Feature Overview (aka. Goal Summary)

LVMS currently only supports thin provisioning. There is customer demand to also allow for thick provisioning.

Goals (aka. expected user outcomes)

Customers can configure LVM deviceClasses to use LVM thick provisioning, to implement two use cases:

1. leverage maximum IO performance, as thin provisioning does induce a performance penalty.
2. Have guarantee / pre-allocated capacity for PVCs. With thin provisioning, there is a risk a running into disk full situations which might not be acceptable. 

Requirements (aka. Acceptance Criteria):

1. Provide a way to configure thick provisioning for a deviceClass. The simplest option could be to make the LVMCluster.storage.deviceClasses.thinPoolConfig attribute optional (which is currently mandatory).
2. A mixture of thin and thick deviceClasses on a single node must be possible.
3. CSI Snapshots do not work with thick provisioning, as LVM does not support that. When a CSI Snapshot is created for a thick storage class, that has to fail immediately with clear error messages.

Out of Scope

Snapshots with thick provisioning 

Customer Considerations

MicroShift already today might use thick provisioning:

https://access.redhat.com/documentation/en-us/red_hat_build_of_microshift/4.14/html-single/storage/index#microshift-lvmd-config-example-basic_microshift-storage-plugin-overview

Documentation Considerations

Documentation has to be clear on how to configure thick provisioning, and make sure that the reader is aware that CSI Snapshots will not work with that.

Interoperability Considerations

Check with the MicroShift team that whatever we change in LVMS does not impact MicroShift usage of LVMS.