Skip to main content

Storage in AKS

Use Azure Disks for databases. Azure Files for shared storage. Blob for large datasets. Everything else is a special case.

Storage classes (built-in)​

AKS ships with these storage classes pre-configured. Don't create your own unless you need custom parameters.

Storage ClassBackendAccess ModeUse Case
managed-csiAzure Disks (Premium LRS)ReadWriteOnceDatabases, single-pod stateful apps
managed-csi-premiumAzure Disks (Premium LRS)ReadWriteOnceSame as above, explicit premium
azurefile-csiAzure Files (Standard)ReadWriteManyShared config, CMS content
azurefile-csi-premiumAzure Files (Premium)ReadWriteManyShared storage needing IOPS
azureblob-nfsBlob NFSReadWriteManyLarge datasets, ML training data
Opinion

Use managed-csi (Azure Disks) as your default for anything stateful. Only reach for Azure Files when multiple pods need simultaneous read/write access to the same data.

PersistentVolume lifecycle​

PersistentVolume Lifecycle

Always use dynamic provisioning unless you have pre-existing disks to import. Dynamic provisioning creates the Azure resource automatically when a PVC is submitted.

Dynamic provisioning example​

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: postgres-data
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: managed-csi
  resources:
    requests:
      storage: 100Gi

That's it. AKS creates a Premium SSD managed disk, attaches it to the node running your pod, and mounts it. No manual disk creation needed.

Reclaim policies​

PolicyBehavior on PVC DeleteUse When
DeleteDisk/share is destroyedEphemeral workloads, dev/test, caches
RetainDisk/share is preserved (orphaned)Production databases, data you cannot lose
warning

The default reclaim policy for managed-csi is Delete. If you delete the PVC, your disk and all data is gone. For production databases, create a custom StorageClass with reclaimPolicy: Retain.

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: managed-csi-retain
provisioner: disk.csi.azure.com
parameters:
  skuName: Premium_LRS
reclaimPolicy: Retain
volumeBindingMode: WaitForFirstConsumer
allowVolumeExpansion: true

Access modes​

ModeMeaningSupported By
ReadWriteOnce (RWO)Single node read/writeAzure Disks
ReadOnlyMany (ROX)Multi-node read-onlyAzure Disks, Azure Files
ReadWriteMany (RWX)Multi-node read/writeAzure Files, Blob NFS
info

Azure Disks are block devices -- they physically attach to one node at a time. If you need multiple pods on different nodes writing to the same volume, you need Azure Files or Blob NFS.

Common mistakes​

  1. Using Azure Files for databases -- Azure Files has higher latency than Disks. Use Disks for anything IOPS-sensitive.
  2. Forgetting volumeBindingMode: WaitForFirstConsumer -- Without this, the disk may provision in a zone where no node can mount it.
  3. Not setting allowVolumeExpansion: true -- You will need to resize disks. Enable this upfront.
  4. Using Delete reclaim policy for production data -- One accidental kubectl delete pvc destroys your database.

Resources​