In this video you will get an overview of the tasks you must perform to prepare your data center networks for automatically deploying VxRail. A VxRail cluster has a dependency on the data center network for workload management and to support continuous operations.
This graphic illustrates the major tasks that should be completed before starting a VxRail Deployment. This starts with the VxRail team coordinating with the network team to jointly develop a networking architecture, and planned after-business and workload requirements are set. The resulting preparation work on the supporting network infrastructure must be completed before starting the VxRail Automated Deployment process.
A VxRail standard cluster depends on one or more Ethernet switches to enable connectivity between the VxRail nodes and enable connectivity upstream. The adjacent Ethernet switches, referred to as Top-of-Rack switches, must be configured before connecting to VxRail nodes. Connecting the iDRAC ports on your VxRail nodes to your network, for hardware management purposes, is recommended but not required.
Before trying to deploy, ensure that every VxRail node is connected to the top-of-rack switches with either two NICs or four NICs. Dual Ethernet switches are strongly recommended for resiliency purposes. You can choose to only use the built-in Ethernet adaptor ports. The other option is to use both the built-in ports and ports on optional PCIe adapter cards.
Whichever option you choose, each switch port must be configured to support VxRail networking as outlined in the joint planning session. VxRail depends on five distinct networks to support cluster operations, as shown in the colored lines. Each VxRail network is assigned to a VLAN. The VLAN's must be able to pass through every switch port that is connected to a VxRail
node port.
Each VxRail node connects to the external management network. The VxRail Manager and the VxRail supplied vCenter on the primary node also connect to this network. Every VxRail node and every management component requires an IP address that can route to data center services. And, every VxRail node and management component requires a host name. Reserve a subnet for this purpose.
The Internal Management network, which is used for internode communications and discovery, requires Multicast to be set on this VLAN. VxRail Manager discovers the powered-on nodes on the network using Multicast in order to form the cluster. You can opt to manually configure the network settings on the nodes if you do not want to enable Multicast on this network. The vSAN network supports a cluster-wide datastore used by the virtual machines.
Unicast must be set on this VLAN. Every VxRail node requires an IP address. If the expectation is the cluster will expand across a single rack, then reserve a public subnet instead of a private subnet for this purpose. The vMotion network enables virtual machine migration between the nodes on the VxRail cluster. Like the vSAN network, every VxRail node requires an IP address. If virtual support migration outside the cluster is required, then reserve a public IP subnet.
Otherwise, a private IP subnet will suffice. The guest networks are where you run applications and workload on the cluster. You can create as many guest networks as needed for your requirements, so long as a VLAN is reserved for each instance. The Guest networks must be configured to pass upstream to reach end users and external applications. After you've recorded the VxRail settings needed for deployment, you'll need to configure a VLAN for each VxRail network on the top-of-the-rack switches.
Then, for each switch port that connects to a VxRail node, configure the port as a truncated port, to pass through all of these VLANs. Plan on configuring inner switch links, such as Virtual Link Trunks, between the switches, to allow for VLAN passage. And, lastly, configure the uplinks to allow passage of the VxRail networks requiring connectivity upstream. Depending on your layer two and layer three boundary, these can be port channels or point to point links.
You will also need to decide on the transmission unit size used for network communication. This unit size can range from 1500 bytes to 9000 bytes. The packet size is set during VxRail automated deployment at both the virtual switch level and at the port group level. Using the larger supported transmission size with the vSAN network, can improve throughput on clusters with a heavy workload.
Ensure the transmission unit size set on the top-of-rack switch is large enough to support the largest transmission unit size set in the virtual network. To prevent the VxRail nodes from being a member of a bridging loop in your network, the VxRail nodes should be configured as Edge of devices for spanning tree purposes. This will prevent an edge of port from participating in topological changes when a link toggles.
If your plain network topology VxRail includes redundant layer two paths, then ensure spanning tree is configured in your network to prevent loops in the data center network. Your routing services need to be configured to enable connectivity for VxRail networks requiring connectivity outside of the cluster. For each guest network you create, make sure the guest virtual machines paying for the network can reach upstream applications and end users.
The VxRail Manager must be able to reach certain data center services. If you plan to use VxRail Call Home features with a secure remote services gateway in your data center, make sure the External Management network can route to this device. If you plan to synchronize the time on your VxRail nodes using NTP, be sure the External Management network can route to this service.
If you plan to deploy the VxRail cluster on a vCenter that already exists in your data center, the External Management network must be able to reach this instance. If you plan to deploy the VxRail cluster using a domain named server in your data center, then it must be reachable on the External Management network. Make sure this DNS instance has forward and reverse look-up records for every VxRail Management component on the cluster.
This must be done before automated deployment. If you want remote access to the VxRail nodes Management interface, a separate one gigabit network is recommended for this purpose. Depending on your planned deployment, you may need to open ports on your firewall between VxRail Manager and certain data center services.
Refer to the VxRail Network Planning Guide. Or the pre-deployment checklist on the configuration portal for information on firewall rules for VxRail. After this, the other recommended tasks are listed here. Take a minute to pause the video and read through these tasks.
In this video, we provided an overview of the tasks you must perform to plan your network before deploying VxRail. A summary of the tasks that need to be worked in tandem with the VxRail planning team and data center administration team was also provided. For more information about networking requirements for VxRail, see the VxRail Network Planning Guide available on: Dell.com/Support.
