Dell SmartFabric Services User Guide Release 10.5.5

Recommendations

See Internal components and networks for more information about the SFS components and networks that are created by SFS.

Recommendations and notes regarding the SFS components are:

• Dell Technologies recommends that you do not change or disable the STP settings on switches in SmartFabric mode. Any change to the STP settings results in reboot of all the switches in the fabric and disabling STP settings cause loops.
• Dell Technologies recommends using uplinks from a leaf pair as a best practice. Uplinks from leaf switches to external switches can be L2 or L3. All uplinks from spine switches to external switches must be L3.
• Except for PowerScale (Isilon) back-end deployment, breakout feature is not supported on the leaf switches for the VLTi or Leaf-Spine link connections. Use the default speed cables of the leaf switch when creating the Leaf-Spine link or VLTi connections in a L3 fabric.
  • Leaf-Spine link example—If the default speed on the leaf switch port is 100G and spine switch port is 400G, use 100G cable for Leaf-Spine link connection. Since autobreakout feature is enabled on the spine switch, the port speed of the spine switch is set to 100G automatically.
  • VLTi example—If the default speed of the leaf switches is 100G, use 100G cable for the VLTi connection.
• Ensure that the Designated Leader configuration is set for the leaf or spine switches in the existing fabric before expanding the fabric to prevent the configuration loss. For more information about how to set the designated leader configuration, see Designated Leader. Once a leader is elected, it initiates all applications to automatically build the network fabric, and leader virtual IP address is advertised for applications to automatically discover the fabric through inband networks.
• The autobreakout feature is not supported on leaf switches. If you want to breakout the ports to configure uplinks, server connections, or jump hosts, use the SFS UI.
• Ensure that the default networks created by SFS do not conflict with any networks in the existing deployment. If so, change the default networks using the instructions that are provided in Edit fabric settings.
• When you want to reboot any of the border leaf switch having L3 uplink with eBGP peer configuration that is connected to the external switch, Dell Technologies recommends shutting down the interface of the external switch that is the connected to that leaf switch before rebooting. After reboot and the BGP session is established within the fabric, bring up the interface of the external switch.
• Dell Technologies recommends not flapping the admin status of the server connected ports using the SFS UI, OMNI UI and REST API to avoid traffic loss on those ports after switch reboot.
• In SmartFabric mode, you can create VLAN using the interface vlan command through OS10 CLI, but you cannot delete the VLAN from the CLI. Therefore, Dell Technologies recommends using the SFS UI to create, edit, or delete a VLAN.

Default settings

• SFS creates VLANs from 4000 to 4094 for internal use. Do not use these VLANs for general use.
• SFS creates default networks for the leaf and spine network configuration for the domain ID.
• By default, autobreakout feature is enabled on the spine switches in SmartFabric mode. Autobreakout works only with DAC and AOC cables.
• In SFS, all network QoS priorities are set to Iron by default, and it cannot be changed.

SFS behavior

SFS elects one switch from the fabric as a leader switch and designates the remaining switches as the backup switches. In the event of a leader reload or failure, a new leader is elected from the backup switches using the keepalive information. The switches that are configured as Designated leader have a higher priority to become the leader switch. If none of the switches are configured as the designated leader, any switch can become the leader.

When you expand the fabric, the newly added switches may come up, form a fabric among themselves, and elect a leader before they are connected to the existing fabric. When the new fabric merges with the existing fabric, SFS elects a new leader switch for the combined fabric. If one of the new switches becomes the leader, it may overwrite the configuration in the existing fabric. To prevent this, ensure that the designated leader flag is set for the current leader in the existing fabric before expanding the fabric to prevent the configuration loss.

Preferred Leader behavior

Starting from OS10.5.5.0P1 release, the Preferred Leader flag is set automatically on all the leaf switches when you create an uplink using SFS UI. The preferred leader flag setting can be viewed in the CLI of each switch using the show smartfabric cluster command.

The provision to configure designated leaders on a fabric is available in UI. For more information, see Designated leader.

Spanning tree considerations

• To avoid loops, SFS does not allow you to configure the same network on multiple uplinks.
• SFS supports RPVST+ and MSTP. The default spanning tree mode in SFS is RPVST+. Once the fabric is created, you can change the STP mode using SFS UI, see Edit default settings.
  NOTE:When you change the mode using SFS UI, the whole fabric goes through a reboot, and the new mode is set to MSTP. The reboot action impacts the traffic flow in the cluster.
• By default, RPVST+ is enabled on the uplink interfaces for L3 VLAN networks. When RPVST+ is enabled on the uplink, the total number of Port VLANs (PV) supported is 400.

BFD support

From OS10.5.3.0, SFS provides an option enable or disable Bidirectional Forwarding Detection (BFD) for external BGP peer routes. You can enable or disable BFD for a eBGP peer route policy when configuring uplinks, server interface profiles, and routing profiles. By default, BFD is disabled.

If you want to edit the BFD configuration for a routing policy, you must delete the existing routing policy and create a policy with the respective BFD configuration.