Linstor is a Software-defined storage a service that orchestrates the creation, life cycle management and integration of DRBD resources with other platforms such as Openstack or Kubernetes. It is maintained by LINBIT, the creators of DRBD.

A typical Linstor cluster consists of two components: the linstor-controller and linstor-satellite services. The controller maintains the configuration and state of the cluster in a database, receives client requests to create and manage resources and orchestrates the execution of those requests by delegating them to satellites. A satellite service runs on a node where Linstor consumes local storage or provides storage to services. Satellites are stateless and receive requests from the controller to realize operations on the node such as creating and mounting volumes or creating DRBD resources.

Linstor uses LVM, LVM thin or ZFS as the storage back-end from which it creates volumes. Those volumes are then assigned to a DRBD resource to be replicated across a set of satellite nodes.

For high availability of the control plane, we have to use an external agent such as Pacemaker to orchestrate the Linstor controller service and configure it to use an external database instead of the default embedded H2 database.

The following table maps between Kubernetes and Linstor concepts:

KubernetesLinstor
podresource
containervolume
deploymentresource-definition
nodenode
labelsaux-properties
storageClassresource-group

When interacting with Linstor, we usually create a resource group, which contains parameters like the storage pool, autoplacement rules and replica count. When we spawn a new resource in this resource group, a new resource definition will be created inheriting the properties from the group. The resource definition will then spawn the resources themselves on the satellite nodes following the autoplacement rules. Linstor always creates a default resource group.

Besides resource groups and definitions, we can also define volume groups and volume definitions to describe how the volumes should be created inside a resource. This enables setting properties such as the size, number and other metadata for volumes. Volume groups must always reference a resource group. Although DRBD supports replicating multiple volumes in a single resource, it is not a very common use case.

References

LINSTOR Is Like Kubernetes, But for Block Devices
https://linbit.com/drbd-user-guide/linstor-guide-1_0-en/#s-linstor-introduction