Fabricwright Design turns a drawing of a network into the configuration for every device in it. You choose the kind of fabric, place the devices, cable them, and describe the services the network carries. Fabricwright works out the rest: interface names, link descriptions, point-to-point addresses, loopbacks, routing protocols, the overlay and management, from templates that have been checked against real platforms.

This guide walks through a design from start to finish. To follow along, open Design and sign in with Google or GitHub.

Before anything goes on a device. Fabricwright generates configuration. It doesn't push it anywhere. Read what it produces, make sure you understand it, and try it in a lab or a maintenance window the way you would any other change.

The screen in one minute

  • The toolbar along the top holds the fabric type and the buttons for Services CSV, Export topology, Clear, Settings, Link mode and Generate configs.
  • The palette on the left holds the platforms. Drag one onto the canvas to add a device.
  • The canvas in the middle is the drawing. Orange links are Layer 3, blue links are Layer 2.
  • The side panel on the right shows whatever is selected: a device's role, addresses and services, or a link's two ends.
  • The project bar, next to the logo, holds your projects: New, Rename, Delete, Export and Import, and a badge that says whether the latest change is saved.

Step 1: Choose the fabric type

The fabric type decides almost everything else: which roles exist, which platform can hold which role, which services can be added, and which settings apply. So choose it first, in the Fabric type menu at the left of the toolbar. The roles it offers are listed beside it.

Fabric What it builds Roles Services
VXLAN An OSPF underlay, a BGP EVPN overlay, anycast gateways, optional TRM multicast Spine, Leaf VLAN, VRF
MPLS-LDP OSPF or IS-IS with LDP, MP-BGP for L3VPN, L2VPN, optional MVPN P, PE VRF, E-Line, E-LAN
SR-MPLS OSPF or IS-IS with segment routing, MP-BGP for L3VPN, L2VPN, optional TI-LFA and MVPN P, PE VRF, E-Line, E-LAN
VRF-LITE A routed core that carries VRFs hop by hop, without MPLS Core VLAN, VRF
PURE-L2 Switching only: trunks, spanning tree, port-channels, SVI gateways Core VLAN

Every fabric also has an Access role, for a switch that hangs off the fabric rather than routing on it. An access switch carries VLANs and nothing else.

Some rules of thumb:

  • A data centre leaf-and-spine on Nexus 9000 or Catalyst 9000 switches is VXLAN.
  • A service provider or large WAN core on ASR 9000, NCS 5000 or Catalyst 8000 routers is MPLS-LDP or SR-MPLS. Choose SR-MPLS for segment routing and TI-LFA fast reroute. A Catalyst 9000 can't be a P or PE on SR-MPLS, only an access switch.
  • A campus or enterprise core that keeps a few VRFs apart without MPLS is VRF-LITE.
  • A plain switched network is PURE-L2.

You can change the fabric type later. Every device is checked again, and one whose role doesn't exist on the new fabric moves to its platform's default role there.

Step 2: Place the devices

Drag a platform from the palette onto the canvas:

Platform Software What it is
Catalyst 9k IOS-XE Switch
Catalyst 8k IOS-XE Router
ASR 9k IOS-XR Router
NCS 5k IOS-XR Router
Nexus 9k NX-OS Switch
IOU L2 IOS-XE Virtual switch, for labs
CSR1000v / C8000v IOS-XE Virtual router, for labs
Firewall none Drawn only, never configured

If a platform can't hold any role on the current fabric, it isn't added, and the status bar at the bottom of the screen says why.

A new device is named after its platform, such as cat9k-1, and given its platform's default role on this fabric. Click it to open it in the side panel:

  • Hostname. Change it to your own naming scheme. It appears in the description of every link that points at this device.
  • Role. Only the roles this platform can hold on this fabric are listed, so a Nexus is never offered as a P router and a router is never offered as an access switch.
  • Chassis, on platforms that come in more than one. It decides how the ports are numbered, so set it before you draw links.
  • INB Management IP and Router-ID. Leave them blank and they're allocated from the pools in the project settings, or type one to pin it.
  • Redundancy role, for a device that's one half of a pair. It decides which side gets the higher HSRP priority, which is the ICCP node, and which is the vPC primary. Auto makes the lower hostname of the pair the primary.
  • Out-of-band management, which has a section of its own below.

A Nexus also has Enable VPC. Tick it, choose the peer, and set the vPC domain, the peer-link ports and the peer-keepalive port. The peer is updated to match. An ASR 9000 or NCS 5000 PE on a fabric with VPLS signalling has Enable ICCP for multihoming, which works the same way.

The Firewall is there so the drawing matches the real network. It holds no role, takes no addresses and gets no configuration. It does occupy the ports it's cabled to, and the switch facing it gets a port described as going to it.

Step 3: Cable it

Click Link mode, then click two devices. The link dialog opens:

  • Link type. Routed is a point-to-point link in the IGP, addressed from the interconnect pool. Trunk is tagged Layer 2, and Access is one untagged VLAN. Only the types both ends can carry are offered: an access switch has no routed links, and a PURE-L2 fabric has nothing routed at all.
  • Module, speed and port for each end. The lists come from the platform's own port layout, so you can only choose a port that exists, and a port another link already uses isn't offered.
  • Bundling, on Layer 2 links: whether these ports form one bundle. How the two ends agree isn't asked, because it follows from the devices: a vPC from the Nexus pair's own vPC settings, an ethernet segment where two leaves reach the same device, and a plain port-channel otherwise.
  • BFD, on routed links. It's on by default, and the timers can be left blank to use the usual values.

Every interface gets a description naming the device and port at the other end, in the form To_<hostname>_<interface>.

To change a link, click it on the canvas and choose Edit link. Click Link mode again when you've finished cabling.

Step 4: Project settings

Settings in the toolbar holds everything that applies to the whole project, and it's applied to every device when the configuration is generated. What you see depends on the fabric type. A PURE-L2 fabric, for example, has no addressing pools and no routing protocols to set.

Addressing pools

  • Interconnect (P2P) pool and prefix length: where routed links get their addresses, as /31s or /30s.
  • Router-ID pool: the Loopback0 address on every routed device.
  • On VXLAN, the VTEP loopback pool, and the anycast VTEP address that a vPC pair shares.

Management

  • Mgmt pool. On a routed fabric, this addresses the in-band management loopback on every routed device. On PURE-L2, it addresses the management SVI on every switch.
  • Mgmt VRF. Management in the global table, or in a VRF of its own. On VXLAN it's always a VRF.
  • Mgmt gateway. Leave it blank to use the last usable address in the pool.
  • Management VLAN. On PURE-L2 it's a field here. On the other fabrics it's a service: add the VLAN (an E-LAN on MPLS fabrics) and tick Management VLAN on it.
  • Out-of-band management. See the next section.
  • Monitoring / control networks: the source networks allowed to reach the devices' VTY lines.
  • NTP servers, syslog servers, and the login and exec banners.

Fabric settings

  • Underlay IGP: OSPF or IS-IS. A VXLAN underlay runs OSPF.
  • BGP ASN.
  • IPv6 support. Prepares the fabric to carry IPv6 unicast. It doesn't address anything by itself: each service takes an optional IPv6 address of its own.
  • Fast-reroute (TI-LFA), on SR-MPLS.
  • L2VPN services, on the MPLS fabrics, with the signalling: VPLS, EVPN, or both.
  • Multicast. On the MPLS fabrics that's MVPN, with a choice of profile 13, 14, 21 or 25; 14 is the recommended one. On VXLAN it's Tenant Routed Multicast, which needs an underlay RP address, and each VRF then opts in on its own. On VRF-LITE it's PIM with an anycast RP on the core.
  • Protocol passwords for BGP, OSPF, IS-IS and LDP.

AAA

  • TACACS+ servers and their key.
  • Fallback account and password, the local login used when no TACACS+ server answers, and the enable secret.
  • AES password encryption, with a master key, for type 6 passwords instead of type 7.
  • 802.1X with MAB fallback, with its RADIUS servers. It's for Catalyst 9000 only, and ports are enabled per VLAN service.

About the secrets in a project. Passwords and keys typed here are saved with the project, so the configuration can be generated again later. As the terms say, they're not separately encrypted. If you'd rather not store real ones, use placeholders and replace them in the generated files.

Out-of-band management

Out-of-band (OOB) management is a second, independent way into every device, through the platform's own dedicated management port. That port carries no fabric links, so it stays reachable when the fabric doesn't, which is exactly when you need it.

It's off by default, because it only exists where a management network is actually cabled. To turn it on:

  1. Open Settings, and in the Management box tick Out-of-band management.
  2. Set the OOB pool, the subnet the management ports sit in. The default is 192.168.255.0/24.
  3. Set the OOB gateway, or leave it blank to use the last usable address in the pool.
  4. Save.

Each device then gets an address from the pool on its management port, in its management VRF, with a default route out of that VRF to the gateway. Which port and which VRF depends on the platform:

Platform Management port VRF
Catalyst 9k GigabitEthernet0/0 Mgmt-vrf
Catalyst 8k GigabitEthernet0 Mgmt-intf
Nexus 9k mgmt0 management
ASR 9k, NCS 5k MgmtEth0/RSP0/CPU0/0 or MgmtEth0/RP0/CPU0/0 OOB
IOU L2, CSR1000v / C8000v none none

On Catalyst and Nexus the VRF comes with the software and the port is already in it, so only the address and the default route are written. IOS-XR has no built-in management VRF, so Fabricwright creates one called OOB, with its own static default route and the SSH server listening in it. The virtual platforms have no management port, so they get no OOB configuration and take no address from the pool. Reach them over the console or over the fabric.

In the side panel, each device shows the port and VRF that will be configured, and lets you:

  • type an OOB Management IP instead of taking one from the pool. A typed address is never handed to another device.
  • type an OOB Gateway for this device only, if it sits in a different management subnet.
  • on IOS-XR, choose the route processor type, which decides whether the port is named RSP or RP, and tick Dual route processors for a chassis with two. That takes three addresses: one for each route processor, and a virtual address that follows whichever one is active.

When you generate, look out for two warnings. If the pool is too small for every device, you're told how many addresses are missing, and the devices it doesn't reach get no OOB configuration. And if the OOB pool overlaps the in-band management pool, you're told that too. The two are meant to fail independently, so they shouldn't share a subnet.

Step 5: Add services

Services are what the network carries: VLANs, VRFs, E-Lines and E-LANs. Select a device, click + Add service in the side panel, choose the type, and fill in its fields.

What's offered depends on the fabric, the device's role and its platform:

  • An access switch carries VLANs only.
  • A router carries VRFs, and on the MPLS fabrics E-Lines and E-LANs, but no VLANs, because it has no VLAN database.
  • A Catalyst 9000 PE on an MPLS fabric carries VRFs and VLANs. On MPLS-LDP, a VLAN can be stretched over the core.

A service is marked Draft until its required fields are filled in, and Ready once they are. The list under the device shows what's still missing.

For a long list, Services CSV in the toolbar downloads a blank template or your current services, and imports a file back in. If the project already has services, you choose whether the file replaces them or is merged in, skipping duplicates.

Step 6: Generate, review, export

Click Generate configs. Every device's configuration appears in a tab of its own, with any warnings underneath: a pool that has run out, a service that's missing something, two links on the same port. Download all (.zip) saves every file at once.

Then review it. Read each device's configuration, check that it says what you meant, and try it in a lab first if you can.

A few more things worth knowing:

  • Export topology saves the drawing, and an inventory of every device and link, as a printable PDF for a design document or a change request.
  • Export in the project bar downloads the whole project as a file, and Import opens one as a new project. An exported project is yours to keep.
  • Clear empties the canvas and puts the settings back to their defaults. It asks first.
  • Projects save themselves as you work, and the badge in the project bar says when the latest change is saved. During the launch, every account can keep three projects.

Where to go next

Questions or corrections: [email protected]. More about Design.