Uninterrupted update of firmware on 9300 stack?

[jmcgrady1]

Say i have a two member C9300 switch stack. Does the latest IOS allow for members to be updated one at a time such that one member is always online and "up"? In the past, both members would be updated and non-operational in sequence such that the whole stack is down.

[Leo Laohoo]

Catalyst 9k do not family no longer support Rolling Stock Upgrade.

Alternatively, what you can do is unpack the firmware and then schedule a reload when no one is at the site like, say, 04:00 in the morning.

[balaji.bandi]

i do not believe that is possible as of now, whole stack reboot is good approach still.

[jmcgrady1]

Do the Catalyst 9300 range support virtual stacking (VSS)? If not, i'll have to use two unstacked switches as any outage is unacceptable in this use case.

[Leo Laohoo]

9300 do not support VSS.  

[balaji.bandi]

Cat 9300 is not designed to work for VSS, you need to upgrade whole stack and reboot in the maintenance window,

[Florin Barhala]

Hey guys,

I read on the 9300 DataSheet on the licensing tab about NSF.

I am puzzled why not software update for 2 members stack would not work without a downtime?

High availability and resiliency Nonstop Forwarding (NSF), Graceful Insertion and Removal (GIR), Extended Fast Software Upgrade (xFSU), Software Patching (CLI Based)

✗

✓

[balaji.bandi]

Not sure what context here, nonstop forwarding as i remember works for any stacking device. upgrade required reload stack.

[Leo Laohoo]

NSF only happens in the event of switch member failures or crashing. 

NSF does not come into effect during firmware upgrade.

[Florin Barhala]

Thank you Leo !

For anyone hitting this thread, here's a link about xFSU feature: https://www.cisco.com/c/en/us/support/docs/switches/catalyst-9300-series-switches/216837-extended-fast-software-upgrade-on-cataly.html#anc3

For what I read so far, it significantly reduces the downtime during a software update.

If anyone has already used it in a stack with Network Advantage license (mandatory) please share the total amount of topology downtime.

[Leo Laohoo]

@Florin Barhala, I have no idea what these questions are for.  It is better if someone "comes clean" and tell us what is going on or what are they trying to achieve. 

Let me tell a small story that happened back in November 2021. 

We had to do a firmware upgrade at a site.  This involved around 12 stacks of 9300 switches and two 9500 in a VSS pair.  

The 9300 upgrade went without a hitch.  Seeing we had so much spare time, we decided to try out Cisco's much-hyped Fast Reload option on the 9500 VSS.

Well guess what, it did not work.  We we triggered the fast reload, the VSS pair CRASHED.  Instead of "saving" 5 minutes of downtime, we waited for the VSS to crash and reboot normally, about 8 minutes.  

ISSU, FSU/eFSU are all hype.  Do not get fooled by it.  These only work in very "corner cases" such as in a lab environment.  If-and-when ISSU or FSU/eFSU work, they work.  But when they fail, they fail spectacularly.  Get ready to drive out to the site because it will be a very long, lonely, quiet and humiliating drive -- At least no one can hear you scream.