Cluster-as-a-Service - Shared storage

This article describes how to deploy and use shared storage clusters using JASMIN Cluster-as-a-Service (CaaS).

In this article:

Introduction

CaaS provides shared storage clusters that can be mounted on multiple nodes to provide common storage across all those nodes.

These storage clusters are not intended to be directly consumed by users, but are taken as cluster configuration options by other clusters. In particular, Slurm clusters take a shared storage cluster as a configuration option - the shared storage is mounted on each cluster node for user home directories.

NFS

Network File System (NFS) is a protocol for accessing remote network-attached storage. NFS is also used to refer to the implementation of the protocol in the Linux kernel.

A CaaS NFS shared storage cluster provides a simple NFS server. A volume is attached of the specified size, formatted as an XFS filesystem, mounted at /srv/home and exported with no authentication.

NFS servers do not get an external IP, and so are only accessible from the tenancy's internal network.

Cluster configuration

The following variables are available to configure an NFS cluster:

Description Required? Can be updated?
Identity manager The CaaS Identity Manager that is used to control access to the cluster. Yes No
Volume size The size of the NFS data volume in GB. Yes No
Size The size to use for the NFS server. Yes No

Gluster

Gluster is an open-source software-defined storage platform, maintained by Red Hat, that provides performant distributed network filesystems. It allows volumes to be created on-demand, and has a number of different options for configuring the available and reliability of those volumes.

In CaaS, Gluster clusters are created with a single "Distributed" volume configured to use all the available bricks. In a "Distributed" volume, data is spread across the bricks with no replication or erasure coding for resilience - this means that all the allocated space (i.e. num_nodes x volume_size) is available for storing your files. In a replicated or erasure-coded volume, the available space is (significantly) less than the allocated space due to the overheads of providing resilience.

Glusters servers do not get external IPs, and so are only accessible from the tenancy's internal network.

Cluster configuration

The following variables are available to configure a Gluster cluster:

Description Required? Can be updated?
Identity manager The CaaS Identity Manager that is used to control access to the cluster. Yes No
Volume size The size of the data volume for each node in GB. Yes No
Server size The size to use for the Gluster nodes. Yes No
Number of nodes The number of nodes in the cluster. Must be at least three.

This number can be increased after initial deployment, but decreasing it may cause data to be lost.
Yes Yes
Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.