Exchanging the FIs with 6248

[adrian.rother]

Hi,

we have old Fabric Interconnects and now I want to exchange them with the new 6248 and also I want to exchange the IO-Modules. But all this without downtime of the servers. So what do I have to do that it will work? Is it possible without downtime?

Is it possible to exchange the secondary FI first, then work with diffrent types and firmware until I exchange the primary one?

How can I upgrade the UCS Manager on the new and secondary FI?

Thanks in advance,

Adrian

[abbharga · Accepted solution]

Adrian,

The following document should help you with most of the details:

http://www.cisco.com/en/US/docs/unified_computing/ucs/sw/upgrading/from1.4/to2.0/b_UpgradingCiscoUCSFrom1.4To2.0_chapter_0101.pdf

Thanks!

./Abhinav

[adrian.rother]

Thanks, but is it necessary to Re-Acknowledge the Chassis?

[abbharga]

per the document the only need to re-ack the chassis is when you change the port-mapping on the FI, which yes would need the re-ack.

Also if you replacing the 2104 with a 2208 and increase the number of links from 4 to 8 plus need a port-channel betweent the Fi and the IOm it would require you to re-ack the chassis.

Thanks!

./Abhinav

[adrian.rother]

Thank you very much for the quick answers!

[adrian.rother]

I have one further question.

The build version of the FIs must be the same, so it is not possible to change from 1.4(3l) to 2.0(1w), is it?

Do I have to update the 1.4 Version before?

[abbharga]

Yes you will have to update to version 2.0 since the 62xx and the 22xx hardware is supported only with the 2.0 version. Asdditionally both the FIs should be on the same firmware version in a clsuter.

./Abhinav

[adrian.rother]

But is it possible to run with the 6140 on 1.4 and with the 6248 on 2.0 only for the Exchange.

Now we have two 6240 on 1.4(3l) and the new one are 6248(1w), so is it possible to do the exchange with these two version or is it mandatory to update the old 6240 to 2.0 before?

[abbharga]

Adrian,

It is mandatory for you to upgrade the previous cluster to 2.0 version.

./Abhinav

[Jeremy Waldrop · Accepted solution]

Adrian, I have done this upgrade twice for customers and both were a success with no downtime. One customer had 10 chassis with around 75 blades and we upgraded them from a pair of 6120s to 6248s. Here are some tips, most of this is in the upgraded guide.

• Before you starte make sure everything is multipathed properly on the LAN/SAN side so that when you upgrade the firmware there is no downtime.
• Fist make sure your 6120s are upgraded to any version of 2.x, I would personally go to 2.01w.
• Make sure the new 6248s are upgraded to the exact same build of 2.x, you can do this by created a temporary UCS cluster with only the L1/L2 and management ports connected. After you get the updated erase the config on each so that they are back to factory.
• Like Abhinav stated, the part about re-ack the chassis is not right, we did not do that but we also made sure the chassis port mappings were the same on the 6248s as they were on the 6120s.
• Don't worry about keeping the LAN/SAN uplink ports the same. In both the upgrades I did we moved those around and added a few.
• When the new 6248 joins the existing cluster as the subordinate switch be patient and let it do its thing. It can take a few minutes to stabilize and you will see a ton of faults.
• Make sure you don't enable any chassis ports until you get both the LAN/SAN uplink ports/port channels reconfigured and are up on both ends.
• Once all chassis are backup with all IOM links up and server back plane ports are up do some host side checks to verify LAN/SAN path redundancy.
• For ESXi hosts check in vSphere that all storage paths are up and vmincs are connected
• Check the SAN switch and SAN to make sure all initiators are logged in
• Once all multipathing is confirmed SSH into the cluster and switch the active role over to the new 6248 and then start the process of replacing the other 6120.
• If you are also upgrading the IOM to the 2208s don't do that upgrade during this same time, wait a week to make sure everything is ok then do the IOM upgrades.

[adrian.rother]

Ok, thanks a lot, for your support!

[pnarayan]

The dme running in 2.0 code is different than in 1.4. As abhargava said it is mandatory you upgrade the old cluster to 2.0 first.

What is your FCOE vlan?.

Make sure your production FCOE vlan before you upgrade is other than VLAN 1. Reason being default FCOE vlan in 2.0 is not VLAN 1.

Please read the 2.0 release notes to understand what is new in 2.0

Thanks

Pnarayan

Sent from Cisco Technical Support iPad App