C9200 mixed stack

[mikhailov.ivan]

Hello colleagues! I am on design stage of a project where a customer needs many endpoint ports and one of the solutions is a stack of c9200-48T (6 items) and I thought that I could add couple of c9200-48PB into the stack. But I am not sure that it will work despite that it's the one series of c9200 switches and it doesn't contrary of the data sheet https://www.cisco.com/c/en/us/products/collateral/switches/catalyst-9200-series-switches/nb-06-cat9200-ser-data-sheet-cte-en.html  I worry about different numbers of supported VNs (vrfs). Even if I assemble the stack and it starts working by some magic,  what if the control plane node (9200PB) gets down and another c9200(not pb) will get the role ? From my logic it shouldn't work, but I didn't find any proofs. Could anyone help please?  Thanks 

[M02@rt37]

Hello @mikhailov.ivan 

From my point of view, stacking different models within the Cisco Catalyst 9200 series, such as mixing the 9200-48T (non-PoE) and 9200-48PB (PoE+) switches, is technically possible and supported by Cisco, as they belong to the same series.

However, there are critical considerations that you need to address before proceeding. One of the primary concerns is the potential disparity in features between these models, particularly the number of supported virtual networks (VRFs). When switches with different capabilities are stacked together, the entire stack generally adopts the characteristics of the least capable switch. This means that even if the PoE models in the stack support a higher number of VRFs, the stack might be limited to the number supported by the non-PoE model. This could limit the functionality and performance of your network, especially in environments where higher VRF counts are necessary. Moreover, if the control plane role shifts to a switch with fewer features during a failover, this could introduce operational disruptions or reduce the overall capabilities of the stack.

In light of these potential issues, best practice suggests using uniform switch models within a stack to ensure consistent feature support and simplify management... 

[mikhailov.ivan]

Thanks for the quick response mate, appreciate it!  Yes from straight logic I understand it too that we shouldn't mix different models , but according to the data sheet we can and the data sheet it's the official document which has a legal power that can be used in a court. I would like to see some official comment from the vendor if it's possible. Thanks! 

[Flavio Miranda]

@mikhailov.ivan 

As per Cisco docs related to restrition, the models you mentioned will work. Now, this part is not clear to me

"Even if I assemble the stack and it starts working by some magic, what if the control plane node (9200PB) gets down and another c9200(not pb) will get the role ?!"

A stack of switch is a group of switch working in one logic entity as it were one switch.

What you need to worry about is related to capacity, for example: " Stacking Bandwidth Support" And from my experience, I would not  recommend use mode than 4 switches in stack.  Put 6 or mode can create a  massige point of failure on the network in case the stack goes down for some reason.

Restrictions for Switch Stacks

The following are the restrictions for switch stack configuration:

• A switch stack can have up to eight stacking-capable switches connected through their StackWise ports.

• Only homogenous stacking is supported, that is, a stack of Cisco Catalyst 9200 Series Switches with only Cisco Catalyst 9200 Series Switches as stack members.

• You cannot have a switch stack containing a mix of different license levels.

• Do not stack Cisco Catalyst 9200L Series Switches with Cisco Catalyst 9200 Series Switches.

• C9200-24PB and C9200-48PB switch models can only be stacked with each other and not with other models of the Cisco Catalyst 9200 Series Switches.

• During a switchover, when the standby device syncs with the active device, the following log message is displayed on the console:

  %SM-4-BADEVENT: Event 'standby_phy_link_up' is invalid for 
  the current state 'NO_NEIGHBOR': rep_lsl_rx Gix/x/x -Traceback=

  Ignore this message. It does not have any functional or operational impact.

[Joseph W. Doherty]

"C9200-24PB-A, C9200-48PB-A SKUs supports 32 Virtual Networks. These SKUs cannot be stacked with C9200 SKUs with 4VNs."

Above, a footnote from your datasheet reference.  Seems to answer your question.

[mikhailov.ivan]

Let's consider exact example:  we have 3 switches,  2x c9200-48T (4 vrfs , w/o poe) and one c9200-48pb(32 vrfs plus poe) , we assemble the stack , configure priority of the c9200-pb to became a master , configure more than 4 vrfs , everything is fine and for some reason the master's c9200-pb gets down (for instance because we didn't pay for an extra subscription of each ports bandwidth and forgot an extra license for the installation kit) what will happen next ? The stack will get ruined completely?  Or will only 4 random vrfs still alive? 

[waqas-khan667898 · Accepted solution]

While it's true that all switches in the Catalyst 9200 series should be able to form a stack, there are potential issues with mixing models that support different features, such as varying numbers of VNs (VRFs).

If a c9200PB is acting as the control plane node and it fails, a non-PB c9200 taking over might indeed cause problems, particularly with feature support. The stack should technically still function, but you might run into limitations or inconsistencies with certain features.

To avoid these risks, it's generally recommended to use switches with consistent capabilities in a stack. However, if you're set on this configuration, testing it thoroughly in a lab environment before deployment could help identify any potential issues.

[mikhailov.ivan]

Thanks,  as I thought. So now it would be great if cisco mentions it in a doc for avoiding sole legal problems.

[mikhailov.ivan · Accepted solution]

I am completely blind 

Sorry guys,  the answer was found 

2 C9200-24PB-A, C9200-48PB-A SKUs supports 32 Virtual Networks. These SKUs cannot be stacked with C9200 SKUs with 4VNs.

Thank everyone for your help! 

[Joseph W. Doherty]

---

@mikhailov.ivan wrote:

I am completely blind 

Sorry guys,  the answer was found 

2 C9200-24PB-A, C9200-48PB-A SKUs supports 32 Virtual Networks. These SKUs cannot be stacked with C9200 SKUs with 4VNs.

Thank everyone for your help! 

---

Just curious, how found?

My earlier reply, an hour before your reply, provides exactly the same information, word-for-word.  (Well, except I didn't include the footnote's number.)

Laugh - but you did also note "I am completely blind".

I think (?) you might be able to mark you own reply as "solution".  If so, you might do that so that anyone else with the same question can easily find an actual "solution" to your OP.  Personally, I don't consider @waqas-khan667898's reply a "solution" to your OP, although certainly it could be considered helpful.  However, "solutions" are in the eye of the OP.

Oh, if you do mark your reply as a "solution", suggest you might reference where that information was found although it's implicit by your "blind" remark.

Also, as to "solutions", @Flavio Miranda reply has embedded within it:

• C9200-24PB and C9200-48PB switch models can only be stacked with each other and not with other models of the Cisco Catalyst 9200 Series Switches.

Which is also a correct answer, i.e. you can not mixes those OP models, although he didn't provide a reference to Cisco literature.

[mikhailov.ivan]

I started reading the data sheet from the begging till the end and just noticed this line somewhere in the begging 

"C9200-24PB-A, C9200-48PB-A SKUs supports 32 Virtual Networks. These SKUs cannot be stacked with C9200 SKUs with 4VNs."  But I didn't see it before, so either I am blind or they could fix the doc when saw the post.

[Joseph W. Doherty]

---

@mikhailov.ivan wrote:

But I didn't see it before, so either I am blind or they could fix the doc when saw the post.

---

Currently at top of document:

"Updated:July 31, 2024"

So, laugh, you missed it an earlier read.  Easiest enough to do.  No fault in that.