Make seLinuxMount optional in CSIDriver for compatibility

[Sebastian-RG]

Describe the bug

According to https://kubernetes.io/docs/reference/command-line-tools-reference/feature-gates/ the seLinuxMount field in the CSIDriver object is introduced as alpha in K8S 1.25, made beta in 1.27 and only starting from 1.28 defaults to true.
If this feature gate is not enabled, ceph-csi-rbd and ceph-csi-fs fail when installing.
I would like for seLinuxMount to be optional but true by default, to have better compatibility

This was enabled on tag 3.10.2. Previous tags did not have this issue. See PR #4346

Environment details

• Image/version of Ceph CSI driver :
• Helm chart version : v3.10.2
• Kernel version :
• Mounter used for mounting PVC (for cephFS its fuse or kernel. for rbd its
  krbd or rbd-nbd) : krbd
• Kubernetes cluster version : v1.25.6
• Ceph cluster version :

Steps to reproduce

Steps to reproduce the behavior:

1. Deploy ceph-csi-rbd on a cluster without the feature flag SELinuxMountReadWriteOncePod enabled
2. Chart will fail while deploying due to an error: unknown field "seLinuxMount" in io.k8s.api.storage.v1.CSIDriverSpec

Actual results

Chart failed to install

Expected behavior

Chart installs correctly or shows a descriptive error about the feature flag

[nixpanic]

Helm chart version : v3.10.2
Kubernetes cluster version : v1.25.6

Note that Ceph-CSI 3.10.x is not tested with Kubernetes 1.25.x. Ceph-CSI 3.10.x expects Kubernetes 1.26 or newer.

The Ceph-CSI CI runs tests on Kubernetes 1.26, so I wonder why there were no issues with deploying there. Maybe @Rakshith-R has an idea?

[Rakshith-R]

Helm chart version : v3.10.2
Kubernetes cluster version : v1.25.6

Note that Ceph-CSI 3.10.x is not tested with Kubernetes 1.25.x. Ceph-CSI 3.10.x expects Kubernetes 1.26 or newer.

The Ceph-CSI CI runs tests on Kubernetes 1.26, so I wonder why there were no issues with deploying there. Maybe @Rakshith-R has an idea?

In k8s 1.26, it seems k8s accepts that option but it doesn't show up if the feature flag is not enabled.

As Niels already mentioned 3.10.x expects k8s 1.26+ .
Please use 3.9.x or upgrade k8s version.
https://github.com/ceph/ceph-csi?tab=readme-ov-file#known-to-work-co-platforms

[dragoangel]

Helm chart version : v3.10.2
Kubernetes cluster version : v1.25.6

Note that Ceph-CSI 3.10.x is not tested with Kubernetes 1.25.x. Ceph-CSI 3.10.x expects Kubernetes 1.26 or newer.
The Ceph-CSI CI runs tests on Kubernetes 1.26, so I wonder why there were no issues with deploying there. Maybe @Rakshith-R has an idea?

In k8s 1.26, it seems k8s accepts that option but it doesn't show up if the feature flag is not enabled.

As Niels already mentioned 3.10.x expects k8s 1.26+ . Please use 3.9.x or upgrade k8s version. https://github.com/ceph/ceph-csi?tab=readme-ov-file#known-to-work-co-platforms

Hi @Rakshith-R I can't agree, as helm allows easily develop backward compatible charts and except this issue there actually no issues in running ceph-csi 3.10.x on K8s 1.24.x for example. The fact that soft was or not was tested doesn't mean we should accept breaking changes for non-tested versions just in helm when helm can be easily backward compatible.

Same approach can be found in many other public helm charts like prometheus, thanos, etc...
They say: we tested version X,Y,Z of our software (it's usually about CDR controllers) over version X of kube, but this doesn't mean that we can't setup version X of software on older version of kube, and their helm charts properly exclude unsupported stuff or properly set correct versions of api's for resources as different versions of Kubernetes has different APIs.