I85PREP
⌘ K
HCIP-Cloud Computing study materials
HCIP-Cloud Computing study resources

HCIP-Cloud Computing: Computing: FusionCompute Virtualization Essentials

HCIP-Cloud Computing · Work through your exam, one topic at a time.

All guide sections

Computing: FusionCompute Virtualization Essentials

The Computing domain of HCIP-Cloud Computing tests how Huawei FusionCompute turns bare-metal x86 servers into a manageable virtual resource pool. Questions target the CNA/VRM split, cluster-level features, VM lifecycle operations, and the exact preconditions that make live migration and HA succeed or fail.

Control plane and data plane

  • VRM (Virtual Resource Management) is the management plane. It is deployed as an active/standby pair (virtual or physical), owns clusters, hosts, datastores, port groups, templates and alarms, and carries no VM traffic.
  • CNA (Computing Node Agent) runs on every host, hosts the KVM-based VMs and reports state to the VRM. If the VRM pair is unavailable, running VMs continue to run, but new management operations stop.
  • A cluster groups hosts with compatible CPUs that share networks and normally shared datastores; cluster-scoped HA and DRS policies apply to all members.

VM resource rules

  • vCPU and memory hot-add require guest OS and tools support; devices such as passthrough GPU or USB pin the VM to one host.
  • Memory overcommit relies on ballooning and page sharing; it raises density but risks host memory pressure.
  • Thin-provisioned disks over-allocate capacity, so a datastore can fill even when provisioned space looks sufficient.

Worked example

VM app01 runs on host A with its disk on a local datastore, and host A must enter maintenance mode. A compute-only hot migration fails. Reasoning: live migration transfers memory state and leaves the disk in place, so the destination host must see the same datastore; a host-local disk is invisible elsewhere. Fix: run a storage live migration to a shared datastore first, then migrate the VM, then place host A in maintenance mode.

Traps that cost marks

  • Confusing VRM HA (management plane failover) with cluster HA (VM restart after host failure).
  • Assuming hot migration also moves disk data; it does not unless storage migration is selected.
  • Forgetting that HA restart needs surviving capacity and shared storage - local-disk VMs cannot fail over.
  • Treating DRS as redundancy; DRS balances load, HA provides restart.
  • Treating snapshots as backups; they live on the same datastore and grow a chain.