Skip to content

Latest commit

 

History

History
79 lines (67 loc) · 4.28 KB

CHANGELOG-1.6.md

File metadata and controls

79 lines (67 loc) · 4.28 KB

Changes since v1.5.0

Deprecations

  • Support for storage.k8s.io/v1beta1.CSINode and the nodeid annotation will be removed in a future release. Only cluster versions that support storage.k8s.io/v1.CSINode will be supported. The topology feature gate will also be removed and will always be on if the plugin supports the VOLUME_ACCESSIBILITY_CONSTRAINTS capability. (#385, #257)
  • The current leader election lock name will be changed in a future release. Rolling upgrades to a newer sidecar will not maintain proper leadership, and a full teardown and bringup will be required if the driver cannot handle multiple leaders. (#295)
  • The --leader-election-type argument will be removed in a future release. Only Leases-type leader election will be supported. The --enable-leader-election argument will also be renamed to --leader-election. (#401)
  • The default value of --worker-threads will be lowered to 10 to match the other CSI sidecars in a future release. (#322)
  • The external-provisioner will no longer default an empty fstype to ext4. A new option to set the default will be added, otherwise the CSI driver must be able to handle an empty fstype according to the CSI spec. (#328)
  • Already deprecated arguments --connection-timeout and --provisioner will be removed in a future release.

New Features

  • StorageClass.allowedtopologies can specify a subset of the topology keys supported by the driver. (#421, @pawanpraka1)

  • Added a new flag, --cloning-protection-threads which defaults to 1, managing how many threads will simultaneously serve the provisioner.storage.kubernetes.io/cloning-protection finalizer removal (#424, @Danil-Grigorev)

  • finalizer provisioner.storage.kubernetes.io/cloning-protection is now set on the source PVC preventing their removal before the cloning finishes. (#422, @Danil-Grigorev)

  • Adds --extraCreateMetadata flag which, when enabled, will inject parameters onto CreateVolume driver requests with PVC and PV metadata. Injected keys:

    • csi.storage.k8s.io/pvc/name
    • csi.storage.k8s.io/pvc/namespace
    • csi.storage.k8s.io/pv/name (#399, @zetsub0u)
  • Default StorageClass secrets are added:

    • csi.storage.k8s.io/secret-name: ${pvc.name}
    • csi.storage.k8s.io/secret-namespace: ${pvc.namespace}

    Note: Default secrets for storage class feature does work only when both parameters are added. (#393, @taaraora)

Bug Fixes

  • If a CSI driver returns ResourceExhausted for CreateVolume, supports topology, and the StorageClass.volumeBindingMode is WaitForFirstConsumer, then the pod will be rescheduled. This may then result in a CreateVolume retry with a different topology. (#405, @pohly)
  • CSI Migration: Fixes CSI migration issue where PVCs provisioned before migration was turned on could not be deleted (#412, @davidz627)
  • Cloning: ensure source and target PVCs have matching volumeMode. (#410, @j-griffith)

Other Notable Changes