Extreme Data Center (VOSS/VSP) cheat sheet

VOSS (Versatile Operating System) is the modular, Linux-based network OS that runs on VSP (Virtual Services Platform) 4000, 7000, 8000, and 9000 series switches; VSP 8000/9000 are the data-center fabric platforms.

Fabric Connect separates service from transport: services are identified by I-SIDs and transported over an SPB (IEEE 802.1aq) backbone, so VLANs are local edge constructs, not end-to-end transport.

The VSP control plane is split into primary and secondary CPU complexes; the secondary CPU runs the management/control-plane functions and can take over on primary failure.

Extreme data-center designs use a two-tier spine-leaf topology; every leaf connects to every spine, and SPB IS-IS runs on spine/leaf fabric links to build the shortest-path tree.

A VSP switch boots into one of two partition modes: a single 4-slot chassis or two independent 2-slot virtual switches, selectable via the boot flag, which changes port numbering and HA options.

Fabric Connect supports L2 VSN (virtual service network) for bridged services and L3 VSN for routed services; both are carried over the same IS-IS SPB backbone.

The VSP 9000 is a chassis-based core/aggregation platform; the VSP 8000 is a stackable/fixed data-center switch; both run VOSS and support Fabric Connect.

Key CLI modes are User EXEC (>), Privileged EXEC (#), Global Configuration (config), and Interface Configuration (config-if); 'enable' and 'configure terminal' move you up the hierarchy.

Out-of-band management uses the dedicated Management port with a default IP of 192.168.10.1/24 on factory-default VSP switches; in-band management uses a management VLAN or loopback.

Default CLI credentials on a factory-default VSP are user 'rwa' with no password (or 'admin' on some releases); change them immediately during initial configuration.

The 'ip address' command under 'interface mgmt' sets the OOB management address; 'ip route' or 'ip default-gateway' sets the management default route.

Save configuration with 'save config' or 'copy running-config startup-config'; unsaved changes are lost on reload. Use 'show running-config' and 'show startup-config' to compare.

Software upgrades use 'download' from a TFTP/FTP/SFTP server, then 'boot' from the new image; verify with 'show boot' and 'show software'.

SSH is enabled with 'ssh' and RSA/DSA key generation; Telnet should be disabled with 'no telnet' for production. SNMPv3 is preferred over v1/v2c.

NTP is configured with 'ntp server ' and verified with 'show ntp status'; accurate time is required for certificate validation and log correlation.

Syslog is configured with 'logging host ' and 'logging severity '; local buffering uses 'logging buffered'. Verify with 'show logging'.

Fabric Connect is enabled globally with 'fabric enable' and an IS-IS instance is created with 'router isis'; the SPB backbone is formed by 'spbm' configuration under the IS-IS instance.

Auto-Sense (Fabric Attach) lets an access port automatically detect a Fabric Attach client and assign it to a service without manual VLAN/I-SID provisioning; enabled with 'auto-sense enable' on the port.

Fabric Attach uses the FA client/server protocol (IEEE 802.1Qcj) over LLDP to signal VLAN/I-SID element requests from an access device to the FA server (the VSP).

Auto-configuration (Zero Touch) uses DHCP options 66/67 to point a booting switch to a configuration file server; the switch downloads and applies a config on first boot.

IS-IS on VSP uses NET (network entity title) addresses; the system ID is derived from the chassis MAC and must be unique per node in the fabric.

'show fabric' and 'show isis adjacencies' verify fabric membership; 'show auto-sense' shows discovered Fabric Attach clients and their assigned I-SIDs.

Fabric Connect requires consistent MTU (typically 9216 jumbo frames) on all fabric links; mismatched MTU prevents IS-IS adjacency formation.

Manual Fabric Attach elements can be pre-provisioned with 'fa element' commands so the FA server has service assignments before the client connects.

SPB (IEEE 802.1aq) uses IS-IS as its link-state protocol; each node computes shortest paths and installs unicast and multicast forwarding entries for all I-SIDs.

SPB uses a backbone VLAN (B-VLAN) for transport and service VLANs (I-SIDs) for services; B-VLANs are configured with 'spbm b-vlan '.

The B-VLAN must be consistent across the fabric; the primary B-VLAN carries unicast and multicast, and a secondary B-VLAN can be used for load sharing.

SPB multicast uses a shared tree rooted at the source; 'spbm multicast' settings control whether the fabric uses head-end replication or the SPB multicast tree.

'show isis spbm' displays the SPB topology, B-VLANs, and I-SID bindings; 'show isis spbm i-sid' shows per-I-SID forwarding state.

SPB equal-cost multipath (ECMP) is automatic across equal-cost spine paths; traffic is hashed across available paths.

A node's SPB nickname is derived from its system ID (default) or set manually with 'spbm nickname '; nicknames must be unique in the fabric.

'show isis interface' and 'show isis adjacencies' confirm adjacency up/down state; an adjacency stuck in Init usually indicates MTU, authentication, or B-VLAN mismatch.

A VLAN is a local Layer 2 broadcast domain on a switch; an I-SID is a fabric-wide service identifier (1-16777215) that maps VLANs across the SPB backbone.

Provision an L2 VSN with: create VLAN, enable 'i-sid ' under the VLAN, then bind the VLAN to the IS-IS SPBM instance with 'spbm i-sid vlan '.

The same I-SID must be configured on every edge switch that participates in the service; the local VLAN ID can differ per switch.

'show i-sid' lists all I-SIDs and their VLAN bindings; 'show vlan i-sid' shows the VLAN-to-I-SID mapping.

A VLAN can be associated with only one I-SID; an I-SID can be associated with multiple VLANs on the same switch only via specific configurations.

For untagged access ports, use 'untagged' on the port; for tagged trunks, use 'tagged' and ensure the VLAN is a member of the port.

Private VLANs and VLAN translation are supported on VSP; VLAN translation maps an ingress VLAN to a different internal VLAN before I-SID mapping.

Verify end-to-end service with 'show mac-address-table i-sid ' and 'show isis spbm i-sid '; a missing remote MAC indicates a service binding or IS-IS problem.

L3 VSN (virtual service network) provides routed services over the SPB backbone; each L3 VSN has its own VRF and IP subnet.

A VRF is created with 'ip vrf '; interfaces are assigned with 'ip vrf forwarding ' under the interface.

An L3 VSN is configured with 'i-sid ' plus 'vrf ' under the IS-IS SPBM instance; the I-SID carries the IP subnet across the fabric.

OSPF and BGP can run over L3 VSNs; OSPF is configured with 'router ospf' and 'network' statements, BGP with 'router bgp ' and 'neighbor' statements.

Static routes are configured with 'ip route '; verify with 'show ip route' and 'show ip route vrf '.

Inter-VRF routing requires route leaking or a shared service; VRFs are isolated by default.

DHCP relay is configured with 'ip dhcp-relay' and 'ip dhcp-relay fwd-path '; verify with 'show ip dhcp-relay'.

VRRP provides first-hop redundancy on VSP; configure 'ip vrrp address ' on the SVI and verify with 'show ip vrrp'.

VSP 8000 supports stacking (up to 8 units) with a single control plane; a stack failure triggers master election and reconvergence.

VSP 9000 uses dual control processors (CP) in active/standby; the standby CP takes over on active failure with sub-second to seconds of disruption.

SMLT (Split Multi-Link Trunking) is Extreme's MC-LAG technology: two aggregation switches form an IST (inter-switch trunk) and present a single logical LAG to access switches.

IST (inter-switch trunk) carries control traffic between SMLT peers; configure with 'interface ist' and 'ist' under the peer switches, and verify with 'show ist'.

RSMLT (Routed SMLT) provides redundant L3 gateways across SMLT peers; both peers answer ARP and forward traffic, eliminating suboptimal routing.

Fabric Connect reconverges on link failure via IS-IS SPF; convergence is typically sub-second for SPB topologies.

BFD (Bidirectional Forwarding Detection) accelerates failure detection for OSPF, BGP, and IS-IS; enable with 'bfd enable' and per-protocol 'bfd'.

'show ist', 'show smlt', and 'show vrrp' verify HA state; 'show log' and 'show tech' capture failure evidence.

'show tech' collects a full diagnostic bundle (config, logs, counters); 'show tech ' narrows the output. Attach it to support cases.

'show log' displays the local syslog buffer; 'show logging' shows configured destinations and severity. Increase verbosity with 'logging buffered '.

'show interface ' displays link state, errors, and counters; 'show interface statistics' shows detailed error counters (CRC, runts, giants).

'show mac-address-table' and 'show mac-address-table i-sid ' verify L2 forwarding; 'show ip arp' and 'show ip route' verify L3.

'show isis adjacencies' and 'show isis spbm' diagnose fabric problems; common causes are MTU mismatch, B-VLAN mismatch, and IS-IS authentication mismatch.

'show cpu' and 'show memory' identify resource exhaustion; high CPU often indicates control-plane churn or a broadcast storm.

'show auto-sense' and 'show fa' verify Fabric Attach operation; a client that does not appear may have LLDP disabled or an FA mismatch.

Maintenance tasks: 'save config', 'copy running-config startup-config', 'download' for upgrades, and 'boot' to activate a new image. Schedule upgrades in maintenance windows.