removing cascaded FC switch and preserving configuration?

[AccuMegalith]

We have a C7000 blade enclosure with a Brocade FC module that's uplinked to a Brocade 4/8 rackmount switch. Our EVA 4100 controllers are connected to the rackmount switch. The switch heirarchy is set so that the rackmount is the root. We only have one of each switch, so that's a single point of failure. The only non-blade connection is the management/CV-EVA server, and we have enough free external ports on the blade module to attach it.

We're about to get another FC blade module (freed up by mirating another blade installation to VC), but can't really justify another rackmount switch to have idetical configurations on both fabrics.

There appears to be no problem with connecting the EVA controllers and management server to the blade module external ports and dispensing with the rackmount switch entirely. I'm not sure what happens to the zoning configuration if we remove the root switch. The way an HP engineer explained it to me over a year ago is that the configuration data are stored on all the switches in the fabric, but we didn't discuss it in any detail.

Is that the case? Does the FC blade module have its own copy of the configuration? Can we remove the rackmount switch and plug everything into the blade switch without any configuration changes?

[Patrick Terlisten · Accepted solution]

Hello,

there is not really a "root switch" in a fabric. There is a switch role called "principal" which is doing some more work then the other switches (zoning changes, time sync etc.), but every switch in a fabric can be the principal. If you remove the B-Series 4/8 from you fabric, your zoning configuration will be remain on the other switches.

So add the second FC module to your C7000, attach the EVA to the both FC modules and remove the B-Series 4/8. This all can be done online.

Best regards,
Patrick

[AWilliams_1]

You would want to check first that the zoning is based on WWN and not switch/port based. If port based then it would be best to first of all change it to WWN (soft) zoning (This can be done online). Then, when the EVA connections are moved to the C-Class switch, the zones will remain intact. If you then add the second C-Class switch to the rack switch it will also get a copy of the zones (be careful about domain ID's). When you remove the rack switch the same zoning info will be on each C-Class switch and can then be cleaned up. Its not particularly complicated, but I would recommend that you find someone with brocade experience to help you.

[JaMeS_311]

hi,

if you have HP-UX servers attached to this SAN, you might also want to check the effect on these servers should you change the EVA connection on the SAN.

HP-UX systems recognize the Domain ID and port of the devices attached to the SAN (described in hardware path), so if you shift your EVA controllers to another switch with a different Domain ID/port, the corresponding device files will also change. This applies to SAN-attached tape libraries too.

This should not be an issue with Windows.

You should check before deploying any changes is to check compatibility of the new switch firmware with the existing ones.

As for zoning, all switches store the zoning configuration. i prefer creating aliases for each connection to help identify connections better and then create zones with these aliases. with aliases you are still not binding a connection to a physical port.

i hope this helps, good luck.

:)

[AccuMegalith]

I guess I should've mentioned that all our zoning is soft zoning, and we have no HP-UX hosts.

Last Thrusday we moved the FC cables to the blade switches without incident, and have decomissioned the rackmount switch. Our management (CV-EVA) server refused to see the active controller, but could after we restarted the server. The boot-from-SAN blades all stayed up. Generally uneventful.

[AccuMegalith]

ummm...