Persistent Storage Class Configuration on Kubernetes
TiDB cluster components such as PD, TiKV, TiDB monitoring, TiDB Binlog, andtidb-backup
require persistent storage for data. To achieve this on Kubernetes, you need to usePersistentVolume (PV). Kubernetes supports different types ofstorage classes, which can be categorized into two main types:
Network storage
Network storage is not located on the current node but is mounted to the node through the network. It usually has redundant replicas to ensure high availability. In the event of a node failure, the corresponding network storage can be remounted to another node for continued use.
Local storage
Local storage is located on the current node and typically provides lower latency compared to network storage. However, it does not have redundant replicas, so data might be lost if the node fails. If the node is an IDC server, data can be partially restored, but if it is a virtual machine using local disk on a public cloud, data cannot be retrieved after a node failure.
PVs are automatically created by the system administrator or volume provisioner. PVs and Pods are bound byPersistentVolumeClaim (PVC). Instead of creating a PV directly, users request to use a PV through a PVC. The corresponding volume provisioner creates a PV that meets the requirements of the PVC and then binds the PV to the PVC.
Recommended storage classes for TiDB clusters
TiKV uses the Raft protocol to replicate data. When a node fails, PD automatically schedules data to fill the missing data replicas. TiKV requires low read and write latency, so it is strongly recommended to use local SSD storage in a production environment.
PD also uses Raft to replicate data. PD is not an I/O-intensive application, but rather a database for storing cluster meta information. Therefore, a local SAS disk or network SSD storage such as EBS General Purpose SSD (gp2) volumes on AWS or SSD persistent disks on Google Cloud can meet the requirements.
To ensure availability, it is recommended to use network storage for components such as TiDB monitoring, TiDB Binlog, andtidb-backup
because they do not have redundant replicas. TiDB Binlog's Pump and Drainer components are I/O-intensive applications that require low read and write latency, so it is recommended to use high-performance network storage such as EBS Provisioned IOPS SSD (io1) volumes on AWS or SSD persistent disks on Google Cloud.
When deploying TiDB clusters ortidb-backup
with TiDB Operator, you can configure theStorageClass
for the components that require persistent storage via the correspondingstorageClassName
field in thevalues.yaml
configuration file. TheStorageClassName
is set tolocal-storage
by default.
Network PV configuration
Starting from Kubernetes 1.11, volume expansion of network PV is supported. However, you need to run the following command to enable volume expansion for the correspondingStorageClass
:
kubectl patch storageclass${storage_class}-p'{"allowVolumeExpansion": true}'
After enabling volume expansion, you can expand the PV using the following method:
Edit the PersistentVolumeClaim (PVC) object:
Suppose the PVC is currently 10 Gi and you need to expand it to 100 Gi.
kubectl patch pvc -n${namespace} ${pvc_name}-p'{"spec": {"resources": {"requests": {"storage": "100Gi"}}}}'视图大小of the PV:
After the expansion, the size displayed by running
kubectl get pvc -n ${namespace} ${pvc_name}
still shows the original size. However, if you run the following command to view the size of the PV, it shows that the size has been expanded to the expected value.kubectl get pv | grep${pvc_name}
Local PV configuration
Currently, Kubernetes supports statically allocated local storage. To create a local storage object, uselocal-volume-provisioner
in thelocal-static-provisionerrepository.
Step 1: Pre-allocate local storage
For a disk that stores TiKV data, you canmountthe disk into the
/mnt/ssd
目录中。To achieve high performance, it is recommended to allocate a dedicated disk for TiDB, with SSD being the recommended disk type.
For a disk that stores PD data, follow thestepsto mount the disk. First, create multiple directories on the disk and bind mount the directories into the
/mnt/sharedssd
目录中。For a disk that stores monitoring data, follow thestepsto mount the disk. First, create multiple directories on the disk and bind mount the directories into the
/mnt/monitoring
目录中。For a disk that stores TiDB Binlog and backup data, follow thestepsto mount the disk. First, create multiple directories on the disk and bind mount the directories into the
/mnt/backup
目录中。
The/mnt/ssd
,/mnt/sharedssd
,/mnt/monitoring
, and/mnt/backup
directories mentioned above are discovery directories used by local-volume-provisioner. For each subdirectory in the discovery directory, local-volume-provisioner creates a corresponding PV.
Step 2: Deploy local-volume-provisioner
Online deployment
Download the deployment file for the local-volume-provisioner.
wget https://raw.githubusercontent.com/pingcap/tidb-operator/v1.5.0/examples/local-pv/local-volume-provisioner.yamlIf you are using the same discovery directory as described inStep 1: Pre-allocate local storage, you can skip this step. If you are using a different path for the discovery directory than in the previous step, you need to modify the ConfigMap and DaemonSet spec.
Modify the
data.storageClassMap
field in the ConfigMap spec:apiVersion: v1 kind: ConfigMap metadata: name: local-provisioner-config namespace: kube-system data: # ... storageClassMap: | ssd存储设备:hostDir: / mnt / ssd mountDir: / mnt /党卫军d shared-ssd-storage: hostDir: /mnt/sharedssd mountDir: /mnt/sharedssd monitoring-storage: hostDir: /mnt/monitoring mountDir: /mnt/monitoring backup-storage: hostDir: /mnt/backup mountDir: /mnt/backupFor more configuration options for the local-volume-provisioner, refer to the配置document.
Modify the
volumes
andvolumeMounts
fields in the DaemonSet spec to ensure that the discovery directory can be mounted to the corresponding directory in the Pod:...... volumeMounts: - mountPath: /mnt/ssd name: local-ssd mountPropagation: “HostToContainer” - mountPath: /mnt/sharedssd name: local-sharedssd mountPropagation: “HostToContainer” - mountPath: /mnt/backup name: local-backup mountPropagation: “HostToContainer” - mountPath: /mnt/monitoring name: local-monitoring mountPropagation: “HostToContainer” volumes: - name: local-ssd hostPath: path: /mnt/ssd - name: local-sharedssd hostPath: path: /mnt/sharedssd - name: local-backup hostPath: path: /mnt/backup - name: local-monitoring hostPath: path: /mnt/monitoring ......
Deploy the
local-volume-provisioner
.kubectl apply -f https://raw.githubusercontent.com/pingcap/tidb-operator/v1.5.0/manifests/local-dind/local-volume-provisioner.yamlCheck the status of the Pod and PV.
kubectl get po -n kube-system -l app=local-volume-provisioner && \ kubectl get pv | grep -e ssd-storage -e shared-ssd-storage -e monitoring-storage -e backup-storageThe
local-volume-provisioner
creates a PV for each mounting point under the discovery directory.
For more information, refer to theKubernetes local storageandlocal-static-provisionerdocuments.
Offline deployment
The steps for offline deployment are the same as for online deployment, except for the following:
Download the
local-volume-provisioner.yaml
file on a machine with Internet access, then upload it to the server and install it.The
local-volume-provisioner
is a DaemonSet that starts a Pod on every Kubernetes worker node. The Pod uses thequay.io/external_storage/local-volume-provisioner:v2.3.4
image. If the server does not have access to the Internet, download this Docker image on a machine with Internet access:docker pull quay.io/external_storage/local-volume-provisioner:v2.3.4 docker save -o local-volume-provisioner-v2.3.4.tar quay.io/external_storage/local-volume-provisioner:v2.3.4Copy the
local-volume-provisioner-v2.3.4.tar
file to the server, and execute thedocker load
command to load the file on the server:docker load -i local-volume-provisioner-v2.3.4.tar
Best practices
- The unique identifier for a local PV is its path. To avoid conflicts, it is recommended to generate a unique path using the UUID of the device.
- To ensure I/O isolation, it is recommended to use a dedicated physical disk per PV for hardware-based isolation.
- For capacity isolation, it is recommended to use either a partition per PV or a physical disk per PV.
For more information on local PV on Kubernetes, refer to theBest Practicesdocument.
Data safety
In general, when a PVC is deleted and no longer in use, the PV bound to it is reclaimed and placed in the resource pool for scheduling by the provisioner. To prevent accidental data loss, you can configure the reclaim policy of theStorageClass
toRetain
globally or change the reclaim policy of a single PV toRetain
. With theRetain
policy, a PV is not automatically reclaimed.
To configure globally:
The reclaim policy of a
StorageClass
is set at creation time and cannot be updated once created. If it is not set during creation, you can create anotherStorageClass
with the same provisioner. For example, the default reclaim policy of theStorageClass
for persistent disks on Google Kubernetes Engine (GKE) is删除
. You can create anotherStorageClass
namedpd-standard
with a reclaim policy ofRetain
and change thestorageClassName
of the corresponding component topd-standard
when creating a TiDB cluster.apiVersion: storage.k8s.io/v1 kind: StorageClass metadata: name: pd-standard parameters: type: pd-standard provisioner: kubernetes.io/gce-pd reclaimPolicy: Retain volumeBindingMode: ImmediateTo configure a single PV:
kubectl patch pv${pv_name}-p'{"spec":{"persistentVolumeReclaimPolicy":"Retain"}}'
删除PV and data
When the reclaim policy of PVs is set toRetain
, if you have confirmed that the data of a PV can be deleted, you can delete the PV and its corresponding data by following these steps:
删除the PVC object corresponding to the PV:
kubectl delete pvc${pvc_name}--namespace=${namespace}Set the reclaim policy of the PV to
删除
. This automatically deletes and reclaims the PV.kubectl patch pv${pv_name}-p'{"spec":{"persistentVolumeReclaimPolicy":"Delete"}}'
For more details, refer to theChange the Reclaim Policy of a PersistentVolumedocument.