VMware VVF (VMware vSphere Foundation) interview questions

One place to manage many hosts and VMs, plus the cluster features that need it: vMotion, DRS, HA, templates, roles and permissions, distributed switches and lifecycle management with vLCM. Without vCenter each host is managed on its own through the host client.

What they are checking: That you see vCenter as the control plane for cluster features, not just a console.

Likely follow-up: What happens to running VMs if vCenter goes down?

HA restarts VMs on other hosts after a host failure, so there is a short outage. DRS balances load by moving running VMs with vMotion and places new VMs. Fault Tolerance keeps a live secondary copy of a VM on another host so a host failure causes no downtime, but it costs double the resources and has vCPU limits.

What they are checking: Clear separation of restart, balance and zero-downtime protection.

Likely follow-up: When would you choose FT over HA for a VM?

The vMotion VMkernel adapters on both hosts (enabled for vMotion, same subnet or a routed vMotion stack, MTU matching end to end), that both hosts see the same datastore and port group, CPU compatibility (EVC mode), and attached devices such as a mounted ISO on local storage. vmkping between the vMotion IPs with the right packet size quickly finds MTU problems.

What they are checking: A network-first, ordered approach and knowing vmkping.

Likely follow-up: What is EVC and why does it matter for mixed CPU generations?

Thin disks save space by allocating on write, so you can overcommit the datastore; the risk is running out of space when VMs grow, which pauses VMs. Thick eager-zeroed gives the best first-write performance and is needed for FT. I use thin with datastore usage alarms and regular capacity reviews.

What they are checking: You know overcommit risk and how to monitor it.

Likely follow-up: What happens to a VM when its datastore fills up?

A standard switch is configured on each host separately, so settings can drift. A distributed switch is defined once in vCenter and pushed to every host, with extra features such as NetFlow, port mirroring, Network I/O Control, LACP and private VLANs. It needs vCenter and the right license.

What they are checking: Consistency at scale and awareness of extra features.

Likely follow-up: How do you migrate VMs from a standard to a distributed switch without downtime?

Do not reboot the host first. Check whether the host management network is reachable (ping, DCUI), then restart the management agents (hostd and vpxa) from the DCUI or SSH. If HA is set up, check it has not isolated the host. Only if agents cannot recover, migrate or shut VMs down gracefully before rebooting.

What they are checking: Protecting running workloads and knowing the management agents.

Likely follow-up: What is host isolation response in HA?

Use vSphere Lifecycle Manager with a cluster image: check compliance, then remediate the cluster. vLCM puts each host into maintenance mode in turn, DRS moves its VMs away, patches and reboots it, then moves on. Check hardware compatibility and back up vCenter before you start.

What they are checking: Maintenance mode, DRS and a safe order of work.

Likely follow-up: Why should you upgrade vCenter before the hosts?

No. A snapshot is a delta file that depends on the base disk on the same datastore; if that storage fails you lose both. Snapshots also grow and slow the VM. Use them briefly before a change and delete them, and use a proper backup tool that copies data elsewhere.

What they are checking: Not confusing snapshots with backups.

Likely follow-up: How do you find VMs with old snapshots?

CPU ready time (VM waiting for a physical CPU, a sign of overcommit), memory ballooning and swapping, disk latency per datastore and network drops. esxtop or the vCenter performance charts show these. Often the fix is right-sizing: too many vCPUs can make a VM slower.

What they are checking: Knowing virtualization-specific metrics, not just guest CPU usage.

Likely follow-up: Why can giving a VM more vCPUs make it slower?

Pick a real example and use the STAR shape: what the change was, what broke, how you restored service, and what you changed afterwards (a change window, a rollback plan, a snapshot before the change, a peer review). Interviewers want honesty and a better process, not a perfect record.

What they are checking: Ownership and learning from mistakes.

Likely follow-up: How do you plan a rollback before a change?