Brocade 4/12 c-Class SAN Switch in one Fabric

[JuanCG]

Hello everyone, I have two Brocade SAN switches each with DOMAIN ID 1 is not connected and provide service to servers.

I have to unite for the server that makes fiber backup to another server to see this on the other switch.

¿Me podeis INDICATE WHAT IS THE PROCEDURE?

THANKS !!!

[Richard Tengdin]

Why do your servers need to see each other? Normally a backup server only needs to see the tape library and its private disks, or the ESX shared storage for VMware Consolidated Backup... I'm not aware of any backup environment where the servers actually need to see each other.

Here is the process, but I have some cautions at the end...

1. Back up the exising zone configuration on one of the two switches (cfgshow and cut/paste to a text file is OK)
2. Clear the active configuration on the switch
3. Disable the switch and set the Domain ID to 3 (we prefer odd/even IDs to identify a switch fabric)
4. Connect the Inter-Switch Links (ISLs). The switches should merge into one fabric
5. Re-create the zones backed up in Step 1 above on the new fabric.

Extending your fabric this way is appropriate if you will still have two separate fabrics at the end.

If you create a single fabric across both SAN switches you have removed one layer of protection against losing access to your data:
*** The common zone configuration is a SINGLE POINT OF FAILURE ***

[JuanCG]

Richard thank you very much for your contribution, you said:

SPOF applies to this configuration, since there are 4 switches at the end there will be 3 Fabric isolated. The reason is that to make the Data Protector backup fiber, is linked to the server backup, NSR andalusia (Library) and local storage of VMware virtual center where copies are kept of VMware Consolidated Backup.

The procedure that tells me is as I had thought.

The other question I have is that versions of Fabric OS (Brocade switch) are different is a v5.2.1by the other switch is v6.1.0.a.

And another question I have is that if you need to stop, and you will see some instant in which the servers are going to have to failover to another HBA and maybe lose information DDBB Oracle, this is what worries me most .

Should we stop systems?

THANK YOU

[Richard Tengdin · Accepted solution]

JuanCG asked: The other question I have is that versions of Fabric OS (Brocade switch) are different is a v5.2.1by the other switch is v6.1.0.a.

Answer: You need to upgrade all the connected switches to the same FOS version before you merge the fabrics. This is totally non-disruptive now (the switch FOS reset does not affect data flow).


JUANCG asked: And another question I have is that if you need to stop, and you will see some instant in which the servers are going to have to failover to another HBA and maybe lose information DDBB Oracle, this is what worries me most .

Should we stop systems?

Answer: If you just connect the two switches without resetting the Domain ID on one you can still merge the fabrics but you will not have control over which switch moves to a different Domain ID. The process of merging the fabrics should not stop data flow so Oracle should not be affected.

For safety's sake you might use the multipath tools for your host operating system to switch I/O traffic to the switches you will not be working on, that way any issues during the fabric merge will not affect production.

You don't have to stop the systems unless your management policies require it.

[JuanCG]

Thanks