A greenfield build on converged or hyperconverged infrastructure goes well or badly long before the kit arrives. Most of the delays I have seen came from missing DNS records, VLANs that were not trunked, or a decision nobody had made. This is the checklist I work through for Cisco UCS Vblock (later renamed VxBlock) and Dell VxRail deployments.
It is not a replacement for the vendor planning guide. Use the current Dell guide for your version as the source of truth, and use this to make sure nothing is forgotten.
Know which platform you are building
The two platforms solve the same problem in different ways, and that changes the plan.
- Vblock is a converged system: Cisco UCS compute, a separate storage array and the network delivered as one engineered and supported unit, with a management platform (AMP) alongside it. Much of the build is done before delivery, so your work is integration with the customer network, storage layout and handover.
- VxRail is hyperconverged: Dell nodes running VMware vSAN, deployed through VxRail Manager. Storage lives in the hosts, so the network and the vSAN design matter more, and the first-run wizard does most of the build.
Before the kit arrives
- Rack, power and cooling. Confirm floor space, power feeds, circuit ratings and cooling with the data centre team, and check the delivery route.
- IP plan. Addresses for ESXi management, vCenter, the VxRail Manager or AMP components, vMotion, vSAN or storage networks, and out-of-band management (iDRAC or UCS Manager). Ping every address first to rule out duplicates.
- DNS. Forward and reverse records for every host, vCenter and the management appliances, checked with
nslookupin both directions. Missing reverse records are the most common reason a first run stops. - NTP. One reliable time source for everything. Hosts, vCenter and the management VMs must agree on the time before you start.
- VLANs. Agree the VLAN IDs for management, vMotion, vSAN or storage, virtual machine networks and any private management network. Trunk them on every switch port and on the links between switches.
- Switch settings. Configure every port that faces the platform the same way, set the MTU end to end if you use jumbo frames, and for VxRail make sure node discovery works. VxRail discovers nodes with IPv6 multicast, so the top-of-rack switches need it enabled on the discovery VLAN.
- Identity and accounts. Service accounts, the Active Directory or Entra ID integration for vCenter, and where the passwords will be kept.
- Decisions in writing. vCenter placement (embedded or existing), cluster names, vSAN storage policies, and who owns the platform after handover.
During the build
- Follow the vendor build guide step by step and record every value you enter.
- On VxRail, confirm VxRail Manager can see every node before you start the wizard, and keep the deployment logs.
- On Vblock, confirm the UCS service profiles, boot policy and storage zoning match the design before any workload lands.
- Turn on DRS and HA, check the vSAN or storage health checks are green, and test vMotion between every host.
Before handover
- Test failure. Pull a cable, put a host into maintenance mode and fail a path. Prove that the platform behaves as designed.
- Patch level. Bring firmware, ESXi and vCenter to the agreed baseline, using the vendor's tested bundle where one exists.
- Monitoring and backup. Add the platform to monitoring and alerting, and make sure the backup of vCenter and the management components works.
- Documentation. An as-built document with the IP plan, VLANs, accounts, versions and support contacts.
- Operations sign-off. Walk the operations team through the platform and hand over the runbooks before the first production workload moves.
The short version
Most greenfield problems are network and DNS problems. Get the IP plan, the DNS records, NTP and the VLAN trunks right and checked before delivery, follow the vendor guide exactly, and test failure before you hand over.