Deploy Storage Nodes

After installing the simplyblock operator, the next step is to create a storage cluster, deploy storage nodes, create a storage pool, and enable volume provisioning via the CSI driver.

Info

In a Kubernetes deployment, not all Kubernetes workers have to become part of the storage cluster. Simplyblock uses node labels to identify Kubernetes workers that are deemed as storage hosting instances.

It is common to add dedicated Kubernetes worker nodes for storage to the same Kubernetes cluster. They can be separated into a different node pool, and using a different type of host. In this case, it is important to remember to taint the Kubernetes worker accordingly to prevent other services from being scheduled on this worker.

OpenShift Prerequisites

If you are deploying onto an OpenShift cluster, ensure that the environment-specific instructions provided in the OpenShift Installation guide are followed.

Networking Configuration

Multiple ports are required to be opened on storage node hosts.

Ports using the same source and target networks (VLANs) will not require any additional firewall settings.

Opening ports may be required between the control plane and storage networks as those typically reside on different VLANs.

Service	Direction	Source / Target Network	Port(s)	Protocol(s)
ICMP	ingress	control	-	ICMP
Storage node API	ingress	storage	5000	TCP
spdk-firewall-proxy	ingress	storage	5001	TCP
spdk-http-proxy	ingress	storage, control	8080-8180	TCP
hublvol-nvmf-subsys-port	ingress	storage, control	9030-9059	TCP
internal-nvmf-subsys-port	ingress	storage, control	9060-9099	TCP
lvol-nvmf-subsys-port	ingress	storage, control	9100-9200	TCP
FoundationDB	egress	storage	4500	TCP
Control plane API	egress	control	80	TCP

Creating a Storage Cluster

Once the operator is running, create a storage cluster by applying a StorageCluster CRD:

Example: storage-cluster.yaml

apiVersion: simplyblock.simplyblock.io/v1alpha1
kind: StorageCluster
metadata:
  name: my-cluster
  namespace: simplyblock
spec:
  clusterName: production
  mgmtIfname: eth0
  haType: ha
  stripe:
    dataChunks: 2
    parityChunks: 1
  fabric: tcp

Apply the cluster resource

kubectl apply -f storage-cluster.yaml

Check the cluster status:

Check cluster status

kubectl get simplyblockstoragecluster -n simplyblock

Cluster Options

For NVMe-oF transport security, backup configuration, and other cluster options, see Cluster Deployment Options.

Deploying Storage Nodes

Apply a StorageNode CRD to deploy storage nodes:

Example: storage-nodes.yaml

apiVersion: simplyblock.simplyblock.io/v1alpha1
kind: StorageNode
metadata:
  name: storage-nodes
  namespace: simplyblock
spec:
  clusterName: production
  clusterImage: "public.ecr.aws/simply-block/simplyblock:26.1.2"
  maxLogicalVolumeCount: 100
  workerNodes:
    - worker-1
    - worker-2
  maxSize: "500G"
  partitions: 1
  coreIsolation: true

Apply the storage node resource

kubectl apply -f storage-nodes.yaml

Storage Node Parameters

Parameter	Description	Default
`clusterName`	Name of the cluster this node belongs to. Required.
`clusterImage`	Storage-node image. Required when `action` is not specified.
`maxLogicalVolumeCount`	Maximum number of logical volumes on this node. Required when `action` is not specified.	10
`maxSize`	Maximum utilized storage capacity (e.g., `500G`). Impacts RAM demand.	150g
`partitions`	Number of partitions per device.	1
`coreIsolation`	Enable CPU core isolation. Requires a node restart after deployment.	false
`corePercentage`	Percentage of total cores allocated to simplyblock.
`pcieAllowList`	List of allowed NVMe PCIe addresses.
`pcieDenyList`	List of blocked NVMe PCIe addresses.
`dataIfname`	Data network interface names for storage traffic.
`socketsToUse`	Number of NUMA sockets to use.
`nodesPerSocket`	Number of storage nodes per NUMA socket.
`workerNodes`	Worker node names for deployment. Required and must be non-empty when `action` is not specified.

For a complete list of fields, see Simplyblock Operator.

Creating a Storage Pool

Apply a Pool CRD to create a storage pool:

Example: storage-pool.yaml

apiVersion: simplyblock.simplyblock.io/v1alpha1
kind: Pool
metadata:
  name: my-pool
  namespace: simplyblock
spec:
  name: production-pool
  clusterName: production
  capacityLimit: "10T"

Apply the storage pool resource

kubectl apply -f storage-pool.yaml

Verification

Check the status of all simplyblock resources:

Verify deployment status

kubectl get simplyblockstoragecluster -n simplyblock
kubectl get simplyblockstoragenode -n simplyblock
kubectl get simplyblockpool -n simplyblock
kubectl get pods -n simplyblock

Multi-Cluster Storage Node Support

A single Kubernetes cluster can host storage nodes connected to multiple simplyblock clusters. To configure this, specify the workerNodes field in the StorageNode CRD:

Multi-cluster storage nodes

apiVersion: simplyblock.simplyblock.io/v1alpha1
kind: StorageNode
metadata:
  name: cluster-a-nodes
  namespace: simplyblock
spec:
  clusterName: cluster-a
  workerNodes:
    - worker-a-1
    - worker-a-2
---
apiVersion: simplyblock.simplyblock.io/v1alpha1
kind: StorageNode
metadata:
  name: cluster-b-nodes
  namespace: simplyblock
spec:
  clusterName: cluster-b
  workerNodes:
    - worker-b-1
    - worker-b-2

Warning

The resources consumed by simplyblock are exclusively used and have to be aligned with resources required by other workloads. For further information, see minimum hardware requirements.

Info

The RAM requirement is split between huge page memory and system memory. Simplyblock manages huge page allocation automatically.

The total amount of RAM required depends on the number of vCPUs used, the number of active logical volumes (Persistent Volume Claims or PVCs) and the utilized virtual storage on this node.