MSA Storage

Best Practices for directyl attached FC Storage to HP 4gb SAN Switches (Brocade)


Best Practices for directyl attached FC Storage to HP 4gb SAN Switches (Brocade)


We have two of these HP 4GB SAN Switchs for our C3000, Blades have FC HBAs installed in Mezz2, WWN are shown in the switch console.
I Plan to connect our FC Storage directly to these switches. Because I'm totally new to FibreCHannel I have a few design/config questions:

1) Actually booth switchs are configured independly, does it make sense, to connect booth switches by one or two external uplinks? I already tried it, but I have no Idea what to do then. The javainterface shows some strange behaviour then: It sometimes shows the switches in the fabric tree under fabric, sometimes one of the switches is listed in the "segmented switch" tree.

2) How do I get booth switches into one (Admin Domain?).
Till now each are configured for their own, but I read in some guides, it's possible to have somethink like a shared config/cluster, but there is no guide howto do that. Furthermore it's hard to find any good configuration guide with examples, like the virtual connect guides :).

Thanks in Advance & Warm regards :)


Johan Guldmyr
Honored Contributor

Re: Best Practices for directyl attached FC Storage to HP 4gb SAN Switches (Brocade)



the norm is to have the two switches separate - in two separate and independent fabrics.


The storage controllers would be connected to both switches/fabrics and if the servers have two fc ports they would also be connected to both fabrics.


This provides redundancy and fault isolation. Servers can use multipathing software to only use one fabric at a time, meaning you can do maintenance on one fabric at a time as well.


It also means each fabric/switch is administrered independently, so some more work. If you use WWN zoning the zoning configs will be different.


If you still want to join the switches into one fabric by connecting the switches there are a few pre-requisites, some errors can be seen in the output of the CLI command "switchshow". Perhaps you could see also why in the Java GUI portinfo by looking at the E_Ports where the switches are connected.

For example unique Domain IDs and identical zoning db (or one of them has no zoning) on the switches is required for switches to join into one fabric. 


If the SAN ever needs to be expanded with more switches it's easier to do so if the switches are in independent fabrics.