Skip to main content

Persistent storage

Ce que nous savons actuellement sur les conteneurs:

  • les conteneurs sont éphémères
  • Les données écritent dans un conteneur sont détruites lors de sa suppression
  • Quand un Pod est supprimé, les conteneurs sont supprimés de la Node
  • Comment pouvons nous persisté de la données entre les différents cycle de vie des Pods ?

Storage API Object​

Kubernetes nous donne des objects permettant de persisté de la donnée:

  • Volume - Stockage persistant qui devient une part du pod (dans les specs)
  • Persistent Volume - Stockage de la donnée en elle même, stockage pour les pods utilisé dans le cluster
  • Persistent Volume Claim - Demande faite par un utilisateur de persister un volume
  • Storage Class - Un moyen de créer des groupes ou classes de stockage disponible dans le cluster

Volume​

  • Persistent storage deployed as part of the Pod spec
  • Implementation details for your storage
  • This can be challenging...
    • Sharing code (very restritive with the Pod itself)
    • Same lifecycle as Pod
  • We can do better...

Persistent Volumes​

  • Administrator defined storage in the Cluster
    • Is the responsibility of the cluster administrator to create and delete storage
    • This is going to be an API object that represents the actual storage itself
  • Implementation details for your storage
  • Lifecycle independent of the Pod
  • Managed by the Kubelet
    • Maps the storage in the Node
    • Exposes Persitent Volume as a mount inside the container - like any other Linux device

For more details https://kubernetes.io/docs/concepts/storage/persistent-volumes/

Types of Persistent Volumes​

For simplify the different types of persistent volumes:

Type of Persistent Volumes

It isn't an exhaustive list, a more complete list is available here: https://kubernetes.io/docs/concepts/storage/persistent-volumes/#types-of-persistent-volumes

Persistent Volumes Claim​

  • A request for storage by a user - is a resquest to the cluster for some amount of storage
    • Size
    • Access Mode - specific node, read-only /read-write
    • Storage Class
  • Enable portability of your application configurations
  • The Cluster will map a PVC to a PV

Access Mode​

In PV & PVC we'll need to define an access mode. Wich controls how a node (or multiple nodes) access a PV.

Because a PV & PVC are mapped to the Node by the kubelet and the exposed into a Pod.

They are 3 access modes:

  • ReadWriteOnce (RWO)
    • One Node can mount a volume for read & write access
  • ReadWriteMany (RWX)
    • More than one Node can be access to the volume for read & write access
  • ReadOnlyMany (ROX)
    • More than one Node can be access to the volume for readOnlyAccess

The volumes are Node level access, not Pod level access

Static Provisioning Workflow​

  • Define a persistent volume - detail for access (ex: NFS, server adress), the size in the spec
  • Define a persistent volume claim - is the resquest for storage in the cluster - define the size, access mode and storage class
  • Define Volume in Pod spec

Storage Lifecycle​

Storage Lifecycle