Extreme Networks Module 06 EXOS Stacking

Slide *Student ObjectivesUpon completion of this module, you will be able to:Describe the benefits of SummitStack Stacking TechnologyExplain how stacking operatesIdentify the various components of stacking Configure a set of devices to employ stacking Verify the stacking configuration Troubleshoot stacking issues

Slide *SummitStack Stacking Technology BenefitsChassis Simplifies management More redundancy and high-availability optionsRequires the allocation of more space than justified by initial needMore expensiveStandalone fixed-format switchesLess expensiveEasier to fit into open rack slotsEach new addition of access ports brings with it another device to manageTypically not as redundant

Slide *SummitStack Stacking Technology Benefits (Continued)Combined ManagementExtremeXOS SummitStack creates a single management point of control Pay as You GrowStart out with a single switch and grow to eight switches in a single stackMix Products to FitMix switches with different interface types and port densities in a single stack10/100/1000BASE-T / TX / mini-GBIC / AC and DC powered / PoE and non-PoEOptimized Stacking Architecture Throughput of up to 320 Gigabits per Second (Gbps) per stackDecreased Connectivity CostUsing Extreme Networks stacking is less expensive than interconnecting the devices using 10 Gigabit Ethernet linksCombined Management

Slide *Hardware RequirementsExtremeXOS allows the following hardware platforms to be joined together in a stackSummit x250eSummit x450, Summit x450a, Summit x450eStacking CableCables with special connectorsminimum length of 0.5 Maximum length of 5 M,

Order #NameDescription16106Stacking Cable, 0.5MSummitStack/UniStack stacking cable, 0.5M16107Stacking Cable, 1.5MSummitStack/UniStack stacking cable, 1.5M16108Stacking Cable, 3.0MSummitStack/UniStack stacking cable, 3.0M16105Stacking Cable, 5.0MSummitStack Stacking Cable, 5.0M (not supported for UniStack)

Slide *Software RequirementsSummitStack Stacking Technology requires ExtremeXOS Software Version 12.0 or laterUnits running less than ExtremeXOS 12.0 will not join stackExact same version requiredUnits with software version greater than or equal to 12.0 but not equal to stack master version become active stack members with disabled front panel portsUnits equal or greater than 12.0 can download software while in stackdownload image slot Stage and configure devices before deploying

Slide *Stacking Ports - Stacking ArchitectureStacking Ports2 stacking ports per device10 Gbps, Full Duplex per port40 Gbps stack capacity per switch320 Gbps stack capacity per 8 devices

20 Gbps per port40 Gbps per switchStacking bandwidth: up to 320 Gbps per stack

Slide *Unit Roles - Stacking ArchitectureMaster UnitManages the configuration of all unitsAcquires ownership of other unit's configurationShare configuration database with Backup Master UnitFunctionality similar to a Master MSMBackup Master UnitReady should the Master Unit failsDoes not own its own configurationAcquires co-ownership of standby unitsFunctionality similar to a secondary MSM in a chassisStandby UnitEssentially Slaves to the Master UnitASICs programmed by Master UnitFunctionality similar to an I/O module in a chassis

Slide *SEGMENTEDRing Topology - Stacking ArchitectureProvides redundancyWorks around stack cable failureWorks around a loss of a unitDaisy chain not supported for continuous operationsNo resiliencyCan lead to stack segmentationBehaves like a chassisStack cable similar to backplaneRINGDAISY-CHAIN

Slide *Topology Traffic - Traffic Handling - Stacking ArchitectureTraffic classified as:Topology TrafficKnown Unicast TrafficKnown Multicast TrafficVLAN Traffic (Unknown Unicast, Unknown Multicast, Broadcast Traffic)Topology Traffic:Uses hop-by-hop protocolManages the stackOnly stack configuration messagesProcessed by the CPUCan be used to configure stacking on devices even if the stack cannot formVery low bandwidth consumption12345678

Slide *Unicast Traffic - Traffic Handling - Stacking ArchitectureUnits send and receive traffic on both portsEnables a shortest path algorithmHandled by switching fabricSource switching ASIC forwards to Destination switching ASICNo stacking ports blockedKnown unicast traffic will not loop because it is not flooded2 "hops" to unit 8 through stack port 13 "hops" to unit 3 through stack port 212345786

Slide *Known Multicast Traffic - Traffic Handling - Stacking ArchitectureHandled by switching fabricEach ASIC aware of relevant Multicast configurations on other unitsBlocks ports to prevent loopsBased upon multicast addressAttempts to distribute multicast traffic evenly 12345678Blocked Ports for Multicast Address 1Blocked Ports for Multicast Address 2

Slide *Unknown Unicast / Broadcast / Unknown Multicast - Traffic Handling - Stacking ArchitectureAlso referred to as VLAN TrafficUnknowns must be flooded since destination location is unknownAll VLANs will have the same stacking ports blockedBlocked port is Stack Port 1 on unit with lowest MAC address plus interconnected port12345678Blocked Ports for all VLAN Traffic

Slide *Stack Join - Stacking OperationsDiscover Stack TopologyElect MasterElect Backup MasterStack ConfigurationOperational Phase12345678

Slide *Discover Stack Topology - Stack TopologyUnits exchange information until they determine stack topologyMust be ExtremeXOS 12.0 or greaterOccurs whether or not stacking is enabledUnits broadcast discovery packetThe CPU on each unit processes the packetEach Unit then increments hop count and forwards packetUnits remove packets with own MAC address12345678

Slide *Master / Backup Master Election - Stacking OperationsElection based uponOnly units with master-capability can participateHighest Stacking PriorityUser configuredPossible Values: AUTOMATIC, 1-100Highest priority winsDefault: AUTOMATICSlot Number (If Priority Tie)User configuredMust be uniqueNot necessarily related to physical position in stack or distance from MasterLowest slot number winsDefault: 1Automatic priority EXOS decides the election priorityCurrently slot number alone used, subject to revisionThe Backup Master is the one that came in second

Slide *Stack Configuration - Stacking OperationsMaster reads the configuration fileMaster shares configuration with backupMaster / Backup Master acquire ownership of other units configurationMaster implements configurationConfigures self (Only master self-configures)Configures backup master with RPC-like callsUses Control VLAN (control channel)Configures standby units with RPC-like callsSwitching engines updated as configuration is implemented

Slide *Operational Phase - Stacking OperationsTopology continuously verifiedRelevant configuration changes propagated throughoutSwitching decision made by ingress switch engineForwarded to destination port(s) Local switching requires no stack portStack is manageableUses VLAN mgmtStack MAC AddressStack IP Addressred vlan limited to single unityellow vlan exists on multiple units - requires stack ports

Slide *Stack Link Failure Recovery - Stack OperationsSingle link failureShortest path recalculated by all unitsBlock ports removedbefore: U5 to U4 - 1 hop via stack port 2after: U5 to U4 - 7 hops via stack port 112345786

Slide *Multiple Link Failure Recovery - Stack OperationsStack segmentation occursMaster / Backup master on different segmentsSegment with master Elect new backup masterPath recalculated on all unitsSwitching ASICs updated for missing unitsSegment with backup master Backup master becomes masterElect new backup masterRecalculated path on all unitsSwitching ASICs updated for missing unitsIssuesBoth segments retain Stack IP addressBoth segments retain Stack MAC addressN E W B A C K U P M A S T E RB A C K U P M A S T E RN E W B A C K U P M A S T E RCalculate new shortest pathM A S T E RBackup Master becomes Master Unit

Slide *Multiple Link Failure Recovery - Stack Operations (Continued)Master / Backup master on same segmentSegment with no mastersElect new masters (using same stack MAC as other segment!)Both SegmentsSwitching ASICs updated for missing unitsRecalculate shortest path on all unitsM A S T E RB A C K U P M A S T E RN E W M A S T E RN E W B A C K U P M A S T E RCalculate new shortest path

Slide *Unit Failure Recovery - Stack OperationsInvolves both the loss of the unit and the loss of the stacking portsRing integrity is compromised due to loss of stacking portStack enters fallback daisy-chain operationLost-Master Recovery TasksBackup Master immediately becomes MasterNew Backup Master unit selectedLost-BU-Master Recovery TasksNew Backup Master electedCommon Lost Unit Recovery TasksBlock ports removedShortest path recalculated by all unitsSwitching ASICs updated for missing unitsIs backup master lost?Is master lost?Lost-Master Recovery TasksYESYESLost-BU-Master Recovery TasksCommon Lost Unit Recovery TasksNONO

Slide *Preparing a Stack for ConfigurationUse the following procedure to prepare a stack for configuration:Select the appropriate stack units for your application and plan to use the stack as if it were a single multi-slot switch. Physically locate the stack nodes adjacent to one another.Ensure the exact same version of software is running on all stack nodes.Ensure all units have a default configuration Connect the stacking cables.If you intend to deploy new units that might be part of a stack in the future, you might want to turn on stacking mode during initial deployment to avoid a future restart.

Slide *Configuring a New StackApply power to all units in the stack.Log in to the intended stack master through the console port.Remove any legacy stacking configuration using the unconfigure stacking command.Verify the stack configuration using the show stacking and show stacking configuration commands.If necessary, configure license level restrictions using the configure stacking license-level command.Issue the enable stacking command.Answer "Yes" to perform easy setupAnswer "Yes" to proceedLog into the switch with admin privileges.safe-default script runs.Select values for normal operationVerify that the master node is the one you intended to be the master using the show stacking command.Verify remainder of configuration.Save the configuration using the save configuration command.enable stacking

Slide *Describing the Easy-Setup OptionThe easiest way to configure the stack is to use the following command:configure stacking easy-setupThe easy-setup option enables the administrator to effectively execute the following five commands:enable stackingconfigure stacking slot-number automaticconfigure stacking mac-address automaticconfigure stacking redundancy minimalreboot stack-topologyconfigure stacking easy-setup

Slide *Enabling / Disabling StackingTo enable stacking on one or all nodes in the stack[ enable | disable ] stacking {node-address }Targets a specific node or all units if node-address is absentWhen Stacking Enabled QoS profiles QP6 and QP7 are unavailablePerforms analysis of current stacking configuration if a node-address is not specifiedPrompts for easy setup if stack is unconfigured or configuration inconsistenciesDisabled nodesDo not forward customer's data through their stacking links Will not become a member of the active topologyWill become their own master and will process and execute their own configuration independentlyCommand takes effect at the node(s) next rebootTo Verifyshow stacking configurationUppercase E flag indicates if enabledDefault: Disabledenable / disable stacking

Slide *Configuring the Stacking Slot-numberConfigure the node's slot numberconfigure stacking slot-number automaticconfigure stacking node-address slot-number Automatic parameter assigns slot numbers automaticallyIn ring configuration: current node is assigned slot 1, unit connected to current node, stack port 2 is assigned slot 2, assignment continues to slot 8In daisy chain configuration: slot 1 is assigned to the node at the end of the chain that begins with the node connected to the current node's stack port 1Node-address parameter allows administrator to target specific unitThe Easy-Setup option executes the configure stacking slot-number automatic commandPorts addressed slot:port_number. Example 3:22Command takes effect at the node(s) next rebootVerify with show stacking configurationDefault: 1

configure stacking slot-number

Slide *Configure the Stacking MAC AddressYou must configure the stacking MAC address to make the stack manageable over the network by selecting a source node The system sets the Universal / Local bit to local to form the stack mac-address and then configure it on every node in the stackTo derive the stack MAC address from the current unit:configure stacking mac-addressThe easy setup option causes the stacking MAC address to be derived from the current unit.To derive the stack MAC address from another unit in the stack:configure stacking node-address macaddressconfigure stacking slot macaddressThis command takes effect only after you restart the node.Verify using: show stacking detailDefault: No stack MAC address is configured.configure stacking node-address

Slide *Configure Stacking RedundancyTo automatically configure master capabilities on all units in the stack, use this command:configure stacking redundancy [ none | minimal | maximal]none - Only one node has master-capability turned onminimal - Two nodes have master-capability turned onmaximal - All nodes have master-capability turned onMust be run on a stack with 8 nodes or lessNone and Minimal configurations use the slot-number and priorityTakes effect on next rebootThe Easy setup option executes the configure stacking redundancy minimal command.Verify using: show stacking configurationlowercase c flag indicates if the command has been executed

configure stacking redundancy

Slide *Rebooting the StackThe reboot option has a number of stack oriented options:reboot slotspecifies the slot number currently being used by the active stack node that isto be rebooted.reboot node-address specifies the MAC address of the SummitStack node to be rebootedreboot stack-topologyspecifies that the entire SummitStack is to be rebooted whether or not nodescare active.reboot stack-topology as-standby specifies that all stack nodes that are to be rebooted are to operate as if configured to not be master-capableThe Easy-Setup executes a reboot stack-topology command after configuring the stackreboot

Slide *Making the Non-Master Nodes IP ManageableConfigure an IP address to make the individual units in the stack manageable over the network.configure stacking alternate-ip-address [ | ] automaticconfigure stacking [node-address | slot ] alternate-ip-address [ | ] Assigns a secondary IP address to the each nodeAutomatic assign the specified IP Address to current nodeSubsequent nodes assigned specified IP Address + 1Alternate not configured if subnet mask differs from "Mgmt" VLANAlternate configured if no "Mgmt" VLAN IP definedConfiguration take effect immediatelyVerify using: show stacking configurationDefault: No address configure stacking alternate-ip-addressconfigure stacking alternate-ip-address

Slide *Configuring Stacking License LevelTo enable a unit to run at a lower license level than is installed:configure stacking {node-address | slot } license-level [core | advanced-edge | edge]Target a particular node using node-address or slot parameters.If node-address or slot parameter not specified, all nodes configured.Does not change the installed license level.Cannot be used to upgrade a nodes behavior.This command takes effect after you restart the node.Verify with: show stacking configurationAll nodes must be at the same effective levelDefault: No license level restriction is configured.configure stacking license-levelunconfigure stacking license-level

Slide *Synchronizing Stacking ParametersTo apply the stacking parameters from one node to another, use the command:synchronize stacking {node-address | slot }Copies NVRAM based configuration parameters to target node:stacking modestack MAC addressfailsafe account and passwordfailsafe account access point permissions the selected partitionTakes effect next node reboot.Verify using: show stacking detailDefault: Not applicable.synchronize stacking

Slide *Verifying Stack ConfigurationStack MAC in use: 02:04:96:27:2c:9fNode Slot Alternate AlternateMAC Address Cfg Cur Prio Mgmt IP / Mask Gateway Flags Lic------------------ --- --- ---- ------------------ --------------- ------- ---*00:04:96:27:2c:9f 1 1 Auto CcEeMm- -- 00:04:96:27:2f:52 2 2 Auto CcEeMm- -- 00:04:96:20:b4:05 3 3 Auto --EeMm- -- 00:04:96:20:aa:25 4 4 Auto --EeMm- --* - Indicates this nodeFlags: (C) master-Capable in use, (c) master-capable is configured, (E) Stacking is currently Enabled, (e) Stacking is configured Enabled, (M) Stack MAC in use, (m) Stack MACs configured and in use are the same, (i) Stack MACs configured and in use are not the same or unknown, (-) Not in use or not configuredLicense level restrictions: (C) Core, (A) Advanced edge, or (E) Edge in use, (c) Core, (a) Advanced edge, or (e) Edge configured, (-) Not in use or not configuredSlot-1 Stack.4 #show stacking configuration

Slide *Troubleshooting Stack OperationIf the stack is having problems forming, you can isolate the issue with the show stacking ports command.Stack Topology is a RingSlot Port Node MAC Address Port State Flags---- ---- ----------------- ----------- -----*1 1 00:04:96:27:2c:9f Operational C-*1 2 00:04:96:27:2c:9f Operational C- 2 1 00:04:96:27:2f:52 Operational C- 2 2 00:04:96:27:2f:52 Operational C- 3 1 00:04:96:20:b4:05 Operational C- 3 2 00:04:96:20:b4:05 Operational CB 4 1 00:04:96:20:aa:25 Operational CB 4 2 00:04:96:20:aa:25 Operational C-* - Indicates this nodeFlags: (C) Control path is active, (B) Port is BlockedSlot-1 Stack.8 #show stacking ports

Slide *Verifying Stack OperationsTo verify stack operation, use the show stacking command.The show stacking command provides:Node MAC Address, Slot, Stack State, Role, FlagsActive Topology is a RingNode MAC Address Slot Stack State Role Flags------------------ ---- ----------- ------- ---*00:04:96:27:2c:9f 1 Active Master CA- 00:04:96:27:2f:52 2 Active Backup CA- 00:04:96:20:b4:05 3 Active Standby CA- 00:04:96:20:aa:25 4 Active Standby CA-* - Indicates this nodeFlags: (C) Candidate for this active topology, (A) Active Node (O) node may be in Other active topologyshow stacking

Slide *Verifying Stack Operations (Continued)Stacking Node 00:04:96:27:2c:9f information:Current: Stacking : Enabled Role : Master Priority : Automatic Slot number : 1 Stack state : Active Master capable? : Yes License level restriction : In active topology? : Yes Factory MAC address : 00:04:96:27:2c:9f Stack MAC address : 02:04:96:27:2c:9f Alternate IP address : Alternate gateway : Stack Port 1: State : Operational Blocked? : No Control path active? : Yes Stack Port 2: State : Operational Blocked? : No Control path active? : YesConfigured: Stacking : Enabled Master capable? : Yes Slot number : 1 Stack MAC address : 02:04:96:27:2c:9f License level restriction : show stacking details

Slide *SummaryYou should now be able to:Describe the benefits of SummitStack Stacking TechnologyExplain how stacking operatesIdentify the various components of stacking Configure a set of devices to employ stacking Verify the stacking configuration Troubleshoot stacking issues

Slide *DemonstrationTurn to the Configuring a Stacked Switch Demonstration in the ExtremeXOS Operations and Configuration - Lab Guide Rev. 12.1

2008 Extreme Networks, Inc. All rights reserved. ExtremeXOS Operation and Configuration, Version 12.1. Part number DOC-00919.

Review Questions

2008 Extreme Networks, Inc. All rights reserved. ExtremeXOS Operation and Configuration, Version 12.1. Part number DOC-00919.

This presentation contains forward-looking statements that involve risks and uncertainties, including statements regarding our expectations as to products, trends and our performance. There can be no assurances that any forward-looking statements will be achieved, and actual results could differ materially from forecasts and estimates. For factors that may affect our business and financial results please refer to our filings with the Securities and Exchange Commission, including, without limitation, under the captions: Managements Discussion and Analysis of Financial Condition and Results of Operations, and Risk Factors, which is on file with the Securities and Exchange Commission (http://www.sec.gov). We undertake no obligation to update the forward-looking information in this release.

2008 Extreme Networks, Inc. All rights reserved. ExtremeXOS Operation and Configuration, Version 12.1. Part number DOC-00919.

The End2008 Extreme Networks, Inc. All rights reserved. ExtremeXOS Operation and Configuration, Version 12.1. Part number DOC-00919.2008 Extreme Networks, Inc. All rights reserved. ExtremeXOS Operation and Configuration, Version 12.1. Part number DOC-00919.

*Welcome to the Extreme Networks SummitStack Stacking Technology Training Module. *Upon completion of this module, you will be able to:

Describe the benefits of SummitStack Stacking TechnologyExplain how stacking operatesIdentify the various components of stacking Configure a set of devices to employ stacking Verify the stacking configuration and Troubleshoot stacking issues*In order to properly evaluate the benefits of SummitStack Stacking Technology, we'll need to compare it to some alternative technologies.

The access layer of a network can be implemented using either chassis or fixed-format switches. A chassis has the advantage of simplifying management and tends employ more redundancy and high-availability options. However, a chassis has the disadvantage of requiring the allocation of more space than what the initial port counts may require. Chassis-based systems also tend to be more expensive than fixed-format switches.

Standalone Fixed-format switches have the advantage of being less expensive and easier to fit into open rack slots, but have a disadvantage in that each new addition of access ports brings with it another device to manage. Also, fixed-format switches are typically not as redundant and do not have the same level of high-availability features as chassis switches.

*There are several benefits to employing ExtremeXOS SummitStack stacking technology in any network. For example:

By using stacking, you have the advantage of using Combined Management over several devices. ExtremeXOS SummitStack creates a single management point of control for configuring and managing all of the member switches in a stack. Configuring Layer 2 VLANs or Layer 3 routing interfaces is simplified with a single management view of all the ports in the entire stack.

Stacking technology allows you to Pay as You Grow. With ExtremeXOS SummitStack, you can start out with a single switch and grow to eight switches in a single stack. Features such as Link Aggregation, Multicast and Port Mirroring operate with ports on a single switch or operate on ports spread across multiple stack member switches. You can add ports to your Extreme Networks SummitStack when extra port density is needed.

Stacking enables you to mix products to fit the unique needs of your business. By using ExtremeXOS SummitStack, you can mix switches with different interface types and port densities in a single stack to support a range of applications.

Interfaces available are:

10/100/1000BASE-T available in AC and DC powered versions

10/100/1000BASE-T with Power Over Ethernet in AC powered versions

100/1000BASE-X mini-GBIC available in AC and DC powered versions

10/100BASE-TX with or without PoE in AC powered versions

Extreme SummitStack provides an optimized stacking Architecture. Other stacking technologies can reduce the management overhead of fixed-format switches, but often at a cost in overall performance and reliability. A stacking architecture that lacks adequate bandwidth or incurs forwarding penalties results in a serious loss of performance, which is unsuitable to support new converged applications.

The SummitStack stacking architecture was designed to provide significant throughput, up to 320 Gigabits per Second (Gbps) per stack, and the distributed, shortest path forwarding can provide performance comparable with chassis switches. Resiliency is of key importance for these applications and is provided by redundant bidirectional ring architecture and n-1 master redundancy, distributed Layer 2 and Layer 3 link aggregation, link redundancy and distributed uplinks.

You will realize Decreased Connectivity Costs with our stacking solution. Using Extreme Networks special stacking cables and built-in stacking ports to connect the supported switches is less expensive than purchasing 10 Gigabit modules to interconnect the devices.

The SummitStack stacking architecture delivers the best of both worlds: the benefits of a chassis at the cost of a stackable, in an architecture designed to support todays evolving LAN applications. The resulting network simplification can provide lower management and maintenance costs while enhancing overall availability.

*There are some hardware requirements in order to benefit from the SummitStack architecture.

The first requirement is that the stack be built from SummitStack Compatible Hardware. As of this writing, ExtremeXOS allows the following hardware platforms to be joined together in a stack:

Summit x250 Series SwitchesSummit x450 Series SwitchesSummit x450a Series SwitchesSummit x450e Series Switches

Each of these products supports various capabilities and media types. Please refer to company literature to ensure that you select switches that best meet your application's needs.

The second hardware requirement is the addition of stacking cables. The SummitStack Stacking Cables contain special connectors that support the high-speed ring topology of the SummitStack Architecture and meet the resiliency requirements of this feature.

Extreme Networks offers Stacking cables in four lengths - from a minimum to a half meter, to a maximum of five meters long. The order numbers are listed on the screen for your convenience and more information is available on the Extreme Networks corporate web site.*SummitStack Stacking Technology requires ExtremeXOS Software Version 12.0 or later. Units running less than ExtremeXOS 12.0 will not join stack.

Whichever version of ExtremeXOS Software you plan to deploy in your network, the SummitStack architecture requires that all units run the exact same version of software in order to ensure device interoperability and stack stability.

Units with software versions greater than 12.0 but not equal to stack master version become active stack members (provided there are no other stack topology issues) with disabled front panel ports. The non-compliant unit is still accessible to the stack master through the stack port. This allows the network administrator to download the appropriate version of code to the device using the:

download image

command shown on the screen. This command loads the appropriate version of software the non-conforming unit requiring that the unit be removed from the stack. NOTE: Since the front panel ports on the target unit are disabled, the tftp server must be accessible though the ports of another unit in the stack.

If possible, stage and configure devices before adding them to a live stack.

*The stacking architecture is very straight forward, but does require a bit of explanation:

First of all, there are two stacking ports per device. Each stacking port provides ten gigabits per second of full duplex bandwidth. Thats ten gigabits of bandwidth available to transmit traffic, and ten gigabits of bandwidth available to receive traffic for a total of twenty gigabits per second of bandwidth per port.

Each switch has two stacking ports, which provides the unit with the capacity to service an aggregate of forty gigabits per second of traffic through the stack connectors.

Finally, In a stack of eight switches, the stacking hardware can support up to three hundred twenty gigabits of throughput.*The SummitStack Architecture defines three types of roles for the units in the stack:

Master UnitBackup Master Unitand Standby Unit

The Master Unit manages the configuration of all units. No unit in the stack is allowed to look at it's configuration file and program its own switching ASICs. This is because the Master Unit has acquired the ownership of the configuration for all other units in the stack.

However, to ensure that the stack can survive it's loss, the master will share the configuration of the stack with the Backup Master Unit. The Backup Master maintains the configuration of the stack in memory, in case it is needed, but does nothing with it unless the Master Unit becomes unavailable.

You can compare the functionality of the Master Unit to that of a Master MSM. Much of the architecture and code for the SummitStack Technology is derived from ExtremeXOS chassis code that is used to manage IO modules.

The Backup Master Unit has a hybrid role in the stack. It must be ready to immediately assume the role of the master unit, should the master unit fail, however, it is like every other Standby Unit in that it does not own its own configuration. Remember, the Master Unit acquires ownership of the configuration of all units in the stack, not just the standby units.

In order to ensure that a failover condition (where the Master Unit fails and the backup unit takes over) is expedited in the most judicious fashion, the backup master unit will acquire co-ownership of the standby units configuration files. The Backup Master Unit will not exercise its right to configure the standby units unless the Master Unit has failed and the Backup Master Unit has assumed the role of Master Unit.

You can compare the functionality of the Backup Master Unit to that of a Backup MSM.

The last unit role we'll examine is that of the Standby Unit. These units are essentially slaves to the Master Unit. The ASICs are configured by the master, but the standby unit doesn't maintains a copy of the database in memory. In other words, the standby unit does what it is told, but doesn't know why it's doing it. However, for stack management purposes, the standby unit is aware of its stack configuration. It knows its own slot number, the stack MAC address, the master unit, and other parameters necessary to maintain stack operations.

In case of a stack error, such as the loss of the Master Unit or Backup Master Unit, a Standby Unit may become a Master Unit or a Backup Master Unit, depending upon its stack configuration.

Finally, you can compare the functionality of the Standby Unit to that of a chassis-based IO module.*To continue with the discussion of the stacking architecture, the SummitStack supports a ring topology during normal operation.

This means that every device is connected to another device in a daisy-chain fashion, and that last device is connected to the first device thus creating a ring. This type of topology provides the stack with the ability to sustain a stack cable failure, or the loss of the a unit and still be able to operate, with all units maintaining a physical and logical association with one another. This break in the ring creates a physical or logical daisy-chain topology.

For this reason, a daisy-chain topology is not supported for continuous or standard operations. A daisy-chain topology is only supported to address a stack failure condition.

A Daisy chain physical topology not supported for continuous operations. Unfortunately, a daisy-chain topology doesn't have the same level of resiliency as a ring, and the loss of another cable or unit could result in the stack becoming segmented and stack members becoming isolated - either physically or logically - from the rest of the stack members or from the network. *SummitStack Architecture has different methods for handling different types of traffic. The architecture classifies traffic into these four categories:

Topology TrafficUnicast TrafficMulticast TrafficVLAN Traffic which consists of Unknown Unicast, Unknown Multicast, and Broadcast Traffic

Topology Traffic is generated by a proprietary hop-by-hop protocol that is used to manage the stack's topology. Among other things, the Topology protocol is responsible for:

Interrogating the various units that are interconnect by the stack cable to determine their capabilities and stack configurationElecting a stack masterElecting a backup masterEnsuring a stack MAC address is configuredEnsuring each switch in the stack has a unique slot numberManaging all stack joins, merges, and failures

Topology traffic does not include any user data, nor does it include any non-stack related configuration messages. Only stack configuration messages are sent using the topology protocol.

Topology traffic is processed by the CPU, not by the switching ASICs.

The topology of the stack must be determined before user data can be forwarded. Therefore, the topology protocol must work before the stack is formed. Since it works before the stack is formed, it can be used to configure stacking options on devices in the stack even if there are stacking issues.

After the stack is formed, topology traffic consumes very little bandwidth.

*To continue on with the topic of how the SummitStack Architecture handles traffic:

The SummitStack Architecture allows the stack units to send and receive traffic using either stack port one or stack port two. This enables the switches in the stack to employ a shortest path algorithm for traffic forwarding operations.

Using the diagram on the slide, in order to reach unit eight, unit six transmits data out of it's stack port one to reach switch eight. However, the same unit, unit six, would use it's stack port two to transmit data to unit three.

If the path to the target is shorter using stack port two, then the system uses stack port two. The same holds true if the path to the target device is shorter using stack port one. Whichever port provides the shortest path, that's the path that the system will use.

Known unicast traffic is handled by the switch fabric.

When a frame enters the system through a non-stack port, the switching ASIC responsible for that port will lookup the destination MAC address in its forwarding database to determine which Module ID and port number is associated with the destination MAC address. The Module ID identifies a specific unit in the stack, as well as a specific switching ASIC in that unit.

The ingress switching ASIC forwards the frame to the egress switching ASIC by pre-pending the Ethernet frame with the destination Module ID and port number and source Module ID and port number. The destination Module ID and Port Number fields enable the frame to navigate the stack fabric to reach the frame's final destination.

Once the frame reaches it's destination, the destination ASIC places an entry in it's forwarding database associating the source Module ID and source port number with the MAC address - if it is not already known.

Because known unicast is not flooded, there is no danger this type of traffic looping around the stack. Therefore no stacking ports are blocked for known unicast traffic. This is not the case for broadcast, multicast, or unknown unicast traffic as you'll soon see.*The Stacking Architecture handles multicast traffic slightly different than it does known unicast traffic.

Multicast traffic is still handled by the switch fabric, and each ASIC in the switch is made aware of configurations that affect traffic forwarding. For example: If a multicast group only exists on one unit within the stack, there is no need for the other units in the stack to be made aware of that multicast group. However, if a multicast group is configured across several units in a stack, then that information must be distributed to the units in the stack. As the multicast frame comes into a unit on a stack, that ASIC will FLOOD traffic to all relevant front panel ports as well as the stacking ports.

Each unit in the stack must be made aware of the multicast group to enable them to either receive the multicast frames and deliver them to its own front panel ports or to allow the multicast traffic to traverse its stack ports on the way to another stack unit.

Because multicast traffic is flooded, and is allowed to transmitted onto the stack fabric, and the typical topology for the stack is a ring, there has to be a mechanism to ensure that multicast traffic will not continue to loop around the ring. The SummitStack Architecture answer is to create a logical break in the ring by blocking connected stacking ports on adjacent units to ensure multicast traffic cannot continuously circulate across the ring.

While this seems straight forward, there is one more optimization that is included in the way the SummitStack Architecture handles multicast traffic.

The decision as to which ports to block is based upon the multicast group address. Let's say that the first multicast group address is 224.0.0.1. The switch will logically block switch port one on unit three, and switch port two on unit four. This creates a logical stack break between units three and four.

The result of this operation is that all the units of the stack will receive the multicast traffic meant for 224.0.0.1, but that traffic is also prevented from looping around the stack.

Let's add to one more piece of data to the equation. Let's say our upstream router, the one behind which all our multicast servers are located, is connected to unit four. That means that multicast traffic destined for unit three would have to exit stack port 1 on unit 4 and traverse the entire stack before reaching unit 3. Now, what if you blocked the same ports for every multicast group? That would cause all multicast traffic from every group to have to traverse the entire stack before reaching unit 3.

But the SummitStack Architecture takes this into account and attempts to distribute multicast traffic more evenly by assigning blocking ports based upon the target address. Now let's say that we get traffic for a second multicast address of 224.0.0.2. The SummitStack Architecture will configure itself to logically block adjacent stacking ports on two units to prevent the multicast traffic from looping over the stack. In this case it may be stack port one on unit five and stack port two on unit six.

So, for packets destined for a device on unit three, and addressed to the multicast group of 224.0.0.2, they can get from unit 4 to unit 3 by exiting unit 4's stack port 2. Essentially one hop away.

*The SummitStack Architecture handles the following types of traffic the same way:

Unknown Multicast Unknown Unicast Broadcast

Because the destination port is unknown, Unknown Multicast, Unknown Unicast, and Broadcast traffic must be flooded to the entire VLAN - in fact this type of traffic is sometimes referred to as VLAN traffic because it is distributed to every port on a VLAN.

This type of traffic can come into from any port, so it doesn't make sense to try to equalize the distribution of traffic in the way that known multicast traffic handles distribution.

Therefore, all VLANs will have the same stacking ports blocked to ensure that frames do not endlessly circulate throughout the stack. The blocked port is Stack Port 1 on the unit with the lowest MAC address plus the interconnected port on the adjacent switch.

*Now that you know how the SummitStack Architecture handles traffic, let's take a look at how it operates. Obviously, the first thing that the stack must do is to join - or to become a single logical unit. From a very high level, the stack software must:

Discover Stack TopologyElect MasterElect Backup MasterStack SynchronizeOperational Phase

In the next few slides, we'll take a look at each one of these tasks in more detail.

*Note to reviewers: This is totally guesswork on my part. There is no document that describes how this is done, so I'm guessing at the design.

After power is applied to a stack, the unit initiate a discovery process to determine the stack's configuration. In order to participate in the stack discovery process, a unit must be running ExtremeXOS 12.0 or greater.

If a unit is running an appropriate ExtremeXOS software version, the stack discovery process occurs whether or not stacking is enabled on the device provided that one of the two stack port is connected to another powered unit. This doesn't mean that a unit that has stacking disabled will join the stack as an active unit, but it will participate in the routines that enable the stack to determine the topology of the stack plane.

First, each unit broadcasts a topology discovery packet (hello packet) out of each port (if connected) with information that describes their configuration. The information provided includes:

Stack Mode (enabled / disabled)Unit MAC AddressStack MAC addressModel NumberSlot #Stack PriorityStack Master CapableHops Count - Starting count of 0Unit License LevelAlternate IP / MaskAlternate Gateway

Next, the units that are connected to the sending device's stack ports receive the packet and forward them to their respective CPUs. The CPU on each unit processes the packet by extracting information about the unit that originated the packet, incrementing the hop count, and retransmitting the packet out of the port that did not receive the packet originally. Using this hop-by-hop forwarding technique, all the units in the stack will eventually have information about all their neighbors.

Finally, a when a unit receives a packet that has its own MAC address as the source, it will remove that packet from the stack plane and will note that the stack is a ring.

In a ring topology, each unit should have a path to every other unit through both of its stack ports. The unit will use the shortest path when making forwarding decisions.

During the discovery process, if the ring is broken, the units in the stack will not receive the topology packets that they sent (so they know that the stack is not a ring configuration), but they build a daisy chain topology based upon the packets received.

Once the discovery process is complete, the units that are qualified to become active stack members proceed to the next step of selecting a stack master. The units that are not qualified to go to the next step are:

Units with stacking mode disabledUnits with the same slot number as another device

*The Master and Backup Master election process is fairly straight forward.

Each unit is able to select the Master Switch based upon the information that they received during the discovery process.

First of all only units that have master-capability turned on can become masters.

Next, the units evaluate the stacking priority of each unit. This parameter is user configurable and can accept the values of AUTOMATIC, and 1 to 100. The unit with the highest priority becomes the stack master. If there are two or more units with the same stack priority (and the value is not automatic), then the system uses the slot number as the tie-breaker.

The slot number is a user configurable parameter. Each unit must be configured with a unique value from 1 to 8 in this parameter. While the slot number is derived from the slot number parameter in an Extreme Networks chassis system, the slot number in a stackable system does not correspond to the unit's physical position in stack or the distance from Master. This number is purely a logical number that is assigned by the network administrator.

In terms of electing the Master Unit, if more than one unit is tied with the highest stack priority value (and the parameter is not set to AUTOMATIC), the unit among these with the lowest slot number will become the Master Unit. At this point, the Master Unit should be selected.

The units in the stack will then go through the same process to elect the Backup Master Unit.

The default value for the stack priority value is AUTOMATIC. The AUTOMATIC setting allows the system to elect the stack master based upon a prioritization algorithm provide by Extreme Networks that may include factors such as CPU speed, memory, or number of ports. As of this writing, the AUTOMATIC setting has not been implemented.

If all units have their Stacking Priority parameter set to AUTOMATIC, the master unit will be elected based upon the unit with the lowest slot number.

*After the Master Unit and the Backup Master Unit have been elected, the Master reads its configuration file and shares it with Backup Master Unit. Now both the Master Unit and the Backup Master Unit have a copy of the stack's configuration.

The next thing that happens is that the Master Unit and Backup Master Unit inform the standby units (non-master units) that they have acquired ownership of the non-master unit's configuration files. This places the standby unit into an ACQUIRED state.

As long as the standby units are in an acquired state, and detect the presence of an acquiring unit on the stack, the units remain operational. If the acquiring units should disappear, and no other Master Units have acquired the standby units, then the standby units reboot because the integrity of their configuration cannot be verified.

After acquiring ownership of the non-master units, the master unit configures itself, the backup master, and the non-master units. To configure other units, the Master Unit uses a Remote Procedure Call (RPC) -like process to program the switching ASICs. Any future configuration changes will be executed using the same process.

Since the master unit is responsible for the non-stacking configuration of all the units in the stack, it is the only unit that configures itself. Even if there is a configuration file on the other units, it will not be used for configuring the unit.

*During the operational phase, units continue to verify the integrity of the stack, update their list of stack members, and maintain a logical association with the Master Units.

If the administrator issues configuration changes, the stack master propagates the changes to the relevant units. For example:If the administrator creates a VLAN that is confined to a single unit, it is not required that the other units in the stack be aware of that VLAN. However, if the administrator creates a VLAN that has port members on various units in the stack, then all the units must have their configuration updated. Obviously, those units with non-stacking ports that are part of the VLAN need to have their configurations update, but the other units need to have their stacking ports configured to allow the VLAN traffic to traverse the stack plane.

Remember, switching decisions are made by ingress switch engine. The switch engine looks up the destination in its forwarding table, and directs the packet to the appropriate port. If the destination port is on another stack member, the packet is sent to the appropriate stack port. If the destination is a port on the same unit, the packet is directed to that port the involvement of the stack plane.

Finally, the stack becomes manageable and accessible via the MGMT (management) VLAN. The stack master uses the configured Stack MAC Address (more on the Stack MAC address later) and Stack IP Address to make the stack manageable over the network. The network administrator can now initiate a network connection to the master unit to manage the stack.

*If there is a single stack link failure, the system can recover fairly easily. When the failure is detected, the units in the stack simple recalculate the shortest path to each of the units.

Since the ring is now physically segmented, there is no reason to logically block ports. Therefore, the logical port blocks for VLAN traffic and known multicast traffic is removed.

In the example on the screen, you can see that the stack cable between unit 4 and unit 5 has failed. Before the failure, unit 5 only needed 1 hop through its stack port 2 to reach unit 4. After the cable failure, unit 5 now finds itself 7 hops away from unit 4 via its stack port 1.

Finally, stacking cable failures are rare. If a stack does experience this type of failure, it is usually because somebody has disconnected the cable for maintenance purposes or by accident.*When a multiple link failure occurs, there are going to be issues that have to be overcome.

First of all, stack segmentation is going to occur. Units will not be able to communicate to one another like they could when the ring was whole or there was only one failed link. The first concern is the location of the master units. One of two scenarios can occur:

One segment has both the master and backup master units and the other segment has no master unit orOne segment has the master unit and the other segment has the backup master unit

If the Master Unit and the Backup Master Unit end up on different stack segments, then the segment with master needs to elect a new backup master, recalculate the path to other units, and update switching ASICs to remove entries for units that are no longer accessible to the units. In order to elect a new backup master, there must be another unit on the stack segment that is master-capable. If no other unit on the stack is configured as master-capable, then a Backup Master Unit cannot be elected.

On the segment with the Backup Master Unit, the Backup Master Unit becomes the Master Unit; a new backup master is elected if a master-capable unit is available, the stack-path to other units is recalculated by all units on the segment, and all units update their switching ASICs to remove entries for units that are no longer available.

This scenario causes two issues that make it difficult to manage either segment. The first issue is that the master on both segments retains the Stack IP Address. Both units could potentially respond to ARP requests. The other problem is that the master unit on both segments retains the Stack MAC address. This, of course, could cause problems with other device on the network with the Stack MAC address may appear to move.

*To continue with the Multiple Link Failure scenario, if a segment ends up with no master units, all units reboot because they have lost contact with both the Master Unit and the Backup Master Unit. From this point, the units in the stack segment will act as if they have just been powered up and go through the standard stack-join process; including attempting to elect master and backup master units. Unfortunately, these units will be using the same Stack MAC address and Stack IP address as the other segment.

Finally, if a segment ends up with both the Master and Backup Master Unit, it simply has to recalculate the shortest path between units.

Obviously, having two stack cable failures in the same stack at the same time is extremely rare. This type of failure is more than likely a result of human activity.*Now, we've looked at what happens if there is a cable failure, now we'll look at the stack's recovery process if a unit fails.

A unit failure involves both the loss of the unit and the loss of the stacking ports. Of course, the ring integrity is compromised due to loss of stacking port so the stack enters fallback daisy-chain operation.

If the lost unit is the Master Unit, the Backup Master Unit immediately becomes the Master Unit. Next, provided a master-capable switch is available in the stack, the stack will elect a unit to take the place of the Backup Master Unit. Finally, all logical blocks are removed from the stack ports, the shortest stack path is recalculated by all units, and the switching ASICs are updated to accommodate the loss of the missing unit.

If the lost unit is the Backup Master Unit, a new Backup Master is elected if there is a master-capable unit available in the stack. Finally, all logical blocks are removed from the stack ports, the shortest stack path is recalculated by all units, and the switching ASICs are updated to accommodate the loss of the missing unit.

If the lost unit is a standby or non-master unit, then all logical blocks are removed from the stack ports, the shortest stack path is recalculated by all units, and the switching ASICs are updated to accommodate the loss of the missing unit.*Before you can begin configuring a stack, you need to prepare and plan for the stack deployment. Use the following procedure to prepare a stack for configuration:

Select the appropriate stack units for your application and plan to use the stack as if it were a single multi-slot switch. You'll have to decide how many ports you'll need , port capacities, power-over-Ethernet capabilities, and power planning considerations. Physically locate the stack nodes adjacent to one another. Stacking cables come in fixed sizes that do not allow the various nodes to be physically distant from one another. It is probably best if the various nodes are physically located right next to one another. Make sure that you implement a stack ring topology. No other topology is supported for normal operations.

Ensure the exact same version of software is running on potential stack nodes. Even though the stack may join if there are different versions one another, it is a good management practice to simply standardize on a particular version of software. Ensure all units have default values. Go to the console port of each unit and execute the following command: unconfigure switch all

This step is technically unnecessary because the stack doesn't use the configuration files found on non-master nodes. However, to simplify the configuration process and minimize potential issues, it is best to return each switch to default values prior to enabling stacking.

If you omit this step, the configuration file will be saved on the unit under a different name.

Connect the stacking cables. Once this is done, you are ready to configure the stack.

If you intend to deploy new units that might be part of a stack in the future, you might want to turn on stacking mode during initial deployment to avoid a future restart.

*The easiest way to set up a stack is to configure the stack as a whole. In this way, all units will start out as if they were brand new, out-of-the-box.

Follow this procedure to configure stacking on a new switch:

Apply power to all units in the stack. Verify all units have power using the front panel LEDs. Remove any legacy stacking configuration using the unconfigure stacking command. This command resets all stacking parameters to the default or unconfigured values shown in the table on the screen. The following parameters are reconfigured by the command, but their new values do not take effect until the nodes reboot:

The stacking mode parameter is set to disabled. The slot-number parameter is set to 1. The master-capable parameter is set to Yes. The license-level restriction is unconfigured.The stack-MAC-Address is unconfigured.

Log in to the intended stack master through the console port. The user name should be admin, and there should be no password. The safe-script command may be executed since there is no configuration on the device. Answer the prompts as you normally would in your network environment.

Verify the stack configuration using the show stacking and show stacking configuration commands. Verify the state of the units in the stack - including which units are master and backup master units - by issuing the show stacking command. If necessary, configure license level restrictions using the configure stacking license-level command. Enable stacking by issuing the following command: enable stacking Since the stack has no configuration, the system will prompt you to use the easy setup option. Answer "Yes" to this prompt. Next, answer "Yes" to proceed to configure the device and reboot.

Log into the switch with admin privileges once the stack reboots. At this point - since the stack has a default configuration - the safe-default script runs. Select values for normal operation.

Verify that the master node is the one you intended to be the master.

Verify reminder of the configuration using the show stacking command.

Save the configuration

*The easiest way to configure the stack is to use the following command:

configure stacking easy-setup

The easy-setup option enables the administrator to effectively execute the following five commands:

enable stackingconfigure stacking slot-number automaticconfigure stacking mac-addressconfigure stacking redundancy minimalreboot stack-topology

The administrator could execute each of these commands on their own if they wanted to.*To enable stacking, use the following command:

enable stacking

This command accepts a node-address argument to allow you to target a specific unit provided you follow the node-address argument with the targeted unit's MAC address. All units are targeted by the command if you omit the node-address argument.

When a node is operating in stacking mode, QoS profiles QP6 and QP7 cannot be created. To ensure that the stack operates under heavy traffic conditions, topology traffic has been assigned the to QoS profile QP7. Configuration messages are assigned to QoS profile QP6.

If a node-address is not specified, this command first performs an analysis of the current stacking configuration on the entire stack. If the stack has not yet been configured for stacking operation, or there are configuration inconsistencies, the user is offered the option of invoking the easy setup function when the following message appears:

You have not yet configured all required stacking parameters.

Would you like to perform an easy setup for stacking operation? (y/N)

If you enter yes to the prompt, the easy setup procedure is invoked and following message is displayed:

Executing "configure stacking easy-setup" command...

If you enter no to the easy-setup prompt, a different message is displayed:

Stacking has been enabled as requested.

Disable Stacking

To disable stacking, use the following command:

disable stacking

This command accepts a node-address argument to allow you to target a specific unit provided you follow the node-address argument with the targeted unit's MAC address. All units are targeted by the command if you omit the node-address argument.

If a unit is a stack has stacking disabled, it will not forward the customer's data through its stacking links and will not become a member of the active topology. Also, a disabled node becomes its own master and processes and executes its own configuration independently.

The command does not take effect immediately, but at the node(s) next reboot. You may verify that this command has executed correctly by issuing the command:

show stacking configuration

The output of this command displays the current setting of the stacking flags. The lowercase e flag should not be set on the target unit(s). The absence of the lowercase e flag indicates that the target unit will have stacking disabled if it is rebooted. The uppercase E flag indicates whether or not stacking is currently enabled on the target device. The presence of an uppercase E flag indicates that stacking is currently enabled. The absence of the uppercase E flag indicates that stacking is currently disabled.

The default setting for stacking is disabled.

You may verify that the enable stacking command has executed correctly by issuing the command:

show stacking configuration

The output of this command displays the current setting of the stacking flags. The lowercase e flag should be set on the target unit(s). The presence of the lowercase e flag indicates that the target unit will have stacking enabled if it is rebooted. The uppercase E flag indicates whether or not stacking is currently enabled on the target device. The presence of an uppercase E flag indicates that stacking is currently enabled. The absence of the uppercase E flag indicates that stacking is currently disabled.

The default setting for stacking is disabled.

*If the network administrator does not choose to use the easy-setup option when enabling the stack, the network administrator must configure the stacking slot-number on each node or unit in the stack.

The slot number is a logical number that is a assigned to each node in the stack to make the node addressable by software. In a chassis system, the slot number represents a physical slot in the switch. However, in a stacked configuration, the slot number is a logical number from one to eight that is assigned to each unit by the administrator. Because the slot number is used to uniquely identify each node in the stack, a slot number can only be assigned to a single node. There is no check for duplicate slot numbers when the configure stacking slot-number command is executed; the number is simply assigned as requested.

To configure a node's slot number, use one of the two following commands:

configure stacking slot-number automatic

configure stacking node-address slot-number

The command using the automatic parameter assigns slot number automatically.

The Easy-Setup option executes the configure stacking slot-number automatic command

The nodes in the stack topology are assigned the numbers in the order in which they would appear currently in the show stacking command output.

If the stack cables are joined in a ring configuration, the current node is assigned slot 1. The node connected to stack port 2 of the current node is assigned slot 2. This numbering scheme continues until all nodes have been assigned slot numbers or slot numbers have been assigned to eight nodes.

If the stack cables are joined in a daisy-chain configuration, slot 1 is assigned to the node at the end of the chain that begins with the node connected to the current node's stack port 1. This should be the node that does not have another node connected to its stack port 1. The node connected to stack port 2 of the node in slot 1 is assigned slot 2. This numbering scheme continues until all nodes have been assigned slot numbers or slot numbers have been assigned to eight nodes.

Before the command is executed when the automatic option was specified, the following confirmation is solicited:

Reassignment of slot numbers may make the stack incompatible with the current configuration file.Do you wish to continue? (y/n)

When this command is executed successfully, the following message is displayed:

This command will take effect at the next reboot of the specified node(s).

The configure stacking slot-number command syntax that uses the node-address parameter allows the network administrator to configure the slot number on a specific unit by specifying the keyword node-address followed by the MAC address of the target device.

The slot number can be used by various commands to target individual units. For example: By using the slot parameter, the network administrator may direct the system to download a new image to a particular node in the stack.

To identify ports in the stack, the network administrator uses the slot:port syntax. For example, to identify port 22 of node 3, the administrator would enter 3:22.

The command does not take effect immediately, but at the node(s) next reboot. You may verify that this command has executed correctly by issuing the command:

show stacking configuration

The output of this command displays the configured slot-number setting and the current slot-number setting. Remember that the configured setting does not take effect until the node is rebooted.

The default setting for the slot-number parameter is 1.

*You must configure the stacking mac-address to make the stack manageable over the network.

You must select a node whose factory assigned MAC address will be used to form a MAC address that will represent the stack as a whole. The system forms the stack MAC address by setting the Universal / Local bit in the specified MAC address. This means that the stack MAC address is a locally administered address, and not the universal MAC address assigned to the selected node. The stack mac-address is then configured on every node in the stack topology.

If you move the unit from which the stack mac-address is derived to another stack, the stack mac-address will not be changed on the original stack. You need to be careful not to derive the mac-address from same unit on two different stacks. This would result in duplicate mac-addresses on the network.

To derive the stack mac-address from the current unit, use the following command:

configure stacking mac-address

To derive the stack mac-address from another unit in the stack, you must identify the source node either by MAC address or slot number. Use one of the following forms of the command to accomplish this task:

configure stacking node-address macaddressconfigure stacking slot macaddress

This command takes effect only after you restart the node. The following message appears after you run the command:

This command will take effect at the next reboot of the specified node(s).

If a stack node that has just joined the stack detects that its stack MAC address is not configured or is different than the stack MAC address in use, it will log the following message at the Error log level:

The stack MAC address is not correctly configured on this node. The stack can not operate properly in this condition. Please correct and reboot.

If you have not configured (or inconsistently configured) the stack MAC address you might encounter difficulty in diagnosing the resulting problems. Whenever the master node (including itself) detects that one or more nodes in its active topology do not have the correct or any stack MAC address configured, it will display the following message to the console every five minutes until you configure a MAC address and restart the node(s):

The stack MAC address is either not configured or its configuration is not consistent within the stack. The stack can not operate properly in this condition. Please correct and reboot.

You may verify the results of this command by using the command:

show stacking detail

By default, no stack mac-address is configured.*In order to simultaneously configure master capability on more than one unit, use the command:

configure stacking redundancy [none | minimal | maximal]

This command defines the level of Master Unit redundancy that will be configured on the stack. A redundancy setting of none configures the stack to have only one unit that is capable of being a master unit. This means that no other unit can provide redundancy and take over the Master Units functionality if the Master unit should fail.

A redundancy setting of minimal configures two nodes in the stack to have master-capability turned on and all other nodes will have master capability turned off. This means the stack will have one Master Unit and one Backup Master Unit. All other nodes in the stack are not able to provide any redundancy should the Master Units fail. The Easy setup option executes the configure stacking redundancy minimal command.

Finally, a redundancy setting of maximal configures all nodes all nodes in the stack to have master-capability turned on.

To run this command, the stack should not contain more than eight nodes in its stack topology. If there are more than eight nodes in the stack topology, the following message appears and the command is not executed:

ERROR: This command can only be used when the stack has eight nodes or less.

Since only eight nodes can be operational in an active topology at a time, you must disconnect the remaining nodes before configuring master-capability using this command.

If you are using the none or minimal redundancy configuration:

The configured values of slot-number and priority decide the nodes on which the master-capability should be turned on.If the priority values are configured on the nodes, the highest priority node(s) will be chosen.If the priority values of all nodes are set to automatic or to the same priority value, the node(s) with the lowest slot number(s) will be chosen. Extreme Networks may change automatic priority behavior in a future release.

If there is a slot number tie or if neither slot-number nor priority were ever configured, the following message appears and the command is not executed:

ERROR: Unique slot numbers must be configured before using this command.

The master-capability setting does not take effect immediately, but at the node(s) next reboot. You may verify that this command has executed correctly by issuing the command:

show stacking configuration

The output of this command displays the current setting of the stacking flags. The uppercase C flag indicates whether or not master-capability is currently enabled on the target node. The presence of an uppercase C flag indicates that master-capability is currently enabled. The absence of the uppercase C flag indicates that master-capability is currently disabled.

The lowercase c flag indicates whether or not master-capability is configured on the target node. The presence of an lowercase c flag indicates that master-capability will be enabled following a reboot. The absence of the lowercase c flag indicates that master-capability will be disabled following a reboot.

If you issued the command with the none parameter, there should be only one unit the lowercase c flag set. The minimal option should leave two units with the lowercase c flag set, and maximal leaves all units with the lowercase c flag set.

*The reboot option has a number of stack oriented options:reboot slotspecifies the slot number currently being used by the active stack node that isto be rebooted.reboot node-address specifies the MAC address of the SummitStack node to be rebootedreboot stack-topologyspecifies that the entire SummitStack is to be rebooted whether or not nodescare active.reboot stack-topology as-standby specifies that all stack nodes that are to be rebooted are to operate as if configured to not be master-capable*Configure an IP address to make the stack manageable over the network. To accomplish this task. use the command:

configure stacking alternate-ip-address

This command assigns an alternate IP address to the management VLAN. As you can see, there are two forms of this command; one using the automatic keyword, and the other using the node-address or slot keywords. When using the node-address or slot keywords, you instruct the system to assign an alternate IP address to a specific unit in the stack. This form of the command operates on one node at a time.

When using the automatic keyword, you instruct the system to assign the specified IP address to the first node listed in the show stacking display. The remainder of the devices on the list are assigned consecutive IP addresses starting with the specified IP address plus 1. For example, if the first unit on the list was assigned the IP address of 10.10.10.1, the next unit on the list would be assigned 10.10.10.2, and so on.

Since there is a specified subnet mask, the address will be checked to insure that the block of IP addresses fits within the specified subnet given the number of nodes in the stack topology. The address block is tested to insure that each address is a valid IP unicast address. If the test fails, no node is configured and an error message is printed.

The other command options are fairly straight forward.

The node-address parameter specifies the MAC address of the target node.

The slot-number parameter identifies the slot number of the target unit.

The ipaddress and netmask parameters enable you to separately identify the IP address and subnet mask you wish to assign to the target node or nodes. For example, you would specify both parameters using dotted decimal notation so that the IP address would appear as 66.77.88.99 and the subnet mask would appear as 255.255.255.0.

The ipNetmask parameter also enables you to identify the IP address and subnet mask you wish to assign to the target node or nodes, but as a single, slash-delimited value. For example: The IP address is specified using standard dotted decimal notation, and the subnet mask is specified as the number of leading bits in the subnetwork address, with a slash (/) in between the two parameters. The result may appear as 66.77.88.99/24. You may use the ipNetmask parameter or the ipaddress and netmask parameter pair when configuring the devices.

The gateway parameter identifiers the address of an IP router.

The alternate-ip-address is not applied if the subnet mask differs from the subnet mask already assigned to the management VLAN. The alternate-ip-address is applied if the command's subnet mask is the same as the subnet mask used to assign an IP address to the management vlan, or there is no IP address assigned to the management VLAN.

The alternate-ip-address and its associated parameters are not used unless the node is operating in stacking mode.

The configuration takes effect immediately after the command is successfully executed.

You may verify the configuration by issuing the command:

show stacking configuration

By default, there is no alternate-ip-address configured.

*To enable a unit to run at a lower license-level than is installed, use the command:

configure stacking license-level

This command has three parameters, node-address, slot, and license level. The node-address parameter specifies the MAC address of the target node. The slot-number parameter identifies the slot number of the target unit. If the node-address or slot parameter is not specified, the command takes effect on every node in the stack topology.

The license-level parameter is a required parameter that specifies the license level to which you wish to restrict the target node or nodes.

This command does not change the installed license level. For example, if a node is configured with the Advanced Edge license and you configure a license level restriction of Edge, the unit is restricted to features available in the Edge license. This license-level restriction is not permanent and may removed or reconfigured at your convenience.

You must purchase and install a new license in order to upgrade a nodes behavior. You may not cause a node with an Edge License installed to expand its capabilities by issuing a command to restrict the license-level to advanced-edge or core. If the installed license level of the target node is lower than the level you are attempting to configure, a message appears warning you that the switch will not operate at a license level beyond that which was purchased.

This command takes effect after you restart the targeted nodes.

To verify the command has taken effect, use the command:

show stacking configuration

All nodes must be at the same effective level in order for the stack topology to operate. The effective license level will appear only when stacking is enabled. The command is node-specific. The effective license level is the level at which the node is restricted to operate, and is not necessarily. the level at which the entire stack is operating.

Default: No license level restriction is configured. If you restart the node without configuring a license level restriction, the node operates at the purchased license level. *When adding a new device to a stack, the network administrator many configure the new device with the stacking parameters already configured on the stack. To apply the stacking parameters from one node to another, use the command:

synchronize stacking

This command has two optional parameters: node-address and slot-number. These parameters are used to target a specific node in the stack to synchronize with the current node. If the administrator uses one of these two parameters, the target node is synchronized with the current node. If the administrator omits these two parameters, all nodes are synchronized with the current node.

This command copies the following NVRAM based configuration parameters to the target node:

stacking modestack MAC addressfailsafe account and passwordfailsafe account access point permissions (whether the failsafe account is allowed over the stacking links, console port, or management port)the selected partition

A default value is not applicable to this command.

*The show stack configuration command enables the network administrator to verify the configuration of the stack options. The command provides a great deal of information to include:

Node MAC AddressConfigured Slot #Current Slot #Slot priorityAlternate Management IP and MaskAlternate Gateway IP addressFlags License level restriction

*Troubleshooting Stack Operation

You may troubleshoot stack operations by executing the following command:

show stacking ports

If there is an issue with the physical operation of the stack and the various stack ports, you can use the command to identify failed units or stacking cables.*To verify stack operation, use the following command:

show stacking

The show stacking command provides:Node MAC AddressSlotStack StateRoleFlags

*To obtain more detailed information about the configuration of various modules in the stack, use the detail option of the show stacking command:

show stacking details

This command provides a comprehensive display of the configuration of each device. *You should now be able to:

Describe the benefits of SummitStack Stacking TechnologyExplain how stacking operatesIdentify the various components of stacking Configure a set of devices to employ stacking Verify the stacking configuration Troubleshoot stacking issues

*