Cisco developed the StackWise interconnect feature for the 3750 model family to overcome several limitations of traditional switching designs. StackWise improves switch management, configuration, and capacity restrictions over independently managed and connected approaches. Stackwise does so by combining 2-9 switches into a logical “stack” that shares a single control plane. This unified switching stack shares configuration and routing information across all switch members and allows for individual switches to be added or removed according to port capacity needs.
Cisco has since expanded StackWise capabilities to the 2960-X, 3750-X, and 3850-X switching platforms that are primarily intended for access or aggregation use.
Benefits of Stacked Switching
When compared to the use of stand-alone switches, switch stacks present a number of advantages. The first is the use of a virtualized control plane across the switches. Stand-alone switches must be individually managed, monitored, and configured. Stacks unify all of those functions into a single touch point resulting in streamlined configuration and operation management.
Cisco StackWise also enables the use of cross-stack EtherChannels across member switches. This redundancy feature allows other network devices like servers or other switches to be connected to multiple member switches using a single logical EtherChannel link. If a single switch fails, hosts connected via EtherChannels will remain connected.
Scalability is another benefit of stacked designs. To increase port capacity in fixed-switch environments, additional switches must be collocated with existing switches – each given limited inter-switch bandwidth and each requiring independent configuration and operational management. If port capacity needs change in a stackable deployment, member switches can be added or removed as needed without adding additional configuration or operational maintenance overhead.
How StackWise Works
The individual switches in a stack are combined using StackWise cables, two per switch. The stack cables are connected in a way that creates a bidirectional closed-loop path among all member switches. This loop of stack cables acts as a switch backplane for the whole unit, similar to the backplane used in modular chassis switches. In 3750s, this backplane supports 32-Gbps. The X series models (3750-X and 3850-X) double that with backplane support for up to 64-Gbps.
In each stack, a single member switch is elected as the master switch. The individual master switch then acts as the control unit for the stack – processing all control plane traffic and using the remaining member to forward traffic as needed. The stack master switch operates similarly to how a supervisor card acts in a chassis switch – using the remaining switches like remote line cards.
The stack configuration is shared among all member switches in the stack, but controlled by the master switch. If for any reason the master switch fails or is removed, a new master switch is promoted and the remaining stack will continue to forward without interruption. This distributed configuration design also allows new switches to be added as stack members in live environments without service disruptions.
Cabling Cisco StackWise Switches
Stacked switches are cabled sequentially, creating a continuous loop as shown below. If any of the cables fails or becomes disconnected, the backplane bandwidth is reduced by half but connectivity is maintained. When the stack detects the connection has been restored, it will fail back to full bandwidth capacity using both paths around the loop.
Master Switch Election
Each stack is controlled by a single member switch known as the master. Any member switch is eligible to become the master. The stack election process occurs after the whole stack is powered up or after a whole-stack reboot. The selection of the master switch is based on the following criteria (in order):
- Priority – Switch priority (1-15) can be manually configured by an administrator on each member switch. Higher priority is preferred. Cisco recommends preselecting the switch you would prefer to be the master and configuring priority values accordingly.
- Non-Default Interface Configuration – Switches with existing interface configurations will be elected master over switches without any interface configuration applied. This prevents newly added switches without any configuration from overwriting the configuration on the rest of the stack.
- Hardware and Software Priority – The switch with the most advanced feature set is preferred. For example, a member switch with IP Services will be selected over a switch running IP Base.
- Longest Uptime
- Lowest MAC Address
After the member switches have been cabled together using the StackWise cables, you should use the show switch stack-ports command to verify the stack cabling connections.
If you need to verify which StackWise ports are connected within the stack, issue the show switch neighbors command.
Renumbering Member Switches
It is common practice during the initial deployment to renumber the switches in order from top to bottom. Since the interface configuration starts with the switch ID (ex. interface Gi3/0/23 is interface 23 on the switch with ID 3), it is important to keep the physical and logical switch numbering consistent.
To start, issue the show switch command to display the IDs and roles of each member in the stack.
In the output above, we can see that the master switch has been manually configured with a priority of 15 (highest) and the second switch in the stack is configured with a priority of 10. This will always elect the top switch in the stack as master and the second as the standby if the top switch fails or is removed.
To renumber a member switch use the switch
Once all the member switches have been renumbered appropriately, issue a copy run start to save the configuration to startup configuration. Finally, issue the reload command to reload the entire stack.
Once the stack reload is complete, use the show switch command to verify the changes.
Configuring StackWise Priority Values
Defining the priority for switches is the best way to define which member switches will be elected to master upon stack reload. This is recommended by Cisco to ensure that the same switch is re-elected master in the event of a re-election. Priorities values range from 1 to 15, with higher values preferred. The default is 1.
To change the priority of a member switch, issue the switch
The new priority value will be immediately applied to the stack member, but it will not affect the status of the switch until a stack re-election occurs.