ZFS Clonning on the Linstor/DRBD9?

[SVronskiy]

Hello!

We perform development with use of one Dev and many Feature environments. Dev environment files/databases are located on ZFS filesystem and exported by shareNFS to Kubernetes pods via dynamic NFS volumes. The task is to automatically create new Feature environments by cloning all files/databases in seconds, then running manual or autotests in that environment. Each Feature enviroment has it own Kubernetes namespace with pods identical to Dev environment pods. Performance of running tests doen't matter in this case, the gole is to make new enviroment avaliable for tests as fast as possible.

My idea is to join HA oportunities of Linstor/DRBD, with ZFS clonning features and Kubernetes balancing.

I'de like to create Kubernetes Pod similar to docker nfs server, but based on top of Linstor PVC and with ZFS cloning oportunities.

The question is how to get working ZFS filesystem on the Linstor persistent volume claim, to be availiable in Kuberntes Pod with purpose to make clone of existing data and export clone using ZFS shareNfs?

Regards, Sergey

[w00jay]

Hi SVronskiy,

Our LINSTOR CSI plugin (https://github.com/LINBIT/linstor-csi) allows use of LINSTOR volumes on Kubernetes. The CSI plugin functions mostly as a 'pass-through' plumbing between Kubernetes and LINSTOR. LINSTOR does support ZFS volumes and therefore, our CSI plugin should support your deployment scenario. Cloning in CSI plugin is done by spawning a new volume after taking a snapshot, and those underlying operations are supported in both ZFS and LINSTOR.

Our DRBD User Guide (https://docs.linbit.com/docs/users-guide-9.0/#s-kubernetes-overview) has a chapter on LINSTOR deployment in k8s environment for in-depth explanation, and the LINSTOR CSI plugin repo has a few example YAML files (https://github.com/LINBIT/linstor-csi/tree/master/examples/k8s) for trying it out.

Let us know if you would like any further information.