Adding SELinux Documentation

[cniackz]

Objective:

When resuming or suspending a drive in an OpenShift cluster, we may encounter a relabel issue. Various solutions exist for this problem, but here I am documenting the one we believe to be the best approach.

[cniackz]

NET:[VulnCheck / Analysis (pull_request) ] fix is on #901

[Praveenrajmani]

PTAL @cniackz

[ravindk89]

Open Question - should we make this a pre-requisite for OpenShift clusters running SELinux?

[balamurugana]

I have modified the text more appropriately as OpenShift < 4.12 are deprecated and added a complete example.

Suggested change

	## SELinux in OpenShift:
	If you encounter the `relabel failed` error after executing the `suspend` or `resume` commands, you should set `spc_t` at Pod's level or Container's level, for example:
	## SELinux in OpenShift
	In a SELinux enabled system, pod may fail to start due to `relabel failed` error on suspended volume. As the suspended volumes are read-only, Kublet tries to do SELinux relabeling by `lsetxattr` system call to write extended attributes. This issue is fixable by adding `spec.securityContext.seLinuxOptions.type: spc_t` at pod level or container level along with appropriate Security Context Constraints (SCCs) in place. A complete detail is available at https://access.redhat.com/solutions/7025337. The following example shows how to set `spc_t` SELinux settings at pod level:

[balamurugana]

Suggested change

	kind: Pod
	spec:
	securityContext:
	seLinuxOptions:
	type: "spc_t" # This applies to all containers unless overridden
	apiVersion: v1
	kind: Pod
	metadata:
	name: sleep-pod
	spec:
	securityContext:
	seLinuxOptions:
	type: "spc_t" # Setting to fix the issue
	volumes:
	- name: sleep-volume
	persistentVolumeClaim:
	claimName: sleep-pvc
	containers:
	- name: sleep-container
	image: example.org/test/sleep:v0.0.1
	volumeMounts:
	- mountPath: "/mnt"
	name: sleep-volume

[balamurugana]

Suggested change

	Ensure that the appropriate Security Context Constraints (SCCs) are in place, as illustrated in the following link: https://access.redhat.com/solutions/7025337.
	Additionally, this issue has been observed in OpenShift version 4.12 and later. Pods that use suspended drives or volumes may fail due to the kubelet attempting to perform SELinux relabeling. This happens because the `lsetxattr` system call fails on a read-only filesystem. The problem can affect any pod that consumes a suspended volume, not just MinIO Tenant pods.