03-30-2020 11:40 AM
If I have two Cat6K/Cat9K chassis, each with only one supervisor (dual-sup altogether), and they form a VSS/Stackwise-Virtual, what happens to the linecards when the active Sup fails?
The control plane will failover to the second sup/chassis, but will the dataplane/linecards in the first chassis continue forwarding traffic (non-disruptive)?
My understanding is the Cat6K & 9K are based on centralized forwarding - all traffic has to go through the supervisor engines (except DFC cards).
So when the original active sup in chassis 1 fails, how would the linecards in chassis 1 continue working?
TIA
03-30-2020 12:09 PM
take example : cat6K, each chasis has single supervisor, they are in VSS , chasis 1 supervisor fails - Second chasis sup will take over, there is no data plane traffic expect to fail.(risk is single supervisor running for now until failed one replaced) - depends on criticality of the business choose how many supervisor required.
Cat9500 different, its simple stack wise virtual.
Cat 96XX / Cat 9400 is modular can work same theory.
03-31-2020 10:17 AM
Thank you for the reply, but it doesn't completely answer the question:
If Catalyst switches (Cat4K/6K/9400/9600) are based on centralized forwarding, how can linecards continue working, when the supervisor in the chassis failed?
Follow-up question:
How can you gracefully take the first chassis out of traffic forwarding path, to replace the failed supvervisor, and configure the new one, and then put the chassis back in service? (same question for both VSS & StackwiseVirtual)
There's a GIR (gracefual insertion removal) feature but it looks like that would take the entire stack out of forwarding path, and not just one of them, and it's meant for physical stack models (Cat9200/9300), and not meant for chassis models (9400/9600)?
03-31-2020 10:53 AM
what Data traffic you were refering, hope you have valid design to take over the Data path like, your Distribution or Access switches or dual homed with VSS to sub millisecond failovers.
here is good explanation for you to understand :
https://www.ciscolive.com/c/dam/r/ciscolive/us/docs/2015/pdf/BRKCRS-3035.pdf
04-02-2020 02:34 PM
I'm sorry if I wasn't clear in my original post, but here's what I'm trying to find out:
Two chassis - SW1 and SW2, each w/ single supervisor engine.
Sup in SW1 fails, what happens to the linecards in SW1 (not SW2)?
Do they continue forwarding traffic, or the linecards stop working because of the sup failure?
04-03-2020 07:26 AM
the switch it self isolated with no control plane (or brian in it) - so Switch 2 take over all the connectiones, if your end device / switches connected dual uplink to both Sw1 and Sw2.
for the best design always recomend to have dual sup engines.
04-03-2020 08:39 AM
Thank you for responding.
In case you haven't noticed, part of this is a C9K hardware architecture discussion.
All I was looking for was a straight forward answer, what would happen to the linecards in SW1 when its sup fails.
Your reply "Switch 2 take over all the connectiones" implies SW1 linecards would be in failed state, and wouldn't forward any traffic, so I'll take that as the answer.
You're correct in ideal world, everything would be redundant but that comes at a cost not everyone can afford.
There are also scenarios where dual-home isn't even possible, especially endpoints (phones, laptops, or legacy devices w/o dual NIC capability).
We just need to document the expected behavior when failure occurs.
04-03-2020 08:56 AM
There are also scenarios where dual-home isn't even possible, especially endpoints (phones, laptops, or legacy devices w/o dual NIC capability).
StackWise virtual is not designed to be used in the access layer. StackWise virtual should only be used on the distro or core layer. All access switches connecting to StackWise virtual need to connect to both switches if not there is no benefit in deploying StackWise virtual. In summary, all end devices like phones, laptops, desktops, printers connect to a single access switch using a single cable and the access switch connect to 2 switches configured in StackWise virtual mode.
HTH
04-03-2020 09:33 AM - edited 04-03-2020 09:37 AM
I agree you normally use VSS or StackwiseVirtual as core or distro, but in a collapsed core/distro/access, endpoints do connect to it.
In the Cat6500 VSS days, it was a perfectly valid design - plenty of PoE variant linecards to support that.
There are also cases like server iDRAC/CIMC port, or phone equipment/gateway that only has single port, but you still want them connected to the core.
04-03-2020 09:56 AM
In the Cat6500 VSS days, it was a perfectly valid design - plenty of PoE variant linecards to support that.
I am not sure where you got this info but VSS never supported POE as it was never designed for access layer.
There are also cases like server iDRAC/CIMC port, or phone equipment/gateway that only has single port, but you still want them connected to the core.
you can, as long as you are using VSS or Stackwise Virtual.
You can use VRRP or HSRP instead if you have to connect devices that need to be singly attached and you don't want to connect them to the access layer.
HTH
04-03-2020 10:08 AM
"VSS never supported POE as it was never designed for access layer."
Not true - one of the CiscoLive slides that BB posted above covers single-attached scenarios.
04-03-2020 11:18 AM
No one says, if you single attach a device to one of the switches in VSS or Stackwise virtual switches, it doesn't work. I think the point you are not getting is that it works but that is not best practice.
04-03-2020 10:10 AM
Not sure where you confused about the Design, Most VSS or high availabilyt on your CORE and Distribution,
Access Layer different all together. please refer the presentation to understand, make sure you designed the network as per industry recomendations.
04-03-2020 10:16 AM
Did you actually look at the slides you posted?
See slide 90
04-03-2020 10:35 AM
"See slide 90"
Sill iam not sure what is the point here to understand, point 2 clearly - for various reason. that means there may be exception cases.
if you do not need high availability / device do not support multi-home favility, my view is waste of money to invest on VSS or Stackwise.
May some organisation have different use case like 1 in 50 or 100. May be this your business case may be. so we onl suggest based on our experience and what we have seen on hands on with.
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