How does stacking provide high availablity for edge devices ? are ports being wasted ?

[SJ K]

Hi all,

Recently, my senior has hand me a set of cisco 2960 which can be stacked.
Beside the ease of management (single management ip, single configuration required) etc, it is said that stacking provide high availability.

But I cannot really understand how does it achieve high availability since each ports in the stacked switches are still individual ports, they do not have the same configurations after all and neither does the interface took over another interface of another stack member (when that particular stack member is down)
 

For the example below

In order to achieve HA, does that means that I have to utilize 2 ports on the server and also have 2 connections to the router for the same network ? Isn't this a waste of ports ?

Also, it means that I have to assign x2 ips as well.

I am sure i am missing something.

Can gurus here please shed some light on me.

Thank you.

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

You get redundancy, on the switch stack, if both links are in the same L2 domain(s).  STP will block one, if the unblocked link fails, the other unblocks. (NB: with pvst or MST, you can also use both links.)  Or, you often place the links into Etherchannel.  This provides both L2 or L3 link redundancy, and more aggregate bandwidth too.

On your router and/or server, redundancy depends on those devices too.  Although a single router or server without similar hardware redundancy, still may obtain some redundancy by having another link, already connected.  Again, in a Etherchannel kind of configuration, there's also the advantage of additional aggregate bandwidth, while avoid some single points of failure.

[edit]

You normally would only need to assign extra IPs if each link is a routed link, often not the case with switches and/or using Etherchannel.

[SJ K]

Hi Joseph,|
 

Thanks for your reply and feedback.

You get redundancy, on the switch stack, if both links are in the same L2 domain(s).

q1) Can you elaborate further ? Are you referring to the switch stack itself or the 2 links from the server to the switch stack ?
 

q2) If we do not use etherchannel -> can we say that it will be an active-passive setup (since 1 link will be block by STP), and 1 port will be wasted on the stackswitch ?

q3) speaking of spanning tree, will the 2 connections to the router be affected by spanning tree ? (since router breakup broadcast domain)  ?

q4) wouldn't it be wierd that a router has 2 interfaces connected to the same subnet ? how does the router know which interface to send out the traffic to ? I don 't think we are able to set 2 interfaces on a router that belong to the same subnet though as well ...

Regards,
Noob

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

q1 Most of the redundancy is for the switch stack, as it not longer has a single point of failure.  Either stack member can run without the other, and as there are connections from each stack member, there's not single point of failure there either.  (Often the last single point of failure you need to consider is power to the switch stack.)

The server too, might be obtaining additional redundancy, e.g. perhaps two NICs.

q2 Yes, if Etherchannel isn't used, generally one link is just on stand by.  As to whether this is wasting a link, it is, and it isn't.  What's the impact to the business if the single link fails?  I.e. how long to server restoration.  So a standby link is "waste" like paying for insurance.

q3 Yes, if the links are L2, and the router is a pure router, by default, such a setup wouldn't work.  However, if the router supports bridging, it will, or if the "router" is a L3 switch, it will too.

q4 Also yes, if we're speaking of routed ports on the router while the connections are just L2.  Again, you would either Etherchannel the ports (allowing L2 or L3), run both ports as routed ports (not possible with a L2 only switch but later 2960 support some basic L3), or bridge the ports on the router (if the router supports that).

[SJ K]

Hi Joseph,

Thanks for your wonderful reply.

q1) Can you elaborate what kind of redundancy does it provide to the switch stack, i understand that they stack together and the failure of 1 switch doesn't bring down the whole stack. But each of the stack's member's interface or configuration is not a replica of the other. So if there is a failure of 1 member, all the connection to that particular member will fail.   Isn't it so ?

q3) i have thought that spanning tree is to prevent L2 broadcast, which i thought router will not forward broadcast and thus no loop will be created.  can you elaborate further on how router is affected by STP ?

q4) do you mean that, the only way is to either

i) etherchannel the 2 ports on the router or

ii) bridge them up (if the router support)

right ?

Regards,

Noob

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

q1: Yes, correct, anything connected to a switch member that fails loses that connection.  But that's were a 2nd (or nth) connection comes in.  When a switch member fails, connectivity won't be lost, although depending on the additional link(s) configuration, there might be a delay for it to be used.

A switch stack, in principle, behave much as a chassis switch, with each switch member being like a chassis line card.

BTW. with multiple switch members, you can "set aside" a switch member (or that number of) ports as "backups" to all the other switch members.  If a switch member fails, you only need to repatch from the failed switch member to the "backup" ports.  Now this would, I'm sure, even seem more of a case of "wasting" ports, but the consideration really is, how long will hosts be down until the failed switch member is replaced, and what's the impact of the business of that.  The advantage of this approach, as the number of switch members increase, there's less "wastage".  (The same can be done with chassis line cards.)

Another variation of the last, when you have multiple stacks, using the same model, you can have a "spare" switch member on hand.  Again decrease the outage time.

All the forgoing vary in outage time.  Active multiple ports might resume in milliseconds to seconds, extra stack member might take minutes to a partial hour, spare on-the-shelf switch might take an hour or more, a replacment service contract might take multiple hours to a day or so.

q3: No, STP isn't just for broadcast, it logically block usage of a link(s) that creates a L2 loop.  Routes aren't part of L2 loops unless they are configured to bridge (when they operate like a switch, remember a switch works as a multi-port bridge).

q4: It depends on how the router is configured.  If the ports are routed, then as you noted, you multiple ports connected to a switch usually isn't useful, for redundancy, (as typically a router will not allow two routed ports on the same network).  The way to make them useful is to Etherchannel them or bridge them (in the router).

[SJ K]

Hi Joseph,

I am sorry for the late reply.  Been tied up with work . Thank you so much for your reply.

So in summary, stacking provides 1) easy management, 2) ports expansion, 3) redundancy (provided if you want have the extra ports on the alternate switch member to spare) - am i right ?

And the common solution is to have an etherchannel setup between

a) the device/server  to across the switch member in the stack &

b) etherchannel trunk across the switch member to the router.

In this case, ports on each member are utilized  to enable higher overall bandwidth.

Right ?

Regards,
Noob

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

"So in summary, stacking provides 1) easy management, 2) ports expansion, 3) redundancy (provided if you want have the extra ports on the alternate switch member to spare) - am i right ?"

#1 yes, #2 it can, but doesn't have to.  For example, if you need 40 ports, you could use a single 48 port switch, or two 24 port switches, stacked.  Same number of ports in the latter, but you would probably do the latter for redundancy.  #3 Yes, if you provide extra ports, but even if not, in my example of having two stack 24 ports switches, if one stack member fails, you can chose which 24 hosts can be kept operational.

"In this case, ports on each member are utilized  to enable higher overall bandwidth."

Yes, that's one of the advantages of Etherchannel, but remember only aggregate bandwidth is increased.  Single flows are kept on one link, so they will not see a bandwidth improvement.

Other advantages, besides redundancy for Etherchannel, usually faster switchover, link failure "invisible" to STP or L3.

BTW, other advantages of stacking switches, they can offer more bandwidth between switch members than provided by ordinary Ethernet ports, you get "sup" redundancy without the need to buy a redundant sup (often a big ticket item on a chassis).

[SJ K]

Hi Joseph,

Thanks for the wonderful reply.

Just a few last query,

q1) I understand during configuration of the stacked switch, only 1 switch configuration need to be done. However, i believe the configuration is actually replicated to the different member so if the master is down, the next master will run the config stored within it ? , am i able to see the indivdual config of the different switch member ? (just to see changes being replicated to different member)

q2) when you mentioned that stack provide more bandwidth between switches, what do you mean ?  Do you mean that the stacking cable provide more bandwidth then daisy chaining switches with Ethernet cable ?

Regards,

Noob

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

q1: Yes, there's only one configuration for the switch stack, when you save it, it's saved on all switch members. If another switch takes over, it too has a copy.

I'm not 100% sure whether you can "see" the save configuration on different switch members, but I think you can.  By default, things like show flash (or directory) show the current master's switch's contents, but you can preface flash with a switch member.  There's also a way, I recall, to logically logon to a selected switch member.

q2: Yes, often the stack cables provide more bandwidth.  For example, in the original 3750 series, each stack port provided, I believe, 8 Gbps duplex.  The later E and -X series doubled that.  Also both stack cables are used when connected.

The newer 3650 and 3850 provided even more bandwidth.  I think the 3850 might be 120 Gbps, duplex, per stack port.

However, I believe the 2960-S series stack ports are "only" 10 Gbps.

[SJ K]

thank you joseph!