This repository has been archived by the owner on Dec 22, 2018. It is now read-only.
Releases: kubernetes-retired/nfs-provisioner
Releases · kubernetes-retired/nfs-provisioner
v1.0.3
- Fix inability to export NFS shares ("1 validation errors in block FSAL") when using Docker's overlay storage driver (CoreOS/container linux, GCE) by patching Ganesha to use device number as fsid. (#63)
- Adds configurable number of retries on failed Provisioner operations. Configurable as an argument to
NewProvisionController
. nfs-provisioner defaults to 10 retries unless the new flag/argument is used. (#65)
v1.0.2
- Usage demo & how-to for writing your own external PV provisioner added here https://github.com/kubernetes-incubator/nfs-provisioner/tree/master/demo
- Change behaviour for getting NFS server IP from env vars (node, service) in case POD_IP env var is not set when needed. Use
hostname -i
as a fallback only for when running out-of-cluster (#52) - Pass whole PVC object from controller to
Provision
as part ofVolumeOptions
, like upstream (#48) - Filter out controller's self-generated race-to-lock leader election PVC updates from being seen as forced resync PVC updates (#58)
- Fix controller's event watching for ending race-to-lock leader elections early. Now correctly discover the latest
ProvisionFailed
/ProvisionSucceeded
events on a claim (#59)
v1.0.1
v1.0.0
- Automatically create NFS PVs for any user-defined Storage Class of PVCs, backed by a containerized NFS server that creates & exports shares from some user-defined mounted storage
- Support multiple ways to run:
- standalone Pod, e.g. for easy dynamically provisioned scratch space
- stateful app, either as a StatefulSet or Deployment of 1 replica: the NFS server will survive restarts and its provisioned PVs can be backed by some mounted persistent storage e.g. a hostPath or one big PV
- DaemonSet, where each node runs the NFS server to expose its hostPath storage
- Docker container or binary outside of Kube
- Race-to-lock PVCs: when multiple instances are running & serving the same PVCs, only one attempts to provision for a PVC at a time
- Optionally exponentially backoff from calls to Provision() and Delete()
- Optionally set per-PV filesystem quotas: based on XFS project-level quotas and available only when running outside of Kubernetes (pending mount option support in Kube)
Docker image:
quay.io/kubernetes_incubator/nfs-provisioner:v1.0.0