3.3. Planning Guidelines

Designate a management server for interaction and control with Microsoft Azure services. For simple Gluster Storage deployments (single site, single NIC), the management platform can be a Linux server/workstation. For more complex deployments, a Windows desktop with Powershell is recommended.

Build custom images based on Red Hat Enterprise Linux 7 with the Hyper-V drivers included within the initramfs file. Instances will fail to start if these drivers are not present.

Use a virtual network for your Red Hat Gluster Storage nodes.

For geo-replication, deploy a common /etc/hosts file to all nodes or use a shared DNS server.

Pricing for standard storage is based on used capacity. It therefore makes sense to use the maximum size for data disks (1023 GB) and allocate as many as the instance supports at install time to minimize future administration overheads.

If NFS is the preferred way to connect to the Gluster Storage nodes, consider using a D series instance that has a more modern CPU with a higher clock speed.

Use availability sets to group Gluster Storage nodes within a replication set together to enhance availability.

Use mdadm to combine disks to form a larger disk.

Use fewer, larger virtual machines to deliver the highest capacity.

For highly available data access, use a replicated GlusterFS volume with the native glusterfs client.

Use a non-default SSH port for public access to the SSH services running on each of the Gluster Storage nodes (that is, use --ssh with vm create ).