cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
17017
Views
25
Helpful
12
Replies

High availability with switch stack

DkNguyen1490
Level 1
Level 1

Hi everyone

I am reading about switch stack technology and i have some questions about high availability feature

I dont really understand how stack technology can provide this feature (i mean HA for data traffic, not only for device management)?

Can somebody help me make clear about this?

 

Thank you very much!

 

TrungNQ

1 Accepted Solution

Accepted Solutions

That is right. If you cross connect your access switches to the core stack using portchannels,, than you don't need to worry about STP, HSRP, VRRP, etc..

The down time is very minimum in case there is a link failure..

HTH

View solution in original post

12 Replies 12

jj27
Spotlight
Spotlight

HA is available for management traffic and any routing configuration as each switch in the stack is capable of being the master should the current active master fail. The next elected master would take over any routing or layer 3 interfaces and continue to function.  Any devices that are connected to a failed switch would lose their connectivity until physically moved to another switch in the stack.  If this is a concern, you would normally dual-home devices connected to the stack, if possible.

Some data high availability is achieved through the StackWise cables on the back of the switch.  All data is efficiently load balanced across the StackWise uplinks for data that needs to traverse through an uplink to somewhere else on the network.  If one StackWise port fails, the traffic would automatically be redirected across the other StackWise port.

Hope this helps.

Thank you very much for all of your answers

 

As I understand, stack will:

- Make several switch act as 1 big switch ( if i have 2 switch 24 port, when they are stacked, i will have a big switch with 48 port)

- Single point of management( all configuation just do one time, upgrade also)

- Stack bus (Cable plug in stack module to connect 2 switch), it just a very high speed link, high bandwidth; it better than us etherchannel to connect 2 standalone switch( maximum 8 link). ???. It also use to "feel" to determine when 1 switch is down and remove it out of stack, right?

 

I have a topology: 

- My server is configed NIC teaming (mode active-standby) with 1 ip address. 2 NIC is linked to 2 difference standalone switch. Each switch have a way out. If i use switch stack, in case of 1 uplink down, switch will feel it and automatic use the uplink in other switch to traffer data from my server in the way out?

 

Thank you for your support!!!

Hello Nguyen,

As per your topology , if we replace stack  switch in place of two different standalone switches. If all the ports in the uplink and the stack switch are in STP the switch will detect  and will change the state of the blocking port to forwarding. 

However i will suggest to use port channel if the upstream topology supports to take the advantage of load balancing of both the links.

 

regards

Parhta 

So the benefit of stack about HA feature is about cross-etherchannel, avoid stp to make the down time smalless, right ?

That is right. If you cross connect your access switches to the core stack using portchannels,, than you don't need to worry about STP, HSRP, VRRP, etc..

The down time is very minimum in case there is a link failure..

HTH

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

Yes, stack cables generally provide more bandwidth.  In original 3750 series, each stack port is (I believe) 8 Gbps, duplex.  In 3750 -E/-X series, each stack port is (I believe) 16 Gbps, duplex.  When both cables are connected, both are used.

 

Unsure about your "feel" or determining when a stack member fails question.

 

For your current server topology, you might find you can use both server ports, actively, if connecting to a stack.  Otherwise it should mimic your two switch topology.

Hi Joseph

My question is: how master know when a stack member down and remove it away from the stack. As i understand, master use the path thought the stack cable to healthcheck other member in stack. Is it corect?

 

Thank you!

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

Yes, stack master will "know" when a provisioned switch member or members is not part of stack.  When both stack ports are connected and only one stack member fails (or stack link), stack will function w/o failed stack member.  If only one stack port is connected and stack member fails (or stack link), stack may partition and run as two stacks (believing themselves to be the "original" stack - you don't want that!).

Reza Sharifi
Hall of Fame
Hall of Fame

Hi,

when you stack multiple switches, they logically become one switch. So you are managing only one switch. As far as the data traffic goes, you need to connected your devices to multiple devices in the stack to take advantage of it. Say you have 2 switches stacked in your core, and 5 switches (not stacked) at the access later, you would need to connect your uplinks from each access switch to both core switches to take advantage of high availability.

HTH

Hello,

 

I'm just taking this information a bit deep. Setup remains the same i.e. two core switches are stacked and 5 switches which are access switches and not stacked. These 5 access switches have one uplink to switch-1 in core stack and second uplink to switch-2 in core stack. Now, if I configured SVI in switch-1 (which is assumed to be master in stack) and the IP address of SVI is used as a default gateway in LAN devices then if switch-1 in stack fails will LAN devices send traffic to switch-2 in core stack?

Yes, although your devices may have already been sending data to core stack via SW-2.

There are some "gotchas" when the stack master dies. For instance on the 3750 series it was possible the stack master and standby switches have different IOS feature set versions installed. If so, the feature set might change with a change of master.

I recall L2 forwarding is done independent of stack master, but L3 functions might be impacted until a standby switch takes over the role of master. I recall (?) there's optional L3 NSF commands to mitigate the loss and impact of the stack master.

At least on the 3750 series, the stack master, bt default, provided the MAC for SVI IPs and when the stack master changed so did the SVI's MAC (again by default). There was/is a command to continue to use the same MAC for the SVI after the election of a new master.

Joseph W. Doherty
Hall of Fame
Hall of Fame

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

When an individual stack members fails, it fails, sort of, like a failed line card on a chassis.  The other stack members continue to function.

 

One of stack members acts as the "brains" for the whole stack.  As noted by the other posters, if this stack member fails, another stack member takes over, much like a second supervisor in a chassis.

 

Lastly, the 3750X series can also share power within their stack.  So, even if a power supply fails in a 3750X, that unit might continue to function normally, even while supporting PoE, much like a chassis with redundant power supplies.

 

As mentioned by the other posters, to fully leverage a stack's redundancy for data traffic, devices connecting to the stack would have at least two connections, and all the connections wouldn't be to the same stack member.  (NB: basically simiar considerations if using a chassis with redundancy or a VSS pair.)

 

Of course you can also provide data traffic redunancy with multiple network devices too, but a stack is one device to manage rather than N devices to manage and often provides improved bandwidth between stack members and sometimes also between the stack and other devices (the latter using Etherchannel).

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: