[c-nsp] 1Gig->10Gig port-channel migration

[Alex Pressé]

It can be sub-second if you configure uplinkfast and backbonefast.

When the link migrates to the new channel, it will flood the CAM table to
the neighbor. This will happen in both directions. This can actually congest
the link momentarily. There is an option to set the packets/second update
rate to uplinkfast to mitigate this.

I would manually set the weights of the current channel interfaces to be
better than the 10-gig (or, set the 10G weights in the other direction).
Then I'd make the physical connections so that the channel comes up but goes
into blocking state after it has learned the addresses.

Once that is done and you've confirmed the interfaces are all up, I'd switch
the weights back to default. 10G will immediately take over.

If you've got a pair of other switches I'd definitely lab it out on some
100/1G channels.


On Thu, Oct 13, 2011 at 10:41, Phil Mayers <p.mayers at imperial.ac.uk> wrote:

> On 13/10/11 16:51, Tim Durack wrote:
>
>> On Thu, Oct 13, 2011 at 11:24 AM, Phil Mayers<p.mayers at imperial.ac.uk**>
>>  wrote:
>>
>>> On 13/10/11 16:03, Tim Durack wrote:
>>>
>>>
>>>> I had hoped I could add the 10Gig links to the existing port-channel,
>>>> and remove the 1Gig links.
>>>>
>>>
>>> No can do, as you've doubtless discovered.
>>>
>>
>> Yes, thanks for confirming ;-(
>>
>>  Next best idea is to migrate the 1Gig port-channel to Po2, and create
>>>> a new Po1 with 10Gig links (I would like to maintain Po1 as the pop
>>>> backbone.)
>>>>
>>>
>>> I would just do this, then let STP choose the higher bandwidth link.
>>>
>>>
>> Any idea what the STP impact is likely to be? Sub-second? 30s? Wild
>> STP oscillations that eventually recover 3 days later after disabling
>> all redundant links? (Our STP is stable, but I still have a healthy
>> fear of "waking the sleeping giant.")
>>
>
> No idea I'm afraid; I haven't timed STP in those circumstances.
>
> In principle it should be fast; although I'm not sure how quickly the FDB
> will re-learn once the 10gig channel takes over.
> ______________________________**_________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/**mailman/listinfo/cisco-nsp<https://puck.nether.net/mailman/listinfo/cisco-nsp>
> archive at http://puck.nether.net/**pipermail/cisco-nsp/<http://puck.nether.net/pipermail/cisco-nsp/>
>



-- 
Alex Presse
"How much net work could a network work if a network could net work?"