Hello,
I was recently upgrading four Nexus switches and could not find much info about it apart of the Cisco guide, which I can tell you is not really accurate for some cases. Let me share you my experience, maybe it will help someone in the future and help you avoid the near death experience I had during the upgrade.
First of all the topology. All switches are with single sup. Old NX OS does not support ISSU, so I had to do disruptive upgrade.
- 2x N9K-C93128TX with NX OS 6.1(2)I3(2)
- 2x N9K-C9396PX with NX OS 6.1(2)I3(2) and FEX module N2K-C2248TP-E-1GE
From the feature point of view, we did not use many of Nexus feature, mostly VPC.
Both pair were interconnectet with VPC to enable HA.
In the end I had to make it in two maintenance windows, since I have failed to upgrade in the first one. I spent 3 hours on call with TAC support, almost went for RMA, but in the end we have managed to upgrade at least one box on first MW. If you follow Cisco guide, after the upgrade you might end up in Reason: Reset Requested due to Fatal Module Error and the switch will be reloading every 30 minutes.
Now to the upgrade. First few steps are following Cisco guide here.
Firstly read the release notes here to verify your HW and features will be supported after the upgrarde.
1. Check you have enough space in bootflash using #dir boothflash:
2. Copy your NX OS to bootflash using for example SCP
3. Verify MD5 or SHA256 checksum
4. Change boot variables #boot nxos bootflash:/nxos.7.0.3.I7.6.bin
5. Copy your running config to bootflash:
6. Do write erease and reload switch (it will come up as a new box)
7. Now you have to do some initial config like admin accout etc. You can skip most of it.
8. Verify upgrade with #show install all impact nxos bootflash:nxos.7.0.3.I7.6.bin command
9. Upgrade the NX OS #instal all nxos bootflash:nxos.7.0.3.I7.6.bin - switch will reload in the end
10. Wait a couple of minutes (usually 2 or 3) until the HW check is permormed after the box will come up alive, you can check the progress using show module, verify new NX OS is there using show version.
11. If the HW check is alright and new OS is there, just copy you config back from bootflash and you are good.
Now few points.
Even when I have followed this, one of the boxes crashed into constant reload - after the upgrade, HW check failed reporting Fatal Module Error and kept relaoding every three minutes. No possible fix for this (not even rollback to old version) and Cisco recommended RMA. However this can be (as we found out) resolved with cutting the power to the switch. We have disconnected both power cables, waited one minute, plugged it back and after the switch came back online, it came back with new NX OS and HW check was performed successfully.
FEX module upgrade will not upgrade during the reload since we have erased the config and FEX is not recognized in blank switch. Once you copy back the initial config from boothflash the FEX will come online, copy the OS to itself and upgrade itself, it will take about 5 minutes and you can check the status using #show fex detail
After the FEX is back with new code, you can copy the lines from config that was rejected due to the module missing.
Now please understand Nexus and NX OS are complex things and you may encounter many issues and this guide might not work with you. This was my personal experience.
My recommendation is to start with the official Cisco guide for upgrading and if that is not working you can try the one I just desrcibed.
Hopefully this will help someone, good luck!
Tomas
