[c-nsp] Nexus 7k Upgrade Path

[Justin M. Streiner]

On Fri, 23 Feb 2018, Bradley Ordner wrote:

> Thanks, I’ll take this all onboard. The reason for the upgrade is we 
> wanted to start using Loopbacks for OTV and finish a DR design that 
> seems to have not been completed. Sent from my iPhone

ISSU is nice in practice, but it's something that's very difficult for 
vendors to get right - kind of like trying to push firmware updates to the 
avionics systems of a plane that's in flight.

Vendors also sometimes conflate "ISSU" and "hitless", or their 
documentation doesn't always make it clear that an ISSU carries the 
potential of outages.  Sometimes that can also cause people further up in 
an organization to have unrealistic expectations.  If possible, it's 
safer to have a design where operations can survive a network device 
being isolated while upgrades are being performed, rather than having to 
rely on the device operating uninterrupted during those upgrades.

This is not Cisco-specific :)

jms

>
>> On 23 Feb 2018, at 5:58 pm, Pete Templin <petelists at templin.org> wrote:
>>
>> I would even go so far as to:
>>
>> load system/kickstart files
>> isolate the box (shutdown all ports)
>> power-cycle the box, let it boot into the new code
>> perform EPLD updates on all cards
>> run the ISSU command to ensure all of the little microcode thingies (PSUs, fans, etc.) are covered
>> unisolate the box
>>
>>> On 2/22/18 11:44 PM, Pavel Skovajsa wrote:
>>> Definitely not a stupid question. While the double ISSU would work we
>>> generally would not do it for big jumps like that.
>>>
>>> The problem is that the whole procedure tended to be buggy so we are too
>>> afraid. Not speaking about crazy bugs we ran into half year later because
>>> "triggered by previous issu upgrade" and we needed to reload anyway.
>>>
>>> So - our recomendation for jumps like this - load the system a kickstart
>>> files and reload the box? Ideally power cycle....there were fw bugs that
>>> needed hard reboot to fix...
>>>
>>> -pavel
>>>
>>>
>>>
>>> Dňa 23. 2. 2018 7:34 používateľ "Justin M. Streiner" <
>>> streiner at cluebyfour.org> napísal:
>>>
>>> On Fri, 23 Feb 2018, Bradley Ordner wrote:
>>>
>>> We have a Nexus 7K with two SUP2Es. We need to get to software version
>>>> 8.1(2). It says that you can't double hop to a software version without an
>>>> outage. Although I have found the following -
>>>> ISSU from 7.2(0)D1(1) to 7.3(1)D1(1) then to 8.1(2).
>>>> We currently are on 7.2(0)D1(1) according to the doco I should be able to
>>>> upgrade as each version can ISSU to the next?
>>>> Has anyone performed this before?
>>>> I have posted this on Cisco Support Community, with no response so either
>>>> it is a stupid question or no one has done it before.
>>>>
>>> I haven't had to do a double-hop upgrade in a while, but my past experience
>>> with ISSUs on the Nexus 7K has been mixed. Sometimes the 7K ecosystem
>>> benefits from a full reboot.  Also, keep in mind that if any of the EPLDs
>>> on your linecards need to be upgraded, the affected linecards will have to
>>> take some amount of outage.  How much of an impact such an outage would
>>> have depends entirely on your network design.
>>>
>>> jms
>>>
>>> _______________________________________________
>>> cisco-nsp mailing list  cisco-nsp at puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>>> _______________________________________________
>>> cisco-nsp mailing list  cisco-nsp at puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>>
>> _______________________________________________
>> cisco-nsp mailing list  cisco-nsp at puck.nether.net
>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>