Cabling a new Fabric Attached MetroCluster

About Fabric Attached MetroClusters

A Fabric Attached MetroCluster is a cluster where the disk shelves and nodes are connected through two Fibre Channel switches per node, instead of the disk shelves and nodes connected to each other. The disk shelves and nodes communicate with each other and to the partner nodes and disk shelves through the switches.

A Fabric Attached MetroCluster has the following characteristics:

Four Fibre Channel switches, either the Brocade 3200 or 3800, two switches of the same type per node.
Cluster interconnect traffic passes through the switches.
One VI-MC cluster adapter per node.
Disk shelves are divided into pools.
Disk shelf ownership by a node is determined by where the disk shelves are plugged into the switch and by where the FC HBA is connected to the switch.
Disk shelves are dual pathed to the same ports on different switches on the same side of the cluster, for example, the first disk shelf in Pool 0 is cabled to port 3 on Switch 1 and also to Port 3 on switch 2.

How switches are divided

The Fibre Channel switches are divided into banks and quadrants. Each switch has two banks and four quadrants, dividing the switch into equal parts. For example, a Brocade 3800 switch has 16 ports, divided into two banks of eight ports per bank.

The banks are further subdivided into two quadrants of four ports each (two for the Brocade 3200). These quadrants are also identified as Pools, which function the same in a Fabric Attached MetroCluster as in a mirrored cluster disk pool. Quadrants 1 and 3 belong to Pool 0, while quadrants 2 and 4 belong to Pool 1. Each switch bank contains a Pool 0 and Pool 1.

See the Data Protection Guide for a detailed discussion of disk pools.

The following illustration shows a Brocade 3800 switch, divided into banks, quadrants, and pools.

What a switch fabric is

A Fabric Attached MetroCluster contains two fabrics. A switch fabric consists of one switch connected to a switch on the partner half of the cluster. The two switches are connected to each other through the long distance ISL cables.

For example, the first fabric in a MetroCluster begins from Switch 1 and is completed by connecting the ISL to the first switch on the partner side--Switch 3. The second fabric consists of the second switch, connected through a second ISL cable, to the second switch on the partner side--Switch 4.

How disk ownership is determined

Disk ownership by a node is determined by where the FC HBAs connect into the switch and by where the disks connect to the switch banks. FC HBAs connect to one bank, and own the disks on the other switch bank. For example, if the local node has its FC HBA adapters connect to Bank 1, it owns the disks connected to Bank 2.

Before you begin

To cable a Fabric Attached MetroCluster, you must be familiar with clustering concepts, the Brocade command-line interface, and synchronous mirroring. You must also have the following information:

Unique domain IDs for the switches
Ethernet IP address for both the switches and nodes
Ethernet subnetmask
Gateway address

For detailed information

For detailed information about setting up and cabling your Fabric Attached MetroCluster, see the following sections: