07-24-2019 01:57 PM
Hello,
I'm going to be upgrading some Cisco N9K's to current gold-star code. They're currently setup in VPC configuration. I've upgraded Nexus switches before, but i'm just wondering if the switches being in VPC configuration would have any impact. Can I do one at a time and have them running on incompatible software for 20-30 minutes without ill effects? Everything is dual-homed. Also, should I shut down interfaces to force traffic one way while the upgrade of one device is occurring?
Thanks,
Shameer
Solved! Go to Solution.
07-24-2019 02:19 PM
Bobby,
Yes, so for example; on a duly connected server, shut the link to the chassis you are about to reload. Even if you don't shut the link before reloading the chassis, the fail-over from one link to another is pretty fast (a few seconds). So, either way, you should be fine as long as everything is duly connected.
Good Luck!
07-24-2019 02:04 PM
Hi,
Yes, with vPC, you can have each switch running a different version for a short time (few hours) until you upgrade both chassis. And yes, it is good idea to shut the interface to the switch you are rebooting to eliminate any longer downtime. If everything connected to both switches, you should not see much of down time (maybe a few seconds). I recommend doing the upgrade in a maintenance window in case something goes wrong.
HTH
07-24-2019 02:10 PM
Thanks Reza. So with shutting the interfaces down, I was thinking to shut down the VPC links (dual 10Gb). Do you agree?
07-24-2019 02:19 PM
Bobby,
Yes, so for example; on a duly connected server, shut the link to the chassis you are about to reload. Even if you don't shut the link before reloading the chassis, the fail-over from one link to another is pretty fast (a few seconds). So, either way, you should be fine as long as everything is duly connected.
Good Luck!
09-10-2019 02:03 PM
Hey Reza,
I had a quick question regarding this upgrade. As you'll recall, I planned to upgrade the 4 nexus 9k's, one at a time. Everything is dual-homed (servers, etc.). I didn't shut down any interfaces prior to upgrading since it was a maintenance window and a few extra seconds of downtime was not a big deal.
Anyhow, the upgrades went fine. But the customer mentioned that he thought failover would be sub-second for hosts connected downstream. It turned out to be about 6 seconds. I've looked for documentation on what the expected behavior would be. Haven't found anything yet. In your experience, is loss of connectivity for about 5-10 seconds the norm? Or should it be something like max 1-2 pings lost?
I appreciate your insight as always.
Thanks,
Bobby
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide