Cisco UCS upgrade plan from 2.1(3a) to 2.2(5b)

haresh.nikumbh · ‎10-19-2015

Hello all,

we are planning to upgrade our vmware env from 5.1 to 6. as per compatibility matrix 2.1 does not support vsphere 6 OS.

is anybody has already done this upgrade? what precaution need to take place before we start?

thanks

haresh n

gkumark · ‎10-19-2015

One of my customer has done this upgrade recently and went on just smooth. This is a straight forward upgrade. Just follow the upgrade guide http://www.cisco.com/c/en/us/td/docs/unified_computing/ucs/sw/upgrading/from2-1/to2-2/b_UpgradingCiscoUCSFrom2-1To2-2.html and you should be good.

harish78us · ‎10-20-2015

Hello Ganesh,

Where do I can confirm if failover is configured for FI, to avoid downtime?

I am not getting option where vnic failure configured on the UCS.

gkumark · ‎10-20-2015

Hi Harish,

Login to FI CLI and run the command "show cluster extended-state" or "show cluster state" command to see if the FI's are in sync.

vNIC failover is configured on the service profile under vNIC.

Hope this helps

harish78us · ‎10-20-2015

Thanks I got it..

Timo Scheller · ‎10-21-2015

Hi,

we did that upgrade in August. We also used the link from above. But we got the info (partner), we had first to upgrade to the latest version of 2.1.3(g or something), then we went directly from that to 2.2.3d (Always use 2 versions down the actual one...) .

One small reminder: Make sure, you have enough space left free on the FI. Remove old FW Versions (only left the last 2). Otherwise you can get stuck while having no more free space.

Had this once, no fun at all....

Then, bring a lot of coffee, wait. Wait a little longer. Stop, don't panik. Wait. And then, after about 20 - 30 min after the failover, your FIs back, cluster is healthy, everything works just fine.

Then wait, again a few to 15 min. until the last errors or warnings are gone.

When this is all done, just went to the blades for the upgrade. But with my experience, i always wait one or two days before going to upgrade the blades.

But all in all, rock solid upgrade. ;-)

HTH,

Timo

haresh.nikumbh · ‎11-03-2015

How much time it took for complete upgrade process?

can you give me estimate time phase wise?

Phase 1 - Infra upgrade

Phase 2 - blade server upgrade

haresh n

Timo Scheller · ‎11-03-2015

Hi,

for Infra-Upgrade, with Downloads, Uploads, Backup-Config etc. i'd say about 1.5 hours.

Of course, while nothing happend.

For each blade, it will take 15 - 20 min, with B200-M3.

I only did the two step upgrade for the Infra, btw. The blades i did directly upgrade after being on 2.2.3d at the Infra from 2.1.3x to 2.2.3x.

HTH,

Timo

rlaplante@bhfs.com · ‎10-20-2015

Double check the cisco upgrade guide. Upgrading to 2.2(5b) is only 'supported' if your are already on 2.1(3g). I had to perform a staged upgrade from 2.1(3c) to 2.1(3g) then to 2.2(5b). Call me chicken but I dont like performing unsupported upgrades.

During the infrastructure upgrade I had a IOM update/activate hang in UCSM process. TAC helped me identify and manually force that one item that was hung to update and the process continued without issue after that.

harish78us · ‎10-20-2015

Hey Russell,

thanks for sharing your experience, and will keep this in mind when we start our upgrade process.

I have tried to find out on cisco upgrade guide but didn't find anything which state from 2.1(3a) to 2.2(5b) not support.I will cross check again with Cisco tech support. earlier tech support said it is single upgrade and can directly go from 2.1(3a) to 2.2(5b).

Keny Perez · ‎10-27-2015

Correct! there is nothing stating that you cannot upgrade from 2.1(3a) to 2.2(5b) in one step, but instead we have this guide http://www.cisco.com/c/en/us/td/docs/unified_computing/ucs/sw/upgrading/from2-1/to2-2/b_UpgradingCiscoUCSFrom2-1To2-2.html showing how to upgrade from 2.1x to 2.2x with the caveat of CSCuh61202:

"If you are upgrading from a release prior to 2.1(3a) or 2.2(1b), you may be impacted by CSCuh61202. This defect affects FC traffic on the Cisco 1240, Cisco 1280, and Cisco M81KR adapters. To avoid this situation and reduce the number of server reboots, you must first upgrade the adapter and server firmware using a host firmware package, and then proceed with the corresponding infrastructure upgrade. This is an exception to the normal upgrade procedure. For more details, refer to https://tools.cisco.com/bugsearch/bug/CSCuh61202. "

Let me know if you have questions.

-Kenny

eglinsky2012 · ‎11-06-2015

We had ESXi 6.0 running fine on 2.1(1e), though that upgrade was done by our service provider who apparently did not check for compatability. We originally had 5.1, then we had to upgrade the management blades to 6.0 in April so we could get large VMDKs. Last week, we went directly from 2.1(1e) to 2.2(5b) without hassle, though I wasn't aware of needing to reverse the order of the upgrade as per CSCuh61202. I was surprised how long it took, maybe that's why. The biggest challenge was figuring out where and how to authorize the Fabric(s) to reboot, since it didn't appear in the pending activities of UCSM GUI. CLI provided no acknowledgement of the command, next thing I knew it was on its way. The servers did appear in the Pending Activities when I started them.

As far as ESXi compatability, this 2.5(5e) document states support for ESXi 5.1 and 6.0 under the "Operating Systems Interoperability Matrix."

http://www.cisco.com/c/dam/en/us/td/docs/unified_computing/ucs/interoperability/matrix/b_hcl_B_rel2-25.pdf

We're still running both ESXi 6.0 and 5.1, no issues since the upgrade.

BTW, the upgrade was in response to one of the fabrics crashing and failing to boot properly. We had to break the bootloader, boot off an image from a TFTP server, and re-write the UCSM, system, and kick start images to the bootflash. After all that, the routine firmware update didn't scare me as much as it should have!