Using Storage Scale storage

IBM® Storage Scale is a cluster file system that provides simultaneous access to a single file system from multiple compute nodes. In the IBM Spectrum LSF offering, Storage Scale applications running in static compute cluster nodes with high-performance access to a shared data space. For more information, see the IBM Storage Scale product documentation.

The IBM Spectrum LSF offering uses IBM Cloud VPC virtual server instance profiles that are provisioned with instance storage for the Storage Scale storage nodes. By virtue of being on instance storage, this storage option can be used for scratch data use cases. When a storage node virtual server instance is rebooted, the data is preserved. However, when the instance is deleted, the instance storage data is lost. There is no automatic replication to persistent storage, for example, IBM Cloud Object Storage, that is provided in the current implementation, but you can add it on your own if required.

Figure 1. Architecture diagram of an IBM Spectrum LSF cluster using Storage Scale storage on IBM Cloud

Before you begin

Before you begin, make sure to complete the steps for Getting started with IBM Spectrum LSF.

Configuring Storage Scale storage

To configure and use Storage Scale storage, the spectrum_scale_enabled deployment value must be set to "true". The remaining Storage Scale storage parameters ("scale_xx") must be set to deploy the wanted storage cluster configuration.

Checking the Storage Scale file system

Complete the following steps to check the Storage Scale file system:

Log in to a Scale storage node by using SSH. Details are available in the logs output with the following spectrum_scale_storage_ssh_command:
```
ssh -J root@52.116.122.64 root@10.240.128.37
```
Display the gpfs cluster setup on the Scale storage node:
```
/usr/lpp/mmfs/bin/mmlscluster
```
Display the file system size:
```
/usr/lpp/mmfs/bin/mmlsfs all -i
```
Display the file server details. This command can be used to validate file block size (node size in bytes):
```
/usr/lpp/mmfs/bin/mmlsfs fs1
```
Log in to the LSF management host by using SSH:
```
ssh -J root@52.116.122.64 root@10.240.128.41
```
Display the gpfs cluster setup on the compute nodes, which include the management host, management host candidate, and worker nodes:
```
/usr/lpp/mmfs/bin/mmlscluster
```
Create a file on the mount point path, for example, /gpfs/fs1, and verify on the other compute nodes that the file can be accessed.