Nexus 7000 virtual Port-Channel Best Practices & Design Guidelines

Nexus 7000 virtual Port-ChannelBest Practices & Design Guidelines Roberto MariTechnical Marketing Engineer Data Center Business Unit November 2009 version 1.1

Agenda • Feature Overview & Terminology • vPC Design Guidance & Best Practices • Building a vPC domain • Attaching to a vPC domain • Layer 3 and vPC • Spanning Tree Recommendations • Data Center Interconnect (& Encryption) • HSRP with vPC • vPC and Services • vPC latest enhancements • ISSU • Convergence and Scalability • vPC Hands-on Lab Information • Reference Material

Feature Overview & TerminologyvPC Definition • Allow a single device to use a port channel across two upstream switches • Eliminate STP blocked ports • Uses all available uplink bandwidth • Dual-homed server operate in active-active mode • Provide fast convergence upon link/device failure • Reduce CAPEX and OPEX • Available on current and future hardware for M1 and D1 generation cards. Logical Topology without vPC Logical Topology with vPC

Feature Overview & Terminology vPC Terminology • vPC peer – a vPC switch, one of a pair • vPC member port – one of a set of ports (port channels) that form a vPC • vPC – the combined port channel between the vPC peers and the downstream device • vPC peer-link – Link used to synchronize state between vPC peer devices, must be 10GbE • vPC peer-keepalive link – the keepalive link between vPC peer devices, i.e., backup to the vPC peer-link • vPC VLAN – one of the VLANs carried over the peer-link and used to communicate via vPC with a peer device. • non-vPC VLAN – One of the STP VLANs not carried over the peer-link • CFS – Cisco Fabric Services protocol, used for state synchronization and configuration validation between vPC peer devices vPC peer-keepalive link vPC peer-link CFS protocol vPC peer vPC vPC member port vPC member port vPC non-vPC device

Following steps are needed to build a vPC (Order does Matter!) Configure globally a vPC domain on both vPC devices Configure a Peer-keepalive link on both vPC peer switches (make sure is operational)NOTE: When a vPC domain is configured the keepalivemust be operational to allow a vPC domain to successfully form. Configure (or reuse) an interconnecting port-channel between the vPC peer switches Configure the inter-switch channel as Peer-link on both vPC devices (make sure is operational) Configure (or reuse) Port-channels to dual-attached devices Configure a unique logical vPC and join port-channels across different vPC peers Building a vPC DomainConfiguration Steps vPC peer-link vPC peer-keepalive link vPC peer Standalone Port-channel vPC member port vPC

Definition: Standard 802.1Q Trunk Can Carry vPC and non vPC VLANs* Carries Cisco Fabric Services messages (tagged as CoS=4 for reliable communication) Carries flooded traffic from a vPC peer Carries STP BPDUs, HSRP Hellos, IGMP updates, etc. Requirements: Member ports must be 10GE interfaces one of the N7K-M132XP-12 modules Peer-link are point-to-point. No other device should be inserted between the vPC peers. Recommendations (strong ones!) Minimum 2x 10GbE ports on separate cards for best resiliency. Dedicated 10GbE ports (not shared mode ports) Building a vPC DomainPeer Link vPC peer-link *It is Best Practice to split vPC and non-vPC VLANs on different Inter-switch Port-Channels.

Building a vPC DomainPeer Link with Single 10G Module • Common Nexus 7000 configuration:1x10G, 7x1G cards • vPC recommendation is 2 10G cards • Potential problem occurs if Nexus 7000 is L3 boundary with single 10G card • Use Object Tracking Feature available in 4.2 • More information from CCO:http://www.cisco.com/en/US/docs/switches/datacenter/sw/4_2/nx-os/interfaces/configuration/guide/if_vPC.html#wp1529488

Building a vPC DomainPeer Link with Single 10G Module – Object Tracking Scenario: • vPC deployments with a single N7K-M132XP-12 card, where core and peer-link interfaces are localized on the same card. • This scenario is vulnerable to access-layer isolation if the 10GE card fails on the primary vPC. vPC Object Tracking Solution: • Leverages object tracking capability in vPC (new CLI commands are added). • Peer-link and Core interfaces are tracked as a list of boolean objects. • vPC object tracking suspends vPCs on the impaired device, so traffic can get diverted over the remaining vPC peer. e1/… e1/… e1/… e1/… e1/… vPC PL e1/… L3 e1/… e1/… L2 vPC PKL e2/… e2/… vPC Primary vPC Secondary rhs-7k-1(config-vpc-domain)# track <object>

Definition: Heartbeat between vPC peers Active/Active (no Peer-Link) detection Messages sent on 2 second interval 3 second hold timeout on peer-link loss Fault Tolerant terminology is specific to VSS and deprecated in vPC. Packet Structure: UDP message on port 3200, 96 bytes long (32 byte payload), includes version, time stamp, local and remote IPs, and domain ID. Keepalive messages can be captured and displayed using the onboard Wireshark Toolkit. Recommendations: Should be a dedicated link (1Gb is adequate) Should NOT be routed over the Peer-Link Can optionally use the mgmt0 interface (along with management traffic) As last resort, can be routed over L3 infrastructure Building a vPC DomainPeer-Keepalive (1 of 2) vPC peer-keepalive link

Cautions/Additional Recommendations: When using supervisor management interfaces to carry the vPC peer-keepalive, do not connect them back to back between the two switches. Only one management port will be active a given point in time and a supervisor switchover may break keep-alive connectivity Use the management interface only if you have an out-of-band management network (management switch in between). Building a vPC DomainPeer-Keepalive (2 of 2) Management Network Standby Management Interface Management Switch Active Management Interface vPC_PK vPC_PK vPC_PL vPC1 vPC2

Building a vPC DomainvPC Member Port • Definition: • Port-channel member of a vPC peer. • Requirements: • Configuration needs to match other vPC peer’s member port config. • In case of inconsistency a VLAN or the entire port-channel may suspend (i.e. MTU mismatch). • Number of member ports on both vPC peers is not required to match. • Up to 8 active ports between both vPC peers (16-way port-channel can be build with multi-layer vPC) vPC member port vPC member port

Building a vPC DomainVDC Interaction • vPC works seamlessly in any VDC based environment. • One vPC domain per VDC is supported, up to the maximum number of VDCs supported in the system. • It is still necessary to have a separate vPC peer-link and vPC Peer-Keepalive Link infrastructure for each VDC deployed. Can vPC run between VDCs on the same switch? • This scenario should technically work, but it is NOT officially supported and has not been extensively tested by our QA team. • Could be useful for Demo or hands on, but It is NOT recommended for production environments. Will consolidate redundant points on the same box with VDCs (e.g. whole aggregation layer on a box) and introduce a single point of failure. • ISSU will NOT work in this configuration, because the vPC devices can NOT be independently upgraded.

Attaching to a vPC domainThe One and Only Rule… ALWAYSdual attach devices to a vPC Domain!!!

Attaching to a vPC DomainIEEE 802.3ad and LACP • Definition: • Port-channel for devices for devices dual-attached to the vPC pair. • Provides local load balancing for port-channel members • STANDARD 802.3ad port channel • Access Device Requirements • STANDARD 802.3ad capability • LACP Optional • Recommendations: • Use LACP when available for better failover and mis-configuration protection vPC vPC member port Regular Port-channel port

Attaching to a vPC Domain”My device can’t be dual attached!” Recommendations (in order of preference): • ALWAYS try to dual attach devices using vPC (not applicable for routed links). PROS: Ensures minimal disruption in case of peer-link failover and consistent behavior with vPC dual-active scenarios. Ensures full redundant active/active paths through vPC.CONS: None • If (1) is not an option – connect the device via a vPC attached access switch (could use VDC to create a “virtual access switch”). PROS: Ensures minimal disruption in case of peer-link failover and consistent behavior with vPC dual-active scenarios. Availability limited by the access switch failure. CONS: Need for an additional access switch or need to use one of the available VDCs. Additional administrative burden to configure/manage the physical/Virtual Device • If (2) is not an option – connect device directly to (primary) vPC peer in a non-vPC VLAN* and provide for a separate interconnecting port-channel between the two vPC peers. PROS: Traffic diverted on a secondary path in case of peer-link failover CONS: Need to configure and manage additional ports (i.e. port-channel) between the Nexus 7000 devices. • If (3) is not an option – connect device directly to (primary) vPC peer in a vPC VLANPROS: Easy deployment CONS: VERY BAD.Bound to vPC roles (no role preemption in vPC) , Full Isolation on peer-link failure when attached vPC toggles to a secondary vPC role. * VLAN that is NOT part of any vPC and not present on vPC peer-link

Attaching to a vPC DomainvPC and non-vPC VLANs (i.e. single attached .. ) P S P S 2. Attached via VDC/Secondary Switch 1. Dual Attached Orphan Ports Orphan Ports P S P S P Primary vPC S Secondary vPC 3. Secondary ISL Port-Channel 4. Single Attached to vPC Device

Attaching to a vPC Domain”My device only does STP!” Recommendations (in order of preference): • ALWAYS try dual attach devices using vPCPROS: Ensures minimal disruption in case of peer-link failover and consistent behavior with vPC dual-active scenarios. Ensures full redundant active/active paths through vPC.CONS: None • If (1) is not an option – connect the device via two independent links using STP. Use non-vPC VLANs ONLY on the STP switch.*PROS: Ensures minimal disruption in case of peer-link failover and consistent behavior with vPC dual-active scenarios. Ensures full redundant Active/Active paths on vPC VLANs.CONS: Requires an additional STP port-channel between the vPC devices. Operational burden in provisioning and configuring separate STP and vPC VLAN domains. Only Active/Standby paths on STP VLANs. • If (2) is not an option – connect the device via two independent links using STP. (Use vPC VLANs on this switch)PROS: Simplify VLAN provisioning and does not require allocation of an additional 10GE port-channel. CONS: STP and vPC devices may not be able to communicate each other in certain failure scenarios (i.e. when STP Root and vPC primary device do not overlap). All VLANs carried over the peer-link may suspend until the two adjacency forms and vPC is fully synchronized". * Run the same STP mode as the vPC domain. Enable portfast/port type edge on host facing ports

vPC Design principlesAttaching to a vPC Domain - vPC and non-vPC VLANs (STP/vPC Hybrid) Non vPC port-channel P SR PR S S P 1. All devices Dual Attached via vPC 2. Separate vPC and STP VLANs SR PR P S P Primary vPC S Secondary vPC PR Primary STP Root SR Secondary STP Root 3. Overlapping vPC and STP VLANs

Attaching to a vPC Domain16-way Port-Channel (1 of 2) • Multi-Layer vPC can join 8 active ports port-channels in a unique 16-way port-channel* • vPC peer side load-balancing is LOCAL to the peer • Each vPC peer has only 8 active links, but the pair has 16 active load balanced links Nexus 7000 16-way port channel Nexus 5000 * Possible with any device supporting vPC/MCEC and 8-way active port-channels

Attaching to a vPC Domain16-way Port-Channel (2 of 2) • 16 active ports between 8 active port-channel devices and 16 active port-channel devices? • vPC peer side load-balancing is LOCAL to the peer • Each vPC peer has only 8 active links, but the pair has 16 active load balanced links to the downstream device supporting 16 active ports • D-series N7000 line cards will also support 16 way active port-channel load balancing, providing for a potential 32 way vPC port channel! Nexus 7000 16-port port-channel Nexus 5000 Nexus 5000 16-port port-channel support introduced in 4.1(3)N1(1a) release

Use separate L3 links to hook up routers to a vPC domain is still standing. Don’t use L2 port channel to attach routers to a vPC domain unless you can statically route to HSRP address If both, routed and bridged traffic is required, use individual L3 links for routed traffic and L2 port-channel for bridged traffic Layer 3 and vPCRecommendations Switch Switch Po2 Po2 7k1 7k2 L3 ECMP Po1 Router Router

Layer 3 and vPCWhat can happen… (1 of 3) vPC view Layer 2 topology Layer 3 topology 7k vPC 7k1 7k2 7k1 7k2 R R R R could be any router, L3 switch or VSS building a port-channel Port-channel looks like a single L2 pipe. Hashing will decide which link to chose Layer 3 will use ECMP for northbound traffic

Layer 3 and vPCWhat can happen… (2 of 3) • Packet arrives at R • R does lookup in routing table and sees 2 equal paths going north (to 7k1 & 7k2) • Assume it chooses 7k1 (ECMP decision) • R now has rewrite information to which router it needs to go (router MAC 7k1 or 7k2) • L2 lookup happens and outgoing interface is port-channel 1 • Hashing determines which port-channel member is chosen (say to 7k2) • Packet is sent to 7k2 • 7k2 sees that it needs to send it over the peer-link to 7k1 based on MAC address S Po2 7k1 7k2 Po1 R

Layer 3 and vPCWhat can happen… (3 of 3) • 7k1 performs lookup and sees that it needs to send to S • 7k1 performs check if the frame came over peer link & is going out on a vPC. • Frame will only be forwarded if outgoing interface is NOT a vPC or if outgoing vPC doesn’t have active interface on other vPC peer (in our example 7k2) S Po2 7k1 7k2 Po1 R

Spanning Tree RecommendationsOverview – STP Interoperability • STP Uses: • Loop detection (failsafe to vPC) • Non-vPC attached device • Loop management on vPC addition/removal • Requirements: • Needs to remain enabled, but doesn’t dictate vPC member port state • Logical ports still count, need to be aware of number of VLANs/port-channels deployed! • Best Practices: • Not recommended to enable Bridge Assurance feature on vPC channels (i.e. no STP “network” port type). Tracked by CSCsz76892. • Make sure all switches in you layer 2 domain are running with Rapid-PVST or MST (IOS default is non-rapid PVST+), to avoid slow STP convergence (30+ secs) • Remember to configure portfast (edge port-type) on host facing interfaces to avoid slow STP convergence (30+ secs) STP is running to manage loops outside of vPC’s direct domain, or before initial vPC configuration vPC vPC

Spanning Tree RecommendationsPort Configuration Overview N Network port E Edge or portfast port type Normal port type - B BPDUguard R Rootguard Data Center Core L Loopguard Primary vPC Secondary vPC vPC Domain HSRP ACTIVE HSRP STANDBY Layer 3 Aggregation SecondaryRoot N N Primary Root Layer 2 (STP + Rootguard) - - - - - - - - R R R R R R R R - Access - L - E E E E E B B B B B Layer 2 (STP + BPDUguard)

N Network port Data Center InterconnectMulti-layer vPC for Agg and DCI E Edge or portfast port type Normal port type - BPDUguard B F BPDUfilter R Rootguard DC 1 DC 2 vPC domain 11 vPC domain 21 Long Distance CORE CORE F - - F - - N N N N - - F F - - AGGR R R AGGR - - R R - - N N N N vPC domain 10 - - vPC domain 20 - - R R R R • Key Recommendations • vPC Domain id for facing vPC layers should be different • No Bridge Assurance on interconnecting vPCs • BPDU Filter on the edge devices to avoid BPDU propagation • No L3 peering between DCs (i.e. L3 over vPC) ACCESS ACCESS - - E E B B Server Cluster Server Cluster

Data Center InterconnectEncrypted Interconnect DC-2 DC-1 Nexus 7010 Nexus 7010 vPC vPC CTS Manual Mode (802.1AE 10GE line-rate encryption) No ACS is required Nexus 7010 Nexus 7010

HSRP with vPCFHRP Active/Active • Support for all FHRP protocols in Active/Active mode with vPC • No additional configuration required • Standby device communicates with vPC manager produces to determine if vPC peer is “Active” HSRP/VRRP peer • General HSRP best practices still applies. • When running active/active aggressive timers can be relaxed (i.e. 2-router vPC case) HSRP/VRRP “Active”: Active for shared L3 MAC HSRP/VRRP “Standby”: Active for shared L3 MAC L3 L2

HSRP with vPCDo NOT use Object Tracking Cautions: • Not recommended using HSRP link tracking in a vPC configuration • Reason: vPC will not forward a packet back on a vPC once it has crossed the peer-link, except in the case of a remote member port failure L3 CORE ACTIVE HSRP STANDBY HSRP GW L2/L3 Aggregation GW GW VLAN 100, 200 VLAN 100 VLAN 200

HSRP with vPCL3 Backup Routing • Use an OSPF point-to-point adjacency (or equivalent L3 protocol) between the vPC peers to establish a L3 backup path to the Core through in case of uplinks failure • A single point-to-point VLAN/SVI will suffice to establish a L3 neighborship. OSPF OSPF VLAN 99 L3 OSPF L2 Primary vPC Secondary vPC

HSRP with vPCDual L2/L3 Pod Interconnect Scenario: • Provide L2/L3 interconnect between L2 Pods, or between L2 attached Datacenters (i.e. sharing the same HSRP group). • A vPC domain without an active HSRP instance in a group would not able to forward traffic. Multi-layer vPC with single HSRP: • L3 on the N7K supports Active/Active on one pair, and still allows normal HSRP behavior on other pair (all in one HSRP group) • L3 traffic will run across Intra-pod link for non Active/Active L3 pair Active Standby Listen Listen

vPC and ServicesCatalyst 6500 Services Chassis w. Services VDC Sandwich • Two Nexus 7000 Virtual Device Contexts used to “sandwich” services between virtual switching layers • Layer-2 switching in Services Chassis with transparent services • Services Chassis provides Etherchannel capabilities for interaction with vPC • vPC running in both VDC pairs to provide Etherchannel for both inside and outside interfaces to Services Chassis • Design considerations: • Access switches requiring services are connected to sub-aggregation VDC • Access switches not requiring services may be connected to aggregation VDC • May be extended to support multiple virtualized service contexts by using multiple VRF instances in the sub-aggregation VDC • Design Cautions: • Be aware of the Layer 3 over vPC design caveat. If Peering at Layer 3 is required across the two vPC layers an alternative solution should be explored (i.e. using STP rather than vPC to attach service chassis)

vPC Latest EnhancementsSummary Several enhancements to vPC: • vPC Object Tracking • vPC Peer-Gateway • vPC Delay Restore • Multi-layer vPC with single HSRP group • vPC unicast ARP handling • vPC Exclude Interface-VLAN • vPC single attached device Listing • vPC Convergence and Scalability For more details: • 4.2 Release Noteshttp://www.cisco.com/en/US/docs/switches/datacenter/sw/4_2/nx-os/release/notes/42_nx-os_release_note.html#wp218085

vPC Latest EnhancementsvPC Peer-Gateway for NAS interoperability Local Routing for peer router –mac Traffic Scenario: • Interoperability with non RFC compliant features of some NAS devices (i.e. NETAPP Fast-Path or EMC IP-Reflect) • NAS device may reply to traffic using the MAC address of the sender device rather than the HSRP gateway. • Packet reaching vPC for the non local Router MAC address are sent across the peer-link and can be dropped if the final destination is behind another vPC. vPC Peer-Gateway Solution: • Allows a vPC switch to act as the active gateway for packets addressed to the peer router MAC (CLI command added in the vPC global config) vPC PL L3 L2 vPC PKL N7k(config-vpc-domain)# peer-gateway

In-Service Software Upgrade (ISSU)vPC System Upgrade/Downgrade 4.1(3) 4.2(1) 4.1(3) 4.2(1) • ISSU is still the recommended system upgrade in a multi-device vPC environment • vPC system can be independently upgraded with no disruption to traffic. • Upgrade is serialized and must be run one at the time (i.e. config lock will prevent synchronous upgrades) • Configuration is locked on “other” vPC peer during ISSU. 4.1(3) 4.2(1)

Agenda • Feature Overview & Terminology • vPC Design Guidance & Best Practices • Building a vPC domain • Attaching to a vPC domain • Layer 3 and vPC • Spanning Tree Recommendations • Data Center Interconnect (& Encryption) • HSRP with vPC • vPC latest enhancements • ISSU • Convergence and Scalability • vPC Hands-on Lab Information • Reference Material

4.2(1) vPC EnhancementsConvergence Topology 20 flows @1000 pps OSPF L3 Core Nexus 7000 E2/14 L2/L3 Aggregation Nexus 7000 vPC E2/14 N7K-1 N7K-2 OSPF Po10 16-way port-channel 4-way port-channel Po20 Po160 L2 Access Nexus 5000 vPC Peer Link LACP Channel (2x10 GigE) vPC Peer-Keepalive (GigE) 20 flows @1000 pps 20 flows @1000 pps

vPC on Nexus 7000Convergence Numbers P S P S P S NOTE: Convergence numbers may vary depending on the specific configuration (i.e. scaled number of VLANs/SVIs or HSRP groups) and traffic patterns (i.e. L2 vs L3 flows).

vPC on Nexus 7000Scalability Number Improvements NOTE: Supported numbers of VLANs/vPCs are NOT related to an hardware or software limit but reflect what has been currently validated by our QA. The N7k BU is planning to continuously increase these numbers as soon as new data-points become available.

vPC Hands-on Lab InformationOn Demand vPC Lab Overview N7K-1POD 1-2 VPC N7K-2POD 1-2 VPC Pod 2 Pod 1 N7K-3POD 3-4 VPC N7K-4POD 3-4 VPC Pod 4 Pod 3 • Instructor-led hands-on lab introducing the vPC (virtual Port-channel) feature for the Nexus 7000. • Participants exposed to the configuration of vPC with NX-OS. • Lab needs to be manually booked through Nexus 7000 TMEs. N7K-7POD 5-6 VPC N7K-8POD 5-6 VPC Pod 6 N7K-Aggr Pod 5 N7K-Aggr Pod 2 Pod 1

Nexus 7000 virtual Port-Channel Best Practices & Design Guidelines