pvc-autoresizer

README

NOTICE

This repo is forked from topolvm/pvc-autoresizer .This one adds storage class level control and automatic restart of workloads on top of that.

pvc-autoresizer

pvc-autoresizer resizes PersistentVolumeClaims (PVCs) when the free amount of storage is below the threshold. It queries the volume usage metrics from Prometheus that collects metrics from kubelet.

Status: beta

Target CSI Drivers

pvc-autoresizer supports CSI Drivers that meet the following requirements:

• implement Volume Expansion.
• implement NodeGetVolumeStats

Prepare

pvc-autoresizer behaves based on the metrics that prometheus collects from kubelet.

Please refer to the following pages to set up Prometheus:

• Installation | Prometheus

Installation

Specify the Prometheus URL to pvc-autoresizer argument as --prometheus-url.

pvc-autoresizer can be deployed to a Kubernetes cluster via helm:

helm install --create-namespace --namespace pvc-autoresizer pvc-autoresizer pvc-autoresizer/pvc-autoresizer --set "controller.args.prometheusURL=<YOUR PROMETHEUS ENDPOINT>"

See the Chart README.md for detailed documentation on the Helm Chart.

How to use

pvc-autoresize

To allow auto volume expansion, the StorageClass of PVC need to allow volume expansion and have resize.kubesphere.io/enabled: "true" annotation. The annotation may be omitted if you give --no-annotation-check command-line flag to pvc-autoresizer executable.

kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: csi-qingcloud
  annotations:
    resize.kubesphere.io/enabled: "true"
provisioner: csi-qingcloud
allowVolumeExpansion: true

To allow auto volume expansion, the PVC to be resized need to specify the upper limit of volume size with the annotation resize.kubesphere.io/storage_limit. The PVC must have volumeMode: Filesystem too. You can also set the upper limit of volume size with .spec.resources.limits.storage, but it is deprecated. If both are present, the annotation takes precedence.

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: test-pvc
  namespace: default
  annotations:
    resize.kubesphere.io/storage-limit: 100Gi
spec:
  accessModes:
  - ReadWriteOnce
  volumeMode: Filesystem
  resources:
    requests:
      storage: 30Gi
  storageClassName: csi-qingcloud

The PVC can optionally have resize.kubesphere.io/threshold and resize.kubepshere.io/increase annotations. (If they are not given, the default value is 10%.)

When the amount of free space of the volume is below resize.kubesphere.io/threshold, .spec.resources.requests.storage is increased by resize.kubesphere.io/increase.

If resize.kubesphere.io/increase is given as a percentage, the value is calculated as the current spec.resources.requests.storage value multiplied by the annotation value.

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: kubesphere-pvc
  namespace: default
  annotations:
    resize.kubesphere.io/storage-limit: 100Gi
    resize.kubesphere.io/threshold: 20%
    resize.kubesphere.io/increase: 20Gi
spec:
  <snip>

workload-autoRestart

The restarter judges the workload that needs to be restarted automatically by checking the status of pvc, and the restarter will stop the workload after successful expansion or timeout and then turn it on again. If you have a storage class that only supports offline expansion, and need to automatically complete the expansion, you can add the following annotations:

kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: csi-qingcloud
  labels:
    app.kubernetes.io/managed-by: Helm
  annotations:
    restart.kubesphere.io/enabled: 'true'
    restart.kubesphere.io/online-expansion-support: 'false'
    restart.kubesphere.io/max-time: '300'

The workload-autoRestart will run when restart.kubesphere.io/enabled is true and restart.kubesphere.io/online-expansion-support is false. restart.kubesphere.io/max-time defines the maximum number of seconds that can be waited for restarting a workload. If the time is exceeded and the restart is not successful, the workload will be skipped afterwards.

Add the following annotations to Deployment or StatefulSet that do not require automatic restart:

kind: 
apiVersion: apps/v1
metadata:
  name: kubesphere-workload
  namespace: default
  annotations:
    restart.kubesphere.io/skip: "true"

Container images

Container images are available on Dockerhub