only one zone in the zoneId parameter of the StorageClass configuration, a disk is created in the specified zone.
  • If you set volumeBindingMode: Immediate and specify multiple zones in the zoneId parameter of the StorageClass configuration, the system attempts to create a disk in the specified zones in a round robin manner.

If your cluster is deployed across zones, we recommend that you specify volumeBindingMode: WaitForFirstConsumer in the StorageClass configurations. You can create a StorageClass based on the type of disk that is required.

Default StorageClass

Kubernetes provides the default StorageClass feature. If a PVC does not specify a StorageClass, the default StorageClass is used to provision a PV for the PVC. For more information, see Default StorageClass.

Note
  • The default StorageClass takes effect on all PVCs. Proceed with caution if your cluster uses PVCs of different storage types. For example, the default StorageClass may create a disk PV for a PVC that defines a NAS file system. Therefore, ACK clusters do not provide default StorageClasses. If you want to configure a default StorageClass, perform the following steps.
  • You can configure only one default StorageClass for each cluster. If you configure more than one default StorageClass for a cluster, all default StorageClasses become invalid.
  1. Run the following command to set alicloud-disk-ssd as the default StorageClass: 
    kubectl patch storageclass alicloud-disk-ssd -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

    After the default StorageClass is configured, alicloud-disk-ssd is marked as (default) 

    kubectl get sc
    Expected output:
    NAME                          PROVISIONER                       AGE
alicloud-disk-ssd (default)   diskplugin.csi.alibabacloud.com   96m
  2. Use the default StroageClass.

    Use the following template to create a PVC without specifying a StorageClass: 

    apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: disk-pvc
spec:
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 20Gi

    The cluster automatically creates a disk PV based on the default StorageClass alicloud-disk-ssd. 

    kubectl get pvc
    Expected output:
    NAME       STATUS   VOLUME                   CAPACITY   ACCESS MODES   STORAGECLASS        AGE
disk-pvc   Bound    d-bp18pbai447qverm3ttq   20Gi       RWO            alicloud-disk-ssd   49s
You can also run the following command to disable the default StorageClass: 
kubectl patch storageclass alicloud-disk-ssd -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"false"}}}'