Cisco 9200 Stacking failure behaviour

[Royls]

Hi Team.

This is a first for me, it's in order to get an understand of the function of stacking when a singular stacked, stack bares victim to a failure to the stacking module or cable, in either switch. What would the behaviour be if you had both switches of the stack in a port-channel, and the cable or module failed, which would also share the same hostname and address?

Would they continue to work? Allowing little time to replace, as once you form a stack the interface range increment is determined T1/1 T2/1 etc.

How would the port-channel respond? 

Addition: And what if they connected to another core stack in different locations!

[Reza Sharifi]

Hi,

There is no Portchannel involved in stacking. When you stack a couple of switches together, you just use the stacking cables to connect them. When you stack multiple switches together, they logically become one switch, meaning you only manage one switch with one IP and one host name. If you have multiple switches in close proximity, stacking is a good solution.

HTH

[Royls]

This has nothing to do with stacking with the use of a port-channel, its the behaviour of the stack with a failure to the stacking cable or module, which connects to another stack in a port-channel, with the 2nd stack in two different locations. 

I'm interested in the response of the first stack.

[Reza Sharifi]

If the 2 stacks are connected together with a Portchannel using 2 different ports on 2 different switches and if stacking cable fails on one of the stacks, it will effect the local switch and the local users on that site but the Portchannel should continue to work. In a case of module failure, as long as you are using one connection from each switch module, the Portchannel should continue to work.

HTH 

[STD_NetWorld]

@Royls : In Case for Physical Stack (Traditional, where we connect Stack cable in cross SW1_Stackport1---to-- SW2_Stackport2 &SW1_Stackport2---to-- SW2_Stackport1). we have management and data plan will be in SNYC and incase of any 1 cable fail still work as it is. if both cable fail then higher priority will be active and second switch will be behave as individual.

for Virtual Stack: their are different failover scenarios, please follow this link for the same.

https://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst9500/software/release/16-11/configuration_guide/ha/b_1611_ha_9500_cg/configuring_cisco_stackwise_virtual.html

[Leo Laohoo]

---

@Royls wrote:
What would the behaviour be if you had both switches of the stack in a port-channel, and the cable or module failed, which would also share the same hostname and address?

---

What cable or module?  

9200 has an optional stacking module.  Some people do not know about this and do not purchase it.  However, some people, after realizing their mistakes, would purchase only one (instead of two).  So they would implement a stack with a single stacking module and make half a ring.  

If this sole stacking module fails, then there is a split brain scenario because the entire stack shares the same config.  

If there were two stacking module and one of them fails, then nothing happens because there is redundancy involved.

[Royls]

Yes, yes, and yes. It's this behaviour I'm trying to understand, as half a ring with stacking failure.

"If this sole stacking module fails, then there is a split brain scenario because the entire stack shares the same config" - This is the behaviour I'm talking about. Because both devices are connected directly to different devices, which are stacked. Would both switches in the broken stack continue to work? Create a loop? What would be their behaviour, if you know?

[Joseph W. Doherty]

Yes, a partitioned stack will continue to try to operate as two of the same logical device, which generally is a bad thing!

[Royls]

Thanks Joseph. It confirms my worries.

[Joseph W. Doherty]

That's one reason having a dual ring is important; precludes one single point of failure causing that.  Also, provides 2x aggregate ring bandwidth.

[Royls]

There are huge delays still on the 9200 / 9300 stacking kits delivery. This was for understanding worst case.

[Joseph W. Doherty]

Ouch!

In that case you might consider whether stack is a "must" requirement, especially with stack of more than two.  (What's special about two?  You can hope one of the stack members completely fails.)

[Royls]

Thanks for all the input guys, enjoy the remainder of your week. R