VCP-DCV cheat sheet

ESXi is a Type 1 hypervisor: the VMkernel runs directly on hardware and schedules CPU, memory, storage, and network resources for every VM.

The vCenter Server Appliance (VCSA) runs Photon OS and is deployed from an OVF/ISO installer; it is required for DRS, vDS, and HA cluster orchestration.

vCenter sizing by hosts/VMs: Tiny 10/100, Small 100/1,000, Medium 400/4,000, Large 1,000/10,000, X-Large 2,000/35,000.

vSphere 7 configuration maximums: 1,024 VMs per host, 96 hosts per cluster (64 with vSAN), 8,000 VMs per cluster.

vSphere 7 virtual machine maximums: 256 virtual CPUs and 6 TB of RAM per VM.

vSphere HA restarts VMs on surviving hosts after a host failure; admission control reserves capacity so those restarts can succeed.

DRS balances compute load across cluster hosts with vMotion; Storage DRS balances capacity and I/O across a datastore cluster.

vMotion migrates only the running compute state of a VM between hosts; Storage vMotion migrates only the virtual disk files between datastores.

A vSphere Distributed Switch is defined once in vCenter and inherited by member hosts; a standard switch is configured separately on each host.

vCenter Server Foundation manages a maximum of 4 ESXi hosts; vCenter Server Standard supports the appliance maximum of 2,000 hosts.

vSphere Standard includes vMotion and vSphere HA; DRS, Storage DRS, Distributed Switch, and host profiles require Enterprise Plus.

vSphere with Tanzu requires Enterprise Plus and creates a Supervisor cluster that runs Kubernetes workloads on vSphere infrastructure.

vSAN is licensed separately from vSphere and requires a cluster whose ESXi hosts contribute local disks to a shared datastore.

VMware Cloud Foundation bundles vSphere, vSAN, NSX, and Aria management into a single private-cloud stack.

VMware vSphere Foundation combines vSphere Enterprise Plus with vSAN and Aria operations and automation components.

Site Recovery Manager orchestrates failover between sites and requires vSphere Replication or array-based replication at both sites.

Aria Operations provides capacity and performance analytics; Aria Automation provides self-service provisioning and infrastructure as code.

NSX provides network virtualization and distributed firewalling, replacing many hardware-centric network segmentation designs.

Cluster design maximums: 96 hosts per cluster (64 with vSAN) and 8,000 VMs per cluster.

HA admission control options are cluster resource percentage (default 25%), slot policy, and dedicated failover hosts.

For N+1 redundancy across four equal hosts, the percentage policy must reserve at least 25% of cluster resources.

A standard vSAN cluster needs at least 3 hosts; a 2-node vSAN configuration requires a witness appliance at a third site.

vSAN design keeps roughly 25-30% of raw capacity free for rebuild slack, snapshots, and future growth.

VMFS6 supports datastores up to 64 TB and virtual disk files up to 62 TB with a 1 MB block size.

Plan at least 250 Mbps of bandwidth per concurrent vMotion operation; 10 GbE is recommended for large-memory VMs.

Separate management, vMotion, storage, and VM traffic onto distinct VLANs and dedicated VMkernel adapters.

The SSO domain name is fixed at deployment, and every host plus vCenter needs resolvable forward and reverse DNS records.

vCenter deployment has two stages: stage 1 deploys the appliance OVF, stage 2 configures SSO and networking and starts services.

The VCSA installer can run unattended using vcsa-deploy install with a JSON template.

The default SSO domain is vsphere.local and the initial administrator account is [email protected].

ESXi 7.0 and later need a boot device of at least 32 GB; SD/USB boot is deprecated in 7.0 and unsupported in 8.0.

The management VMkernel adapter (vmk0) is created during ESXi installation using the IP address entered in the DCUI.

SSH and the ESXi Shell services are stopped by default and must be started explicitly before remote command-line access.

A cluster can use vSphere Lifecycle Manager baselines or a single cluster image, but not both at the same time.

Host profiles capture a reference host's configuration; host-specific values such as IP addresses are supplied by an answer file.

VAMI on port 5480 configures VCSA backups, updates, certificate replacement, and service restarts.

esxtop interactive views: c for CPU, m for memory, n for network, d for disk, i for interrupts.

Sustained %RDY above roughly 10% indicates CPU contention; %MLMTD shows time a VM was held back by its own CPU limit.

%CSTP (co-stop) reflects SMP scheduling delay and rises when a VM has more vCPUs than available physical cores can co-schedule.

Memory reclamation escalates from transparent page sharing to ballooning (MCTLSZ), compression, then host-level swapping (SWCUR).

Storage I/O Control uses a congestion threshold (default 30 ms latency) to trigger share-based I/O throttling on a datastore.

The round-robin path policy (VMW_PSP_RR) is recommended for ALUA arrays; the fixed policy pins I/O to one preferred path.

VMXNET3 and PVSCSI paravirtual drivers reduce CPU overhead compared with e1000 and LSI Logic emulated adapters.

Upgrade order is vCenter first, then ESXi hosts; an older vCenter cannot manage a newer ESXi version.

The DRS migration threshold ranges from 1 (conservative) to 5 (aggressive) and controls how eagerly vMotion recommendations are applied.

ESXi logs live in /var/log and include vmkernel.log, hostd.log, vpxa.log, and fdm.log for vSphere HA agent activity.

vCenter logs are under /var/log/vmware/ on the appliance, with vpxd logs in /var/log/vmware/vpxd.

Collect a support bundle with vm-support on a host or from the vCenter UI to gather logs across the environment.

A purple diagnostic screen records a VMkernel panic; the core dump and stack trace identify the failing module.

APD means the datastore is temporarily unreachable and ESXi retries; PDL means the LUN is permanently gone and HA can restart VMs.

HA failover failures commonly trace to admission control settings, missing datastore heartbeats, or an unreachable isolation address (default gateway).

vMotion failures often come from EVC incompatibility, absent vMotion VMkernel adapters, or mismatched port group names on source and target.

AD/Kerberos logins fail when vCenter time differs from a domain controller by more than 5 minutes; verify NTP on the appliance.

Check vCenter service state with service-control --status --all and start everything with service-control --start --all.

Permissions propagate from parent inventory objects to children; an explicit child assignment replaces rather than merges with the parent's.

Roles are sets of privileges: system roles are read-only, sample roles are editable starting points, and custom roles are user-created.

Lockdown mode limits host access to vCenter; normal mode permits users with privileges, strict mode permits only exception users.

VAMI on port 5480 handles VCSA backup and restore, updates, certificate management, and service control.

A VCSA restore must use the same deployment size and the same FQDN as the backup being restored.

Host profiles check configuration compliance and remediate drift, using answer files for host-specific values.

Alarms trigger on events, conditions, or state changes and can send SNMP traps or email notifications to monitoring tools.

ESXi and vCenter both run in evaluation mode for 60 days before a license key must be assigned.

Maintenance mode with DRS evacuates VMs to other hosts via vMotion and blocks new VM placements on the host.