Understanding single and dual pathing in clusters

What a dual-path configuration is

Some clusters have two paths to each of their disks; this is called a dual-path configuration. Fabric Attached MetroClusters and clusters using the SharedStorage functionality are dual-path configurations.

Why understanding your paths is important

If you want to remove a module from your cluster, you need to know whether the path you will disrupt is redundant. If it is, you can remove the module without interfering with the storage system's ability to serve data. On the other hand, if that module provides the only path to any of the disks in your cluster, you must take action to ensure that you do not incur system downtime.

How to determine path status for your cluster

You can determine whether any module in your system provides the only path to any disk by using the storage show disk -p command at your system console. This command displays the following information for every disk in the cluster:

Primary port
Secondary port
Shelf
Bay

Dual-path example output

The following example shows what the storage show disk -p command output might look like for a dual-path cluster:

PRIMARY PORT SECONDARY PORT SHELF BAY

------- ---- --------- ---- ---------

0c.112   A    0b.112    B     7    0

0b.113   B    0c.113    A     7    1

0b.114   B    0c.114    A     7    2

0c.115   A    0b.115    B     7    3

0c.116   A    0b.116    B     7    4

0c.117   A    0b.117    B     7    5

0b.118   B    0c.118    A     7    6

0b.119   B    0c.119    A     7    7

0b.120   B    0c.120    A     7    8

0c.121   A    0b.121    B     7    9

0c.122   A    0b.122    B     7   10

0b.123   B    0c.123    A     7   11

Notice that every disk (e.g. 0c.112/0b.112) has two ports active. This means that you do not need to fail over one system before removing modules from the system.

Attention
Make sure that every disk has two paths. Even in a cluster configured for dual paths, a hardware or configuration problem can cause one or more disks to have only one path. If any disk in your cluster has only one path, you must treat that loop as if it were in a single-path cluster when removing modules.

Example output for a cluster that does not have dual paths

The following example shows what the storage show disk -p command output might look like for a cluster that does not use dual paths:

filer1> storage show disk -p

PRIMARY PORT SECONDARY PORT SHELF BAY

------- ---- --------- ---- ---------

5b.16     B                    1    0

5b.17     B                    1    1

5b.18     B                    1    2

5b.19     B                    1    3

5b.20     B                    1    4

5b.21     B                    1    5

5b.22     B                    1    6

5b.23     B                    1    7

5b.24     B                    1    8

5b.25     B                    1    9

5b.26     B                    1   10

5b.27     B                    1   11

5b.28     B                    1   12

5b.29     B                    1   13

5b.32     B                    2    0

5b.33     B                    2    1

5b.34     B                    2    2

5b.35     B                    2    3

5b.36     B                    2    4

5b.37     B                    2    5

5b.38     B                    2    6

5b.39     B                    2    7

5b.40     B                    2    8

5b.41     B                    2    9

5b.42     B                    2   10

5b.43     B                    2   11

5b.44     B                    2   12

5b.45     B                    2   13

For this cluster, there is only one path to each disk. This means that you cannot remove a module from the cluster, thereby disabling that path, without first performing a takeover.