Production deployments

External key-value store

StorageOS uses a key-value store to store cluster configuration. For ease of use, etcd is embedded into the container image.

If you are already familiar with running etcd clusters, you may prefer to run etcd yourself. Doing so will give you greater flexibility in how and where it is deployed, allow custom tuning, and may be easier to integrate into operational tools and processes.

Deploying compute-only nodes

As your cluster scales, you may wish to reserve specific nodes for storage and other nodes for application workloads.

By default, StorageOS nodes both present and consume storage (storageos.com/deployment=mixed). You can specify that a node should only consume storage by setting the storageos.com/deployment=computeonly label.

You can specify that a node should only consume storage by setting the storageos.com/deployment=computeonly label on the node. This can be done by setting the label on the StorageOS container at startup (LABELS=’storageos.com/deployment=computeonly’), or by applying the label once it is running. Note that volumes must be drained off the node first.

Rules

Rules are used for managing data policy and placement using StorageOS features such as replication, QoS and compression.

Rules are defined using labels and selectors. When a volume is created, the rules are evaluated to determine whether to apply the action.

An example business requirement might be that all production volumes are replicated twice. This would be defined with a selector env==prod, and the action would be to add the label storageos.com/replicas=2.

Rules can be created with the CLI or Web UI. See Rules for details.