Need Help: MDS9509 firmware upgrade

you need to find a server where you will run Fabric Manager from ..as it will no longer reside on the switch itself. That aside ..have you looked in E-Lab to make sure all other components in your SAN are compatible with this new version (HBA drivers/firmware, failover software, array flare/microcode levels ..etc)

Beyond what dynamox pointed out, unless you have special modules in your 9509 like the SSM or ASM, the upgrades are non disruptive and actually go fairly quickly. The dual supervisor of the 9509 make upgrades fairly painless and will be transparent to the hosts and arrays since the version of the switch OS is transparent to the hosts and arrays.

That being said, there have been a few notices from Cisco about problems with 9500 series upgrades, there were some issues with flash modules that I recall. EMC can check your switches to determine if these notices apply to your switches.

But EMC should ensure that there are also no incompatible components in your infrastructure. If they have not checked, ask them to prior to the upgrade.

This is common sense, but make sure you also only upgrade one fabric at a time. We had an issue when we had our MDS 9513 (4 total) updated to version 3.02 that resulted in one of the blades on the chassis hard resetting itself.

Since all but 1 server were dual connected there was no harm other than a lot of powerpath error messages. There was an alert on the issue we had and it was on one of the earlier versions of 3.02 and I believe has been resolved in later releases.

Related note - checking compatability is critical as well. We did have a case where the combination of the initial Cisco code we had (can't recall exact version), Solutions Enabler, and AIX without a certain patch could result in syminq hanging and possibly crashing the host.

We ran across this after having a host with a syminq command that hung and luckily were able to get the correct patches and firmware installed prior to having issues.

But once again - both you and EMC need to do your homework before upgrading

Thank you for the additional information.
I will have to check out all the known problems and confirm them with EMC...
It's my first time to arrange such task.. tough but I am learning many things!

Thank you so much for the advices!
Actually, it's the first project I am in charge (!!!), it's pretty stressful...
especially when our servers are not all installed with powerpath... I am really worried...

Make sure you leverage EMC to assist you in analyzing your environment prior to the upgrade. The maintenance contracts for the switches cover this service and they have the expertise to help you through this. You should have been assigned a project manager for this upgrade right? That person will help you get the resources you need to confirm that your environment is ready for an upgrade.

I actually was in the same boat you are in now. I just lead a project to upgrade 3 of our 9509s from 3.0.3 to 3.2(2c). Everything went smoothly except 1 module in 1 9509 had to reboot causing those 16 ports to drop. The install stopped at that point leaving us at a mixed release on that 9509. We couldn't find any documenation on that error and when we continued we didn't see any issues.

As far as your original question about things to do prior or doing the upgrade. Backup all the running configs is something we always do. If possible try this upgrade process on a test switch if you have one. In another thread the talk about the fact that the 3.2 code will not fit on the Generation 1 bootflash with the current release (so 3.0.2 kickstart.bin and sys.bin + 3.2 kickstart.bin and sys.bin). We had this problem also and left only the old kickstart in place during the upgrade. This should only be an issues with the Generation (SUP1) 1 supervisors, not enough space in the bootflash.

Hello,

I don't have all the details on exactly what issue you ran into on the upgrade but it could possibly be the following.

ETA emc185406: MDS-9000 Switches: Under certain circumstances, Generation 1 modules may reload during a SAN-OS upgrade, due to Compact Flash unable to partition for new code.

These and other issues are documented in both Cisco's and EMC release notes and in our knowledgebases. Did you open a case with EMC on the issues you encountered?

Thank you.

The switch in question is supported by another vendor and we are just starting to move this along to them. Thanks for the issue log. We couldn't find anything the day of.