rolling updates with persistent storage require RWX

[C0nsultant]

Currently, all peristent storage volumes are RWO. This works fine with update strategy recreate, but fails intermittently for rollingUpdate at least (when the pod is is on a different node, since the volume obviously would need to be bound on both the old and the new nodes simultaneously until the new pod is healthy).

Possible options:

• make sure all deployments use an update strategy compatible with RWO
• switch all relevant PVs and PVCs to RWX

Not idential, but related to the Node Failure scenarios in longhorn.

[C0nsultant]

The performance differences between RWO and RWX in longhorn are quite stark. Using RWX for everything is probably not the way to go.