Skip to content

The cubefs-helm project helps deploy a CubeFS cluster orchestrated by Kubernetes.

License

Notifications You must be signed in to change notification settings

cubefs/cubefs-helm

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

98 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

cubefs-helm

Deploy Cubefs using Kubernetes and Helm

The cubefs-helm project helps deploy a Cubefs cluster orchestrated by Kubernetes.

Governance

  • CubeFS-helm is a sub-project of the CubeFS main project, complying with the rules of CubeFS main projects.
  • CubeFS-helm does not have independent leadership, adopting the same leadership strategy as the CubeFS main project.
  • Contributions to CubeFS-helm are equivalent to contributions to the main project and can be used as a basis for community role promotion.

Cubefs Components

Cubefs Components

Cubefs Deployment

Cubefs Deployment

Prerequisite

  • Kubernetes 1.14+
  • CSI spec version 1.1.0
  • Helm 3

Download cubefs-helm

git clone https://github.com/cubefs/cubefs-helm
cd cubefs-helm

Create configuration yaml file

Create a cubefs.yaml file, and put it in a user-defined path. Suppose this is where we put it.

vim ~/cubefs.yaml 
# Select which component to install
component:
  master: true
  datanode: true
  metanode: true
  objectnode: true
  client: false
  provisioner: false
  monitor: false
  ingress: true

# store data,log and other data, these directory will be
#  mounted from host to container using hostPath
path:
  data: /var/lib/cubefs
  log: /var/log/cubefs

datanode:
  # Disks will be used by datanode to storage data
  # Format: disk_mount_point:reserved_space
  # disk_mount_point: the mount point of disk in machine
  # reserved_space: similar to metanode reserved space, if disk available
  # space less than this number, then the disk will be unwritable
  disks:
    - /data0:21474836480
    - /data1:21474836480

metanode:
  # Total memory metanode can use, recommended to be configured
  # as 80% of physical machine memory
  total_mem: "26843545600"

provisioner:
  kubelet_path: /var/lib/kubelet

Note that cubefs/values.yaml shows all the config parameters of Cubefs. The parameters path.data and path.log are used to store server data and logs, respectively.

Add labels to Kubernetes node

You should tag each Kubernetes node with the appropriate labels accorindly for server node and CSI node of Cubefs.

kubectl label node <nodename> component.cubefs.io/master=enabled
kubectl label node <nodename> component.cubefs.io/metanode=enabled
kubectl label node <nodename> component.cubefs.io/datanode=enabled
kubectl label node <nodename> component.cubefs.io/objectnode=enabled
kubectl label node <nodename> component.cubefs.io/csi=enabled

Deploy Cubefs cluster

helm upgrade --install cubefs  -f ~/cubefs.yaml -n cubefs --create-namespace cubefs

The output of helm install shows servers to be deployed.

Use the following command to check pod status, which may take a few minutes.

kubectl -n cubefs get pods
NAME                         READY   STATUS    RESTARTS   AGE
cfs-csi-controller-cfc7754b-ptvlq   3/3     Running   0          2m40s
cfs-csi-node-q262p                  2/2     Running   0          2m40s
cfs-csi-node-sgvtf                  2/2     Running   0          2m40s
client-55786c975d-vttcx             1/1     Running   0          2m40s
consul-787fdc9c7d-cvwgz             1/1     Running   0          2m40s
datanode-2rcmz                      1/1     Running   0          2m40s
datanode-7c9gv                      1/1     Running   0          2m40s
datanode-s2w8z                      1/1     Running   0          2m40s
grafana-6964fd5775-6z5lx            1/1     Running   0          2m40s
master-0                            1/1     Running   0          2m40s
master-1                            1/1     Running   0          2m34s
master-2                            1/1     Running   0          2m27s
metanode-bwr8f                      1/1     Running   0          2m40s
metanode-hdn5b                      1/1     Running   0          2m40s
metanode-w9snq                      1/1     Running   0          2m40s
objectnode-6598bd9c87-8kpvv         1/1     Running   0          2m40s
objectnode-6598bd9c87-ckwsh         1/1     Running   0          2m40s
objectnode-6598bd9c87-pj7fc         1/1     Running   0          2m40s
prometheus-6dcf97d7b-5v2xw          1/1     Running   0          2m40s

Check cluster status

helm status cubefs

Use Cubefs CSI as backend storage

After installing Cubefs using helm, the StorageClass named cfs-sc of Cubefs has been created. Next, you can to create a PVC that the storageClassName value is cfs-sc to using Cubefs as backend storage.

An example pvc.yaml is shown below.

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: cfs-pvc
spec:
  accessModes:
  - ReadWriteMany
  volumeMode: Filesystem
  resources:
    requests:
      storage: 5Gi
  storageClassName: cfs-sc
kubectl create -f pvc.yaml

There is an example deployment.yaml using the PVC as below

apiVersion: apps/v1
kind: Deployment
metadata:
  name: cfs-csi-demo
spec:
  replicas: 1
  selector:
    matchLabels:
      app: cfs-csi-demo-pod
  template:
    metadata:
      labels:
        app: cfs-csi-demo-pod
    spec:
      nodeSelector:
        cubefs-csi-node: enabled
      containers:
        - name: cfs-csi-demo
          image: nginx:1.17.9
          imagePullPolicy: "IfNotPresent"
          ports:
            - containerPort: 80
              name: "http-server"
          volumeMounts:
            - mountPath: "/usr/share/nginx/html"
              name: mypvc
      volumes:
        - name: mypvc
          persistentVolumeClaim:
            claimName: cfs-pvc
kubectl create -f deployment.yaml

Config Monitoring System (optional)

Monitor daemons are started if the cluster is deployed with cubefs-helm. Cubefs uses Consul, Prometheus and Grafana to construct the monitoring system.

Accessing the monitor dashboard requires Kubernetes Ingress Controller. In this example, the Nginx Ingress is used. Download the default config yaml file, and add hostNetwork: true in the spec section.

spec:
  # wait up to five minutes for the drain of connections
  terminationGracePeriodSeconds: 300
  serviceAccountName: nginx-ingress-serviceaccount
  hostNetwork: true
  nodeSelector:
    kubernetes.io/os: linux

Start the ingress controller

kubectl apply -f mandatory.yaml

Get the IP address of Nginx ingress controller.

kubectl get pods --all-namespaces -o wide | grep nginx-ingress-controller
ingress-nginx   nginx-ingress-controller-5bbd46cd86-q88sw    1/1     Running   0          115m   10.196.31.101   host-10-196-31-101   <none>           <none>

Get the host name of Grafana which should also be used as domain name.

kubectl get ingress -n cubefs
NAME      HOSTS                  ADDRESS         PORTS   AGE
grafana   monitor.cubefs.com   10.106.207.55   80      24h

Add a local DNS in /etc/hosts in order for a request to find the ingress controller.

10.196.31.101 monitor.cubefs.com

At this point, dashboard can be visited by http://monitor.cubefs.com.

Uninstall Cubefs

uninstall Cubefs cluster using helm

helm delete cubefs