Cisco Catalyst 2900 XL Series Switches

Catalyst Switch GigaStack Configuration and Implications

Document ID: 22360

Updated: Sep 01, 2005



The GigaStack Gigabit Interface Converter (GBIC) is a Cisco proprietary GBIC. The GigaStack GBIC provides low cost, high bandwidth connections between supported devices. Gigabit connections can be configured in a point-to-point or cascaded stack configuration, with or without redundancy.

The GigaStack GBIC allows you to stack up to nine switches to form an independent backbone that can be managed with a single IP address. This stack gives the appearance of a single large switch for network management purposes. You can also form a point-to-point link between two switches. The GigaStack GBIC supports one full-duplex link (in a point-to-point link), or up to eight half-duplex links (in a stack configuration) to other Gigabit Ethernet devices.

This document outlines the necessary steps to insure a working configuration and possible issues related to a misconfiguration.



The following table shows the list of Cisco switches that support the GigaStack GBIC, along with the IOS that supports redundant loop configuration. Redundant loop configuration is not supported in any prior release. The GigaStack GBIC is not supported in the Catalyst 4000, Catalyst 5000, and Catalyst 6000 product line.

Model Minimum Release Required to support GigaStack Minimum Release Required for Loop Detection Minimum Recommend Release
Catalyst 2912MF XL 11.2(8)SA6 12.0(5)XU 12.0(5)WC3b
Catalyst 2924M XL 11.2(8)SA6 12.0(5)XU 12.0(5)WC3b
Catalyst 2924M XL DC 12.0(5)XU 12.0(5)XU 12.0(5)WC3b
Catalyst WS-X2931-XL 11.2(8)SA6 12.0(5)XU 12.0(5)WC3b
Catalyst 3508G XL 11.2(8)SA6 12.0(5)XU 12.0(5)WC3b
Catalyst 3512 XL 11.2(8)SA6 12.0(5)XU 12.0(5)WC3b
Catalyst 3524 XL 11.2(8)SA6 12.0(5)XU 12.0(5)WC3b
Catalyst 3524 PWR XL 12.0(5)XU 12.0(5)XU 12.0(5)WC3b
Catalyst 3548 XL 12.0(5)XP 12.0(5)XU 12.0(5)WC3b
Catalyst 2950G-12-EI 12.1(6)EA2 12.1(6)EA2 12.1(6)EA2c
Catalyst 2950G-24-EI 12.1(6)EA2 12.1(6)EA2 12.1(6)EA2c
Catalyst 2950G-24-EI-DC 12.1(6)EA2 12.1(6)EA2 12.1(6)EA2c
Catalyst 2950G-48-EI 12.1(6)EA2 12.1(6)EA2 12.1(6)EA2c
Catalyst 3550-12G 12.1(6)EA1 12.1(6)EA1 12.1(8)EA1c
Catalyst 3550-12T 12.1(4)EA1c 12.1(4)EA1c 12.1(8)EA1c
Catalyst 3550-24-SMI Catalyst 3550-24-EMI 12.1(6)EA1a 12.1(6)EA1a 12.1(8)EA1c
Catalyst 3550-24-SMI-DC 12.1(6)EA1b 12.1(6)EA1b 12.1(8)EA1c
Catalyst 3550-48-SMI Catalyst 3550-48-EMI 12.1(8)EA1b 12.1(8)EA1b 12.1(8)EA1c

Note: Upgrade all the switches in a series to the same software version. Otherwise, the loop-breaking algorithm might never resolve itself. For example, if the stack includes only Catalyst 2900 series XL and 3500 series XL switches, they must run Release 12.0(5)XU or later. If the stack includes a mixture of Catalyst 2900 series XL, 3500 series XL, 2950, and 3550 switches, all the 2900 XL and 3500 XL switches must run Release 12.0(5)XW or later, all 2950 switches must run 12.1(6)EA2 or later, and all the Catalyst 3550 switches must run Release at least 12.1(4)EA1 or later, depending on model.

Components Used

The GigaStack GBIC is an adapter that can only be installed into the GBIC modules slot of any supported switch. The maximum distance for a GBIC-to-GBIC connection is one meter. The GigaStack GBIC requires Cisco proprietary signaling and cabling. Each GigaStack GBIC consists of two ports, port 1 and port 2. One or both ports can be used, depending on cabling configuration.

  • GBIC: WS-X3500-XL (Cisco GigaStack GBIC and 50 centimeter cable for GigaStack GBIC)

  • Cable: CAB-GS-1M (one meter cable/39.3696 inches)

  • Cable: CAB-GS-50CM (50 centimeter cable/19.6850 inches)



caution Caution: Do not use standard IEEE 1394 cables with the GigaStack GBIC. You must use one of the Cisco proprietary cables (CAB-GS-50CM or CAB-GS-1M). If you use any other cable, you will not have connectivity.

caution Caution: Do not use the GigaStack GBIC with standard IEEE 1394 equipment. You might damage the equipment or lose data.


For more information on document conventions, refer to the Cisco Technical Tips Conventions.

Cabling Configuration

It is extremely important to understand how to physically connect the GigaStack GBICs and what mode of operation will occur. Each physical configuration is is cabled somewhat differently and may require the use of more than one GigaStack per switch. Failure to comply may result in a loss of connectivity. The GigaStack GBIC can be cabled in one of the following three ways:

  • Point-to-point

  • Cascaded stack

  • Cascaded stack with a redundant link

The GigaStack GBIC will auto negotiate to either full-duplex or half-duplex. The software determines the mode of operation by the way the cables are connected and the number of switches in the stack.

Note that although duplex can be forced, it is not supported. The only supported option for duplex is to leave it to auto-negotiate.

Point-to-Point Connection

Point-to-point connections consist of one GigaStack GBIC per switch and only one connection exists between these modules. The additional available port on the Gigastack is not used and should not be connected to any other device. In this configuration, the GigaStack interface will automatically negotiate to full-duplex. One switch can have multiple point-to-point connections with neighbors as long as you are only using one per port on separate GigaStacks to neighboring switches.

In point-to-point operations, it does not matter what port is connected to the neighbor's port. For example, GigaStack port 1 on switch A can be connected to GigaStack port 2 on switch B.



Cascaded Stack Connections

Cascaded stack configurations consist of one GigaStack GBIC per switch. Multiple connections exist between these modules. By using one module per switch, you are able to interconnect between three and nine switches using only one GigaStack module per switch. The cable is daisy chained between switches by connecting opposite port numbers between switches. Port 1 to neighboring switch port 2 and so on. There is no physical redundancy.

In this configuration, all GigaStack interfaces will automatically negotiate to half-duplex. All GigaStack ports will act as a repeater and share the one Gbps channel. The shared Gigabit connection between all switches is called a multidrop backbone.


caution Caution: GBICs are sensitive to power spikes. It is advised to regulate the power to the devices to avoid GBIC failures.

Cascaded Stack Connections with a Redundant Link

Cascaded stack configuration with a redundant link consists of one GigaStack GBIC per switch. Multiple connections exist between these modules. By using one module per switch, you are able to interconnect between three and nine switches using only one GigaStack module per switch. The cable is daisy chained between switches by connecting opposite port numbers between switches. Port 1 to neighboring switch port 2 and so on. Physical redundancy occurs when you complete the connection between the first and last switch. Every port in all GigaStack modules will have a connection in use.

In this configuration, all GigaStack interfaces will automatically negotiate to half-duplex. All GigaStack ports will act as a repeater and share the one Gbps channel. A loop configuration will be detected by a loop detection algorithm and one of the links on a GigaStack module will be used as the redundant link. The redundant link in the stack is disabled for data transmission while all other links in the stack are up. When another link in the stack goes down, the redundant link will be enabled for data transmission. No user intervention is required.


Note: This configuration is only supported if all switches in the stack run software that supports the loop detection mechanism.

Using the Cross-Tack Uplink Fast Feature

Cross-stack UplinkFast (CSUF) provides a fast spanning-tree transition (fast convergence in less than one second under normal network conditions) across a stack of switches that use the GigaStack GBICs connected in a shared cascaded configuration (multidrop backbone). During the fast transition, an alternate redundant link on the stack of switches is placed in the forwarding state without causing temporary spanning-tree loops or loss of connectivity to the backbone. With this feature, you can have a redundant and resilient network in some configurations. On the catalyst 2900/3500XL, 2950, and 3550 series switches you enable CSUF by first entering the spanning-tree uplinkfast command in global configuration mode, and then using the spanning-tree stack-port interface configuration command on the Gigastack interface desired. On catalyst switches 3750 CSUF is automatically enabled when you enable the UplinkFast feature by using the spanning-tree uplinkfast global configuration command. The CSUF feature is supported only when the switch is running PVST.

Note: The 2900xl and 3500xl switches require code version 12.0(5)XW or higher to support the Cross-stack uplinkfast feature.

How CSUF Works

CSUF ensures that one link in the stack is elected as the path to the root. As shown in the figure below, Switches A, B, and C are cascaded through the GigaStack GBIC to form a multidrop backbone, which communicates control and data traffic across the switches at the access layer. The switches in the stack use their stack ports to communicate with each other and to connect to the stack backbone; stack ports are always in the spanning-tree forwarding state. The stack-root port on Switch A provides the path to the root of the spanning tree; the alternate stack-root ports on Switches B and C can provide an alternate path to the spanning-tree root if the current stack-root switch fails or if its link to the spanning-tree root fails.

Link A, the root link, is in the spanning-tree forwarding state; Links B and C are alternate redundant links that are in the spanning-tree blocking state. If Link A, Switch A or its stack-root port fails, CSUF selects either the Switch B or Switch C alternate stack-root port and puts it into the forwarding state in less than one second.

Cross-Stack UplinkFast Topology


CSUF uses the Stack Membership Discovery Protocol to build a neighbor list of stack members through the receipt of discovery hello packets. When certain link loss or spanning-tree events occur (described in the Events that Cause Fast Convergence section of this document), the Fast Uplink Transition Protocol uses the neighbor list to send fast-transition requests on the stack port to stack members.

The switch sending the fast-transition request needs to do a fast transition to the forwarding state of a port that it has chosen as the root port, and it must obtain an acknowledgement from each stack switch before performing the fast transition.

Each switch in the stack determines if the sending switch is a better choice than itself to be the stack root of this spanning-tree instance by comparing the root, cost, and bridge ID. If the sending switch is the best choice as the stack root, each switch in the stack returns an acknowledgement; otherwise, it does not respond to the sending switch (drops the packet). The sending switch then has not received acknowledgements from all stack switches.

When acknowledgements are received from all stack switches, the Fast Uplink Transition Protocol on the sending switch immediately transitions its alternate stack-root port to the forwarding state. If acknowledgements from all stack switches are not obtained by the sending switch, the normal spanning-tree transitions (blocking, listening, learning, and forwarding) take place, and the spanning-tree topology converges at its normal rate (2 * forward-delay time + max-age time).

The Fast Uplink Transition Protocol is implemented on a per-VLAN basis and affects only one spanning-tree instance at a time.

Events that Cause Fast Convergence

Depending on the network event or failure, the CSUF fast convergence may or may not occur. Fast convergence (less than one second under normal network conditions) occurs under these circumstances:

  • The stack-root port link fails.

If two switches in the stack have alternate paths to the root, only one of the switches performs the fast transition.

  • The failed link, which connects the stack root to the spanning-tree root, recovers.

  • A network reconfiguration causes a new stack-root switch to be selected.

  • A network reconfiguration causes a new port on the current stack-root switch to be chosen as the stack-root port.

Note: The fast transition might not occur if multiple events occur simultaneously. For example, if a stack member switch is powered off while the link connecting the stack root to the spanning-tree root comes back up, the normal spanning-tree convergence occurs.

Normal spanning-tree convergence (30 to 40 seconds) occurs under the following conditions:

  • The stack-root switch is powered off, or the software failed.

  • The stack-root switch, which was powered off or failed, is powered on.

  • A new switch, which might become the stack root, is added to the stack.

  • A switch other than the stack root is powered off or failed.

  • A link fails between stack ports on the multidrop backbone.


These limitations apply to CSUF:

  • CSUF uses the GigaStack GBIC and runs on all Catalyst 3550 switches, all Catalyst 3500 XL switches, Catalyst 2950 switches with GBIC module slots, and only on modular Catalyst 2900 XL switches that have the 1000BASE-X module installed.

  • Up to nine stack switches can be connected through their stack ports to the multidrop backbone. Only one stack port per switch is supported.

  • Each stack switch can be connected to the spanning-tree backbone through one uplink.

  • If the stack consists of a mixture of Catalyst 3550, Catalyst 3500 XL, Catalyst 2950, and Catalyst 2900 XL switches, up to 64 VLANs with spanning tree enabled are supported. If the stack consists of only Catalyst 3550 switches, up to 128 VLANs with spanning tree enabled are supported.

  • The CSUF feature is supported only when the switch is running PVST .

  • If a stack that has Catalyst 2955, Catalyst 2950, or Catalyst 2940 switches also has Catalyst 2900 XL or Catalyst 3500 XL switches, cross-stack UplinkFast (CSUF) does not function if the management VLAN on the Catalyst 2900 XL or Catalyst 3500 XL switches is changed to a VLAN other than VLAN 1 (the default). The workaround is to make sure that the management VLANs of all Catalyst 2900 XL or 3500 XL switches in the stack are set to VLAN 1. (CSCdv82224)


The following are troubleshooting steps to take on a case by case scenario:

Error Message: NO_LOOP_DETECT

NO_LOOP_DETECT, GIGASTACK, LOG_ALERT, 0, The link neighbor of link %d of GigaStack

This error message indicates that there is no acknowledgment for GigaStack loop detection request received from one of the links on a GigaStack GBIC. Either the neighboring switch does not support the GigaStack loop breaking algorithm, or the link between the two GigaStack GBICs is broken. Under this condition, a GigaStack loop topology is not be automatically detected, and the connectivity between switches in the stack could be lost.

If loop topology is used in the GigaStack, make sure the latest software is running on all switches in the stack. Check the GigaStack GBICs involved to make sure they are functioning.

If no loop exists in the GigaStack GBIC stack, as a workaround, you can disable the loop-breaking algorithm by issuing the no gigastack loop-breaking [interface-id] global configuration command. This is meant as a workaround until you can upgrade all switches to support loop breaking.

One Way Traffic Flow

When using the GigaStack GBIC with a Catalyst 3500 XL switch or 2900 XL modular switch and IOS Release 11.2(8.1)SA6, you may experience a problem with traffic flow in one direction when you insert or remove a cable. This is resolved in 11.2(8.2) SA6 and greater.

Link Flaps and Never Stabilizes

When two switches are connected by a GigaStack GBIC and duplex is forced to full on both ends, the link may flap and never stabilize. When reconfigured to auto duplex on both ends, the link still will not stabilize. After reinserting one of the razors, the link finally stabilizes. This is also seen when link negotiation is shut off. Duplex must always be auto-negotiated when a GigaStack GBIC is inserted.

Configure the GigaStack ports to auto-negotiate duplex by issuing the interface command duplex auto. Configure the GigaStack port to negotiate the link by issuing the interface command negotiation auto. Once complete, remove and reinsert the GigaStack GBIC.

The show interface command will show the negotiated state for the GigaStack. Check for the following values in the show interface command output:

Auto-duplex (Half), 1000Mb/s, media type is CX_GIGASTACK

Note: Link and duplex auto negotiation are the default configuration for the GigaStack interfaces.


In a redundant loop configuration, an election of which GigaStack link is to be blocked occurs, also known as master loop breaker. All the LEDs on all GigaStacks will flash amber. Once the master loop breaker has been elected, all GigaStack LEDs will be solid green with the exception of the blocked master loop breaker. This link will be solid amber.

The following message is displayed on the elected master loop breaker switch:

%GIGASTACK-6-LOOP_DETECTED: Gigastack GBIC in  is selected as Master Loop Breaker.

Link 2 of the master loop breaker GigaStack GBIC is disabled to break the loop. Issuing the show interface command on the master loop breaker port will display the following:

GigaStack module(0.1) in GBIC slot. link1 is up, link2 is redundant link.

This is normal operation and the message is for informational purposes only. During this time, you may see the links go up/down multiple times. This is normal behavior. ]

The loop breaking algorithm occurs between the GigaStack GBICs by exchanging MAC addresses using their A- and B-ports. The winner of the election, the one with the lowest byte-order reversed MAC address, will become the master loop breaker and disable its port 2. For example, if a gig port has the MAC address 00:d0:79:6b:ee:01 (00d0.796b.ee01), the byte-order reversed MAC address is ee:01:79:6b:00:d0 (ee01.796b.00d0).

To determine the Master Loop Breaker, the reversed MAC addresses are broken into seven data bit chunks, with the first chunk having the highest significance as Master Loop Breaker. The loop breaker will be the seven data bit chunk with the lowest value. If the first seven data bit chunks are identical, then compare the next seven bits until there is a loop breaker.

To find out which switch will disable its B-port, issue the sh int gig x/y command on each switch and make note of the MAC address in parentheses (the one that says bia and then a MAC address). With this information, perform the byte-order reversing operation and then compare the first seven data bit chunks to find the lowest value. The reversed MAC address with the lowest value in its first seven data bit chunk will be the master loop breaker and will disable its port 2. If the first seven data bit chunks are identical, compare the next group until a non-identical comparison value can be found.

Related Information

Updated: Sep 01, 2005
Document ID: 22360