I85PREP
⌘ K
Dell VxRail Operate study materials
Dell VxRail Operate study resources

Dell VxRail Operate: VxRail Networking and vSAN Operations

Dell VxRail Operate · Work through your exam, one topic at a time.

All guide sections

VxRail Networking and vSAN Operations

Networking and vSAN are where design errors surface later as performance or availability incidents. Expect questions on physical switching, vmkernel layout, jumbo frames, storage policies, and maintenance-mode behaviour.

Network rules

  • VxRail clusters use top-of-rack switch pairs with LAG (Dell VLT, Cisco vPC) so a single uplink or switch loss does not isolate a host.
  • Separate vmkernels for management, vSAN, vMotion and VM traffic, each on its own VLAN; NIOC shares decide who wins under contention.
  • Jumbo frames (MTU 9000) must be consistent end to end: host vmkernel, distributed or standard switch, uplinks, ToR ports, and any routed path used by stretched or witness traffic.
  • vSAN has used unicast since 6.6 - multicast is not required, and legacy multicast guidance is a distractor.

vSAN behaviour

  • Original storage architecture uses disk groups: one cache device per group plus one or more capacity devices. Cache is the write buffer, so size it for sustained writes and resync.
  • Storage policies set FTT and failure tolerance method: mirroring needs 2n+1 hosts, RAID-5 needs 4, RAID-6 needs 6.
  • Stretched clusters place components per site with a witness holding metadata; witness traffic can share the vSAN vmkernel or use a dedicated one.
  • Maintenance mode offers ensure accessibility (fast, leaves the cluster exposed) and full data migration (slow, restores policy compliance).

Worked example

Six hosts, 10 TB usable required, FTT=2. Mirroring costs 3x, so 10 TB usable needs 30 TB raw and each copy is spread thin; RAID-6 costs 1.5x, so 15 TB raw suffices and still survives two failures. Choose RAID-6 and confirm all six hosts contribute capacity. If one host is then placed in maintenance mode with ensure accessibility, compliance drops immediately and the cluster is temporarily at FTT=1.

Traps that cost marks

  • Setting jumbo frames on hosts but not on the ToR switch ports.
  • Designing vSAN on 1 GbE, or sharing a vmkernel with management and no NIOC shares.
  • Assuming RAID-5 delivers FTT=2 on a four-host cluster - it does not.
  • Using ensure accessibility for a planned disk replacement.
  • Forgetting that witness traffic needs a reachable path in stretched designs.