Table of Contents
1. The concept of data storage
2. Basic storage
1. EmptyDir storage volume
2. hostPath storage volume
3. nfs shared storage volume
3. Advanced storage
1. PV (persistent volume)
2. PVC (persistent volume statement)
3. Static PV experiment
4. Dynamic PV experiment
4.1 Install nfs on the stor01 node and configure the nfs service
4.2 Create Service Account
4.3 Use Deployment to create NFS Provisioner
4.4 Create StorageClass
4.5 Create PVC and Pod tests
1. The concept of data storage
The life cycle of files on the container disk is short-lived, which causes some problems when running important applications in the container. First, when a container crashes, the kubelet will restart it, but the files in the container will be lost – the container is restarted in a clean state (the original state of the image). Secondly, when multiple containers are running simultaneously in a Pod, files usually need to be shared between these containers. The Volume abstraction in Kubernetes solves these problems very well. Containers in the Pod share the Volume through the Pause container.
2. Basic Storage
1. EmptyDir storage volume
EmptyDir is the most basic Volume type. An EmptyDir is an empty directory on the Host.
EmptyDir is created when a Pod is assigned to a Node. Its initial content is empty, and there is no need to specify the corresponding directory file on the host, because kubernetes will automatically allocate a directory. When the Pod is destroyed, the data in EmptyDir will also be permanently deleted.
The uses of EmptyDir are as follows:
- Temporary space, such as a temporary directory that is required when some applications are running and does not need to be retained permanently
- A directory where one container needs to obtain data from another container (multi-container shared directory)
Next, let’s use EmptyDir through the case of file sharing between containers and.
- Create a pod-emptydir.yaml
apiVersion: v1 Kind: Pod metadata: name: pod-emptydir namespace:default labels: app: myapp tier: frontend spec: containers: - name: myapp image: ikubernetes/myapp:v1 imagePullPolicy: IfNotPresent ports: - name: http containerPort: 80 #Define container mounting content volumeMounts: #The storage volume name used. If the value of the volume field name below is the same, it means that the storage volume of the volume is used. - name: html #Which directory to mount to in the container mountPath: /usr/share/nginx/html/ - name: busybox image: busybox:latest imagePullPolicy: IfNotPresent volumeMounts: - name: html #Define the mount storage name and mount path within the container mountPath: /data/ command: ['/bin/sh','-c','while true;do echo $(date) >> /data/index.html;sleep 2;done'] #Define storage volume volumes: #Define storage volume name - name: html #Define storage volume type emptyDir: {}
kubectl apply -f pod-emptydir.yaml #Create pod kubectl get pods -o wide #View details
Two containers are defined above, one of which is to input the date into index.html, and then verify whether the date can be obtained by accessing the nginx html. To verify that the emptyDir mounted between the two containers is shared.
?
2. hostPath storage volume
The data in EmptyDir will not be persisted, it will be destroyed when the Pod ends. If you want to simply persist the data to the host, you can choose HostPath.
HostPath is to hang an actual directory in the Node host into the Pod for use by the container. This design can ensure that the Pod is destroyed, but the data basis can exist on the Node host.
- Create a mounting directory on the node01 node
mkdir -p /data/pod/volume1 echo 'node01.kfc.com' > /data/pod/volume1/index.html
- Create a mounting directory on the node02 node
mkdir -p /data/pod/volume1 echo 'node02.kc.com' > /data/pod/volume1/index.html
- Create Pod resources
vim pod-hostpath.yaml apiVersion: v1 Kind: Pod metadata: name: pod-hostpath namespace:default spec: containers: - name: myapp image: ikubernetes/myapp:v1 #Define container mounting content volumeMounts: #The storage volume name used. If the value of the volume field name below is the same, it means that the storage volume of the volume is used. - name: html #Which directory to mount to in the container mountPath: /usr/share/nginx/html #Read-write mounting mode, the default is read-write mode false readOnly: false The #volumes field defines the host or distributed file system storage volume associated with the paues container volumes: #Storage volume name - name: html #Path, storage path for the host machine hostPath: #The path to the directory on the host machine path: /data/pod/volume1 #Define the type, which means that if the host does not have this directory, it will be created automatically type: DirectoryOrCreate
kubectl apply -f pod-hostpath.yaml
- Access Test
kubectl get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-hostpath 2/2 Running 0 37s 10.244.2.35 node02 <none> <none> curl 10.244.2.35 node02.kgc.com
- Delete the pod and rebuild it again to verify whether the original content can still be accessed
kubectl delete -f pod-hostpath.yaml kubectl apply -f pod-hostpath.yaml kubectl get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES pod-hostpath 2/2 Running 0 36s 10.244.2.37 node02 <none> <none> curl 10.244.2.37 node02.kgc.com
3. nfs shared storage volume
HostPath can solve the problem of data persistence, but once the Node node fails and the Pod is transferred to another node, problems will occur again. At this time, a separate network storage system needs to be prepared, and NFS and CIFS are commonly used.
NFS is a network file storage system. You can build an NFS server and then directly connect the storage in the Pod to the NFS system. In this case, no matter how the Pod is transferred on the node, as long as the connection between Node and NFS is OK, , the data can be accessed successfully.
- Install nfs on the stor01 node and configure the nfs service
mkdir /data/volumes -p chmod 777 /data/volumes vim /etc/exports /data/volumes 192.168.10.0/24(rw,no_root_squash) systemctl start rpcbind systemctl start nfs showmount -e Export list for stor01: /data/volumes 192.168.10.0/24
- Master node operation
vim pod-nfs-vol.yaml apiVersion: v1 Kind: Pod metadata: name: pod-vol-nfs namespace:default spec: containers: - name: myapp image: ikubernetes/myapp:v1 volumeMounts: - name: html mountPath: /usr/share/nginx/html volumes: - name: html nfs: path: /data/volumes server: stor01
kubectl apply -f pod-nfs-vol.yaml kubectl get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE pod-vol-nfs 1/1 Running 0 21s 10.244.2.38 node02
- Create index.html on nfs server
cd /data/volumes vim index.html <h1>nfs stor01</h1>
- Master node operation
curl 10.244.2.38 <h1>nfs stor01</h1> kubectl delete -f pod-nfs-vol.yaml #Delete nfs-related pods and re-create them to obtain persistent storage of data kubectl apply -f pod-nfs-vol.yaml
3. Advanced Storage
We learned earlier about using NFS to provide storage. At this time, users are required to build an NFS system and configure nfs in yaml. Since there are many storage systems supported by k8s, it is obviously unrealistic for customers to master them all. In order to shield the details of the underlying storage implementation and facilitate user use, k8s introduces two resource objects, PV and PVC.
- PV (Persistent Volume): Persistent storage volume. It is used to describe or define a storage volume, which is usually defined by operation and maintenance engineers.
- PVC (Persistent Volume Claim): It is a request for persistent storage. It is used to describe what kind of PV storage you want to use or what conditions you want to meet.
The usage logic of PVC: Define a storage volume in the Pod (the storage volume type is PVC), specify the size directly when defining, the PVC must establish a relationship with the corresponding PV, and the PVC will be based on the configuration definition Go to PV to apply, and PV is created from the storage space. PV and PVC are storage resources abstracted by Kubernetes.
The interaction between PV and PVC follows this life cycle:
Provisioning —> Binding —> Using —> Releasing —> Recycling
- Provisioning, is the creation of PV. PV can be created directly (statically) or dynamically using StorageClass.
- Binding, Assign PV to PVC
- Using, Pod uses the Volume through PVC, and can prevent deletion of the PVC in use through admission control StorageProtection (PVCProtection for 1.9 and previous versions)
- Releasing, the Pod releases the Volume and deletes the PVC
- Reclaiming, recycling PV, you can keep the PV for next use, or you can delete it directly from the cloud storage
According to these 5 stages, there are 4 states of PV:
●Available: Indicates available status and has not been bound by any PVC.
●Bound: Indicates that the PV has been bound to the PVC
●Released: Indicates that the PVC has been deleted, but the resources have not yet been reclaimed by the cluster.
●Failed: Indicates that the automatic recycling of the PV failed
The specific process of a PV from creation to destruction is as follows:
1. After a PV is created, its status will change to Available, waiting to be bound by the PVC.
2. Once bound by PVC, the status of PV will change to Bound, and it can be used by Pods with corresponding PVC defined.
3. After the Pod is used, the PV will be released, and the status of the PV will change to Released.
4. The PV that becomes Released will be recycled according to the defined recycling strategy. There are three recycling strategies, Retain, Delete and Recycle. Retain means to retain the scene. The K8S cluster does nothing and waits for the user to manually process the data in the PV. After the processing is completed, the PV is manually deleted. Delete policy, K8S will automatically delete the PV and the data in it. In Recycle mode, K8S will delete the data in the PV, and then change the status of the PV to Available, which can then be bound and used by new PVCs.
1. PV (persistent volume)
PV is an abstraction of storage resources. The following is the resource list file.
apiVersion: v1 kind: PersistentVolume metadata: name: pv2 #Note: pv is an abstraction of storage resources, it does not store the definition of namespace spec: nfs: #Storage type, corresponding to the underlying real storage (there are many, such as NFS, GFS, CIFS, etc.) path: (define the mount volume path) server: (define server name) cappcity: #Storage capacity, that is, the amount of storage space Storage: 2Gi accessModes: #Access mode storageClassName: #Storage class persistentVolumeReclaimPolicy: #Recycling policy
Description of key configuration parameters of PV
Storage type
- The actual underlying storage type, k8s supports multiple storage types, and the configuration of each storage type is different.
Storage capacity
- Currently, only storage space settings (storage=1Gi) are supported. In the future, configuration of IOPS, throughput and other indicators may be added.
Access Modes
User describes the user application’s access permissions to storage resources. Access permissions include the following methods:
- ReadWriteOnce (RWO): Read and write permissions, but can only be mounted by a single node
- ReadOnlyMany (ROX): Read-only permission, can be mounted by multiple nodes
- ReadWriteMany (RWX): read and write permissions, can be mounted by multiple nodes
Reclaim Policy (persistentVolumeReclaimPolicy)
How to deal with PV when it is no longer used. Currently three strategies are supported
- Reatin: Retain data and require administrator to manually clean the data
- Recycle: Clear the data in PV, the effect is equivalent to executing rm -rf /thevolume/*
- Delete: The back-end storage connected to the PV completes the volume deletion operation. Of course, it is common to use storage services with cloud service providers.
Storage Class
PV can specify a storage class through the storage Name parameter.
- PVs with a specific category can only be bound to PVCs that requested that category
- PVs with no category set can only be bound to PVCs that do not request any category.
2. PVC (persistent volume statement)
PVC is an application for resources, used to declare information on storage space, access mode, and storage category requirements.
Resource manifest file
apiVersion: v1 kind: PersistentVolumeClaim #Define the resource type of pvc metadata: name: pvc namespace: dev #Namespace can be defined sepc: accessModes: #Access mode selector: #Use tags to select PV storageClassName: #Storage class resources: #Request space requests: Storage: 5Gi
Description of key configuration parameters of PVC
Access Modes
- Used to describe the user application’s access rights to storage resources.
Selector
- By setting the Label Selector, PVC can be used to filter PVs that already exist in the system.
Storage Class (storageClassName)
- When defining PVC, you can set the required back-end storage class. Only PVs with this class set can be selected by the system.
Resources
- Describes a request for a storage resource
3. Static PV experiment
Use NFS as storage to demonstrate the use of PV, create 3 PVs, corresponding to the three exposed paths of NFS
1. Configure nfs storage
mkdir v{1,2,3,4,5} vim /etc/exports /data/volumes/v1 192.168.10.0/24(rw,no_root_squash) /data/volumes/v2 192.168.10.0/24(rw,no_root_squash) /data/volumes/v3 192.168.10.0/24(rw,no_root_squash) /data/volumes/v4 192.168.10.0/24(rw,no_root_squash) /data/volumes/v5 192.168.10.0/24(rw,no_root_squash) exportfs -arv showmount -e
Official documentation: https://kubernetes.io/zh-cn/docs/tasks/configure-pod-container/configure-persistent-volume-storage/#create-a-persistentvolume
2. Define PV
Five PVs are defined here, and the mounting path and access mode are defined, as well as the size of the PV division.
vim pv-demo.yaml apiVersion: v1 kind: PersistentVolume metadata: name: pv001 labels: name: pv001 spec: nfs: path: /data/volumes/v1 server: stor01 accessModes: ["ReadWriteMany","ReadWriteOnce"] capacity: Storage: 1Gi --- apiVersion: v1 kind: PersistentVolume metadata: name: pv002 labels: name: pv002 spec: nfs: path: /data/volumes/v2 server: stor01 accessModes: ["ReadWriteOnce"] capacity: Storage: 2Gi --- apiVersion: v1 kind: PersistentVolume metadata: name: pv003 labels: name: pv003 spec: nfs: path: /data/volumes/v3 server: stor01 accessModes: ["ReadWriteMany","ReadWriteOnce"] capacity: Storage: 2Gi --- apiVersion: v1 kind: PersistentVolume metadata: name: pv004 labels: name: pv004 spec: nfs: path: /data/volumes/v4 server: stor01 accessModes: ["ReadWriteMany","ReadWriteOnce"] capacity: storage: 4Gi --- apiVersion: v1 kind: PersistentVolume metadata: name: pv005 labels: name: pv005 spec: nfs: path: /data/volumes/v5 server: stor01 accessModes: ["ReadWriteMany","ReadWriteOnce"] capacity: Storage: 5Gi
kubectl apply -f pv-demo.yaml
kubectl get pv NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE pv001 1Gi RWO,RWX Retain Available 7s pv002 2Gi RWO Retain Available 7s pv003 2Gi RWO,RWX Retain Available 7s pv004 4Gi RWO,RWX Retain Available 7s pv005 5Gi RWO,RWX Retain Available 7s
3. Define PVC
The access mode of pvc is defined here as multi-channel read and write. This access mode must be among the access modes defined by pv previously. Define the size of the PVC application to be 2Gi. At this time, the PVC will automatically match the multi-channel read and write PV with a size of 2Gi. The status of the successful matching of the PVC is Bound.
vim pod-vol-pvc.yaml apiVersion: v1 kind: PersistentVolumeClaim metadata: name:mypvc namespace:default spec: accessModes: ["ReadWriteMany"] resources: requests: Storage: 2Gi --- apiVersion: v1 Kind: Pod metadata: name: pod-vol-pvc namespace:default spec: containers: - name: myapp image: ikubernetes/myapp:v1 volumeMounts: - name: html mountPath: /usr/share/nginx/html volumes: - name: html persistentVolumeClaim: claimName: mypvc
kubectl apply -f pod-vol-pvc.yaml
kubectl get pv NAME CAPACITY ACCESS MODES RECLAIM POLICY STATUS CLAIM STORAGECLASS REASON AGE pv001 1Gi RWO,RWX Retain Available 19m pv002 2Gi RWO Retain Available 19m pv003 2Gi RWO,RWX Retain Bound default/mypvc 19m pv004 4Gi RWO,RWX Retain Available 19m pv005 5Gi RWO,RWX Retain Available 19m kubectl get pvc NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE mypvc Bound pv003 2Gi RWO,RWX 22s
4. Test access
Create index.html on the storage server, write data, and view the corresponding page by accessing the Pod.
cd /data/volumes/v3/ echo "welcome to use pv3" > index.html kubectl get pods -o wide pod-vol-pvc 1/1 Running 0 3m 10.244.2.39 k8s-node02 curl 10.244.2.39 welcome to use pv3
4. Dynamic PV experiment
The PV and PVC modes introduced above require operation and maintenance personnel to create PVs first, and then developers define PVCs for one-to-one bonding. However, if there are thousands of PVC requests, thousands of PVs need to be created. Maintenance costs are very high for operation and maintenance personnel. Kubernetes provides a mechanism to automatically create PVs called StorageClass, which is used to create PV templates.
Creating a StorageClass requires defining the attributes of the PV, such as storage type, size, etc.; in addition, creating such a PV requires the use of storage plug-ins, such as Ceph, etc. With these two pieces of information, Kubernetes can find the corresponding StorageClass based on the PVC submitted by the user. Then Kubernetes will call the storage plug-in declared by the StorageClass to automatically create the required PV and bind it.
Build StorageClass + NFS to realize dynamic PV creation of NFS
The dynamic PV creation supported by Kubernetes itself does not include NFS, so you need to use an external storage volume plug-in to allocate PVs. For details, see: https://kubernetes.io/zh/docs/concepts/storage/storage-classes/
The volume plug-in is called Provisioner (storage allocator), and NFS uses nfs-client. This external volume plug-in will automatically create PVs using the configured NFS server.
4.1 Install nfs on the stor01 node and configure the nfs service
mkdir /opt/k8s chmod 777 /opt/k8s/ vim /etc/exports /opt/k8s 192.168.10.0/24(rw,no_root_squash,sync) systemctl restart nfs
4.2 Create Service Account
Create a Service Account to manage the permissions of NFS Provisioner to run in the k8s cluster, and set nfs-client rules for PV, PVC, StorageClass, etc.
vim nfs-client-rbac.yaml #Create a Service Account to manage the permissions of NFS Provisioner running in the k8s cluster apiVersion: v1 kind: ServiceAccount metadata: name: nfs-client-provisioner --- #Create cluster role apiVersion: rbac.authorization.k8s.io/v1 kind:ClusterRole metadata: name: nfs-client-provisioner-clusterrole rules: - apiGroups: [""] resources: ["persistentvolumes"] verbs: ["get", "list", "watch", "create", "delete"] - apiGroups: [""] resources: ["persistentvolumeclaims"] verbs: ["get", "list", "watch", "update"] - apiGroups: ["storage.k8s.io"] resources: ["storageclasses"] verbs: ["get", "list", "watch"] - apiGroups: [""] resources: ["events"] verbs: ["list", "watch", "create", "update", "patch"] - apiGroups: [""] resources: ["endpoints"] verbs: ["create", "delete", "get", "list", "watch", "patch", "update"] --- #Cluster role binding apiVersion: rbac.authorization.k8s.io/v1 kind: ClusterRoleBinding metadata: name: nfs-client-provisioner-clusterrolebinding subjects: - kind: ServiceAccount name: nfs-client-provisioner namespace:default roleRef: kind:ClusterRole name: nfs-client-provisioner-clusterrole apiGroup: rbac.authorization.k8s.io
kubectl apply -f nfs-client-rbac.yaml
4.3 Use Deployment to create NFS Provisioner
NFS Provisione (ie nfs-client) has two functions: one is to create a mount point (volume) under the NFS shared directory, and the other is to associate the PV with the NFS mount point.
#Since selfLink is enabled in version 1.20, an error will be reported when k8s 1.20 + version dynamically generates pv through nfs provisioner. The solution is as follows: vim /etc/kubernetes/manifests/kube-apiserver.yaml spec: containers: - command: - kube-apiserver - --feature-gates=RemoveSelfLink=false #Add this line - --advertise-address=192.168.80.20 ... kubectl apply -f /etc/kubernetes/manifests/kube-apiserver.yaml kubectl delete pods kube-apiserver -n kube-system kubectl get pods -n kube-system | grep apiserver
#Create NFS Provisioner vim nfs-client-provisioner.yaml Kind: Deployment apiVersion: apps/v1 metadata: name: nfs-client-provisioner spec: replicas: 1 selector: matchLabels: app: nfs-client-provisioner strategy: type: Recreate template: metadata: labels: app: nfs-client-provisioner spec: serviceAccountName: nfs-client-provisioner #Specify Service Account account containers: - name: nfs-client-provisioner image: quay.io/external_storage/nfs-client-provisioner:latest imagePullPolicy: IfNotPresent volumeMounts: - name: nfs-client-root mountPath: /persistentvolumes env: - name: PROVISIONER_NAME value: nfs-storage #Configure the provisioner Name and ensure that the name is consistent with the provisioner name in the StorageClass resource - name: NFS_SERVER value: stor01 #Configure the bound nfs server - name: NFS_PATH value: /opt/k8s #Configure the bound nfs server directory volumes: #Declare nfs data volume - name: nfs-client-root nfs: server: stor01 path: /opt/k8s \t
kubectl apply -f nfs-client-provisioner.yaml kubectl getpod NAME READY STATUS RESTARTS AGE nfs-client-provisioner-cd6ff67-sp8qd 1/1 Running 0 14s
4.4 Create StorageClass
Create StorageClass, responsible for establishing PVC and calling NFS provisioner to perform scheduled work, and associate PV with PVC
vim nfs-client-storageclass.yaml apiVersion: storage.k8s.io/v1 kind: StorageClass metadata: name: nfs-client-storageclass provisioner: nfs-storage #The name here should be consistent with the environment variable PROVISIONER_NAME in the provisioner configuration file parameters: archiveOnDelete: "false" #false means that the data will not be archived when the PVC is deleted, that is, the data will be deleted kubectl apply -f nfs-client-storageclass.yaml kubectl get storageclass NAME PROVISIONER RECLAIMPOLICY VOLUMEBINDINGMODE ALLOWVOLUMEEXPANSION AGE nfs-client-storageclass nfs-storage Delete Immediate false 43s
4.5 Create PVC and Pod test
vim test-pvc-pod.yaml apiVersion: v1 kind: PersistentVolumeClaim metadata: name: test-nfs-pvc spec: accessModes: - ReadWriteMany storageClassName: nfs-client-PROVISIONER #Associate StorageClass object resources: requests: Storage: 1Gi --- apiVersion: v1 Kind: Pod metadata: name: test-storageclass-pod spec: containers: - name: busybox image: busybox:latest imagePullPolicy: IfNotPresent command: - "/bin/sh" - "-c" args: - "sleep 3600" volumeMounts: - name: nfs-pvc mountPath: /mnt restartPolicy: Never volumes: - name: nfs-pvc persistentVolumeClaim: claimName: test-nfs-pvc #Consistent with the PVC name
kubectl apply -f pod-hostpath.yaml