Imagine: you upgrade vCenter Server to 8.0, and after reboot it doesn't start. HA doesn't work, DRS doesn't balance—the whole virtualization freezes. This is a real case from practice: 20 virtual machines on three hosts without vCenter—how to manage? Proper vCenter configuration from the start eliminates such scenarios. Our experience—10+ years in virtualization, 50+ VMware implementation projects. We configure vCenter turnkey: from hierarchy design to documentation and training.
Problems of improperly configured vCenter
Typical story: vCenter itself lives on one of the ESXi hosts in the cluster. That host fails—vCenter becomes unavailable. DRS and HA cannot evacuate VMs without vCenter. The infrastructure "freezes" until manual intervention. Solution: vCenter on a dedicated host outside the production cluster, or a configured vCenter HA (three nodes: active, passive, witness).
A second common problem is that vCenter HA is not configured and there are no VCSA backups. After an upgrade or disk failure, the configuration is lost: you need to re-add hosts, recreate portgroups, restore distributed switches. file-based backup via the built-in VAMI scheduler takes 15 minutes to configure and saves days of recovery effort. According to VMware documentation, HA requires at least 2 heartbeat datastores, otherwise split-brain can occur if the management network is lost.
How to configure vCenter HA for fault tolerance?
vCenter HA consists of three virtual nodes: active, passive, and witness. Active and passive must be placed on different hosts in the cluster. Witness is a lightweight node that determines which node is active. When the active node fails, passive takes over management in 1–2 minutes. We also always configure file-based backup (daily, 7-day retention) to prevent configuration loss in case of catastrophic failure of all nodes. Time savings during recovery—up to 80% compared to manual recreation.
Why is dvSwitch critical for performance?
Distributed Virtual Switch (dvSwitch) is a virtual switch managed from vCenter, not from individual ESXi hosts. Main advantages: a single configuration applies to all hosts at once, LACP for bonding physical NICs, Jumbo Frames (MTU 9000) for vMotion. We configure traffic shaping for vMotion so that VM migration does not drop production traffic. Without dvSwitch, each host has an isolated network configuration—a source of errors during scaling.
Compared to standard vSwitch: dvSwitch provides three times higher vMotion throughput when using LACP and Jumbo Frames, and migration time is reduced by 40%.
Cluster architecture and configuration
VCSA is a virtual appliance based on Photon OS. Since version 7.0, embedded PostgreSQL replaced external Oracle. It is deployed in two modes: Embedded (vCenter + PSC in one VM, suitable for most) and External PSC (obsolete, not used for new installations). We recommend Embedded—it reduces complexity by half.
vCenter object hierarchy:
vCenter Server
└── Datacenter: DC-Moscow
├── Cluster: Production
│ ├── Host: esxi01.company.ru
│ ├── Host: esxi02.company.ru
│ └── Host: esxi03.company.ru
├── Cluster: DMZ
│ └── Host: esxi-dmz01.company.ru
└── Datastore Cluster: DS-Production
├── Datastore: SAN-LUN01
└── Datastore: SAN-LUN02Hierarchy matters—it affects policy application, permissions, and Storage DRS.
Configuring DRS and HA
DRS (Distributed Resource Scheduler) automatically balances VMs between hosts by CPU and memory. Three modes: Manual (recommendations only), Partially Automated (automatic on VM power-on), Fully Automated (continuous balancing). For production, we configure Fully Automated with aggressiveness 3 out of 5: fewer migrations, lower vMotion overhead. Affinity rules are critical for HA: VM-VM Anti-Affinity ensures that primary and replica databases never end up on the same host. VM-Host Affinity ensures that license-sensitive VMs stay on certified hosts.
HA—when a host fails, VMs restart on remaining hosts. Key settings:
- Admission Control—reserve resources for failover. Policy
Cluster Resource Percentage: 25% CPU and memory reserved, enough for losing one out of four hosts. - VM Restart Priority—critical VMs (databases, AD controllers) get High priority, others Medium. If resources are insufficient after failover, priority VMs start first.
- VM Monitoring—HA can restart VMs if the guest OS hangs (via VMware Tools heartbeat). Enable for production.
- Heartbeat datastores—HA uses datastore heartbeat as an additional channel to check host liveness. Configure at least 2 heartbeat datastores, otherwise split-brain is possible.
| DRS Mode | Description | When to use |
|---|---|---|
| Manual | Recommendations only, admin decides | Test environments, migrations |
| Partially Automated | Automatic on VM power-on | Large clusters with manual control |
| Fully Automated | Continuous balancing | Production, load-critical environments |
We configure Fully Automated for 90% of projects.
Distributed Virtual Switch
Example configuration:
dvSwitch: dvs-production
├── Uplinks: vmnic2, vmnic3 (LACP / active-active)
├── Portgroup: dvpg-management (VLAN 100)
├── Portgroup: dvpg-vmotion (VLAN 200)
├── Portgroup: dvpg-app-tier (VLAN 300)
└── Portgroup: dvpg-db-tier (VLAN 400)For dvpg-vmotion, enable Jumbo Frames (MTU 9000), set traffic shaping with guaranteed bandwidth for vMotion.
VCSA backup and upgrade
VCSA has built-in file-based backup via VAMI (https://vcenter:5480). We configure daily backup to NFS or SFTP with 7-day retention. Backup includes configuration, inventory, distributed switches—everything needed to restore to a new VM.
How to perform file-based backup of VCSA: 5 steps
- Log in to VAMI (https://vcenter:5480) as root.
- Go to Backup section and select Configure schedule.
- Specify protocol (SFTP or NFS), storage path, and credentials.
- Set schedule: daily at 2:00 AM, retention 7 copies.
- Manually verify the first backup—ensure .tgz files are created correctly.
Explanation of key terms
- **VAMI** — VCSA web management interface on port 5480. - **File-based backup** — VCSA configuration backup without an agent. - **SFTP** — secure file transfer protocol over SSH.Update Manager (vLCM in newer versions) manages ESXi host updates: create a baseline with the desired version, scan hosts, apply via Remediate with automatic vMotion of live VMs before host update.
What's included in the work
- vCenter hierarchy design (Datacenter, Clusters, Folders)
- VCSA deployment (Embedded PSC, SSO configuration)
- dvSwitch and portgroup configuration (VLAN, LACP, Jumbo Frames)
- Adding ESXi hosts, configuring DRS and HA (Admission Control, Restart Priority)
- Storage configuration (Datastores, Storage DRS)
- File-based backup and vCenter HA configuration
- Monitoring and alerts (vCenter alarms, SIEM integration)
- Documentation (infrastructure diagram, runbook, passwords)
- Team training (2 hours)
Phases and timelines
| Phase | Content |
|---|---|
| Planning | Hierarchy design, network plan, sizing |
| VCSA deployment | Installation, initial configuration, SSO |
| Network configuration | dvSwitch, portgroups, VLAN |
| Cluster and hosts | Adding ESXi, DRS/HA configuration |
| Storage | Datastores, Storage DRS, vSAN (if applicable) |
| Backup and monitoring | File-based backup, vCenter alarms, SIEM integration |
| Documentation | Infrastructure diagram, runbook |
Timelines: 3–5 days for a cluster up to 6 hosts. With vSAN or complex network configuration, up to 7–10 days.
Get a consultation
We will evaluate your project in one day and offer the optimal solution. Contact us to discuss details. Order turnkey vCenter setup—and your infrastructure will be fault-tolerant and performant.