Check every item below before running the installer. The installer performs an automated prerequisite scan on its first screen, but fixing prerequisites after the fact is much slower than getting them right up front.
The installer needs Administrator rights and makes substantial changes to the host, so read the list in Installer Overview before you run it.
Installer Host
The machine on which you run vamctl.exe.
Requirement | Minimum |
|---|---|
OS | Windows Server 2022 or Windows 11 (Pro / Enterprise) |
CPU | 4 cores |
RAM | 8 GB |
Disk | 50 GB free on the system drive for installer state and logs |
Privileges | An account with local Administrator rights |
Network | Outbound HTTPS (443) to the Vizrt registry for online installs |
Virtualization | Hardware virtualization enabled in BIOS/UEFI, required by both Hyper-V and WSL 2 |
WSL 2 | Available and set as the default version. The installer runs Podman on it. |
The installer host can be the same machine that runs VAM (for a single-node install on Hyper-V), or a separate management workstation (for VMware vSphere or remote Hyper-V deployments).
Podman and WSL 2
Every installation needs a working Podman machine, including a plain online install onto Hyper-V. The installer uses Podman to build the Talos boot images that provision the cluster nodes. Without it, the install cannot proceed past preparation.
The installer manages the machine for you: it creates one if none exists, sets it rootful, sizes it, and starts it. It adds user-mode networking only when it creates the machine and only when VPN Compatible Mode is on. That setting is off by default and is in the Podman group of the Settings workflow. You do not configure Podman by hand. What you must provide is the host support it depends on:
WSL 2 installed, current, and the default version (
wsl --status).Hardware virtualization enabled, because WSL 2 needs it just as Hyper-V does.
Enough headroom for the machine on top of Windows itself. On WSL 2 the machine's memory and CPU come from
%UserProfile%\.wslconfig, not from Podman.
Podman on WSL 2 is the component that most often breaks after a Windows or WSL update. If preparation fails on the Podman row, see Troubleshooting the Installer > Podman: Machine Will Not Start.
Cluster Nodes
VAM runs as a set of containerized services on a Kubernetes cluster. The nodes are the machines, physical or virtual, that make up that cluster. The requirements below are per node.
A cluster has these kinds of node:
Control plane nodes run Kubernetes itself (the API server, scheduler and the etcd datastore). The installer creates 1, 3 or 5 of them: 1 for an evaluation or a single-node install, 3 so etcd keeps quorum when one control plane node fails, 5 only for large sites. A cluster with 3 or 5 control plane nodes needs at least 2 worker nodes, and 3 or more are recommended. The Multi-node preset proposes 3 control plane nodes and 3 worker nodes. The wizard reports a layout it does not create below the node counts and does not continue with it.
Worker nodes run the VAM services and hold the persistent storage for media, indexes and databases.
A typical resilient production layout is therefore 3 control plane + 3 or more worker nodes. A single-node install collapses all roles onto one machine and is intended for evaluation and small productions.
Single-Node Evaluation
Requirement | Minimum |
|---|---|
CPU | 8 cores |
RAM | 32 GB |
Disk | 130 GB SSD (200 GB recommended) |
OS | Provisioned by the installer (Talos Linux) |
Multi-Node Production
Role | CPU | RAM | Disk | Count |
|---|---|---|---|---|
Control plane | 4 cores | 16 GB | 200 GB SSD | 3 |
Worker | 16 cores | 64 GB | 1 TB+ SSD | 3 or more |
Production sizing depends heavily on the size and volume of media your installation handles, so the numbers above are a starting point.
How the Installer Splits Disk per Node
You size each node with one number: its total disk. The installer divides that total into the kinds of disk a node actually gets:
a system disk (the Talos OS / ephemeral disk) that holds the container images plus every non-media persistent volume: OpenSearch, SurrealDB, RabbitMQ and friends all live here, because that storage class is backed by the node's own filesystem.
one or more object-storage drives, separate disks that hold the S3 object store (media, proxies, artifacts). Only workers get these drives, and only when the cluster has at least 2 workers. You choose how many drives each worker gets. The default is 1.
The system disk grows with the total only until it reaches 120 GB. Everything you add above that becomes asset storage, so a bigger node means a bigger S3 store, not a bigger operating system.
This split applies to Hyper-V clusters with at least 2 workers. A single-node cluster gets no object-storage drives: the node keeps its whole total as the system disk, and the asset store shares it. The installer does not create a cluster with exactly 1 worker. On VMware vSphere the installer also creates only the system disk, sized to the full total. See VMware vSphere Deployment > Sizing.
Node role | Total minimum | Total recommended | Split at the recommended total |
|---|---|---|---|
Single-node (control plane runs workloads) | 130 GB | 200 GB | all system, asset store included |
Control plane in a cluster that has workers | 40 GB | 60 GB | all system (no object storage) |
Worker with 1 object-storage drive | 130 GB | 200 GB | 120 GB system + 80 GB object storage |
The minimums come from the floors of the disk parts: 80 GB of system disk and 50 GB per object-storage drive. Each additional object-storage drive adds 50 GB to the minimum and 80 GB to the recommended total. A node without drives still needs that room for the asset store, so its minimum is the same. On a node with a drive, at exactly 130 GB the split lands on both floors at once. Above that the system disk grows with the total until it reaches 120 GB, and the rest goes to the asset store.
Every worker needs at least 1 object-storage drive. With 1 drive per worker the object store has no parity, and a failed drive or worker loses data. With 2 or 3 workers and 1 drive each, the installer splits every drive into 2 volumes, because the object store needs at least 4 volumes on 2 or 3 workers. Each volume gets half of the drive. With 2 or more drives per worker the installer sets the parity so that the data stays readable when one worker fails. For a production system, give every worker at least 2 object-storage drives.
A control plane only stays small when the cluster actually has worker nodes. It is tainted and runs no application pods, and it receives no object-storage drive. In a single-node install there are no workers, so the control plane is schedulable, carries the whole stack and holds the asset store on its system disk.
The installer refuses a node disk below its minimum total. The wizard accepts anything at or above the minimums and warns (without blocking) below the recommended totals.
On Hyper-V these are dynamically expanding VHDX files: a 120 GB system disk only occupies as much host disk as it actually contains, so sizing generously up front costs nothing until the data exists. Sizing too small does cost you, because existing VHDs are never resized by a later run of the installer.
The cluster nodes are created and managed for you by the installer and they run Talos Linux, a minimal immutable Kubernetes OS, so you do not install or patch a node operating system by hand. The only exception is the bring-your-own Kubernetes option below, where you supply the cluster yourself.
Deployment Target
The installer can either provision the cluster nodes for you (on Hyper-V or VMware vSphere) or deploy into a Kubernetes cluster you already run. Pick the option that matches your infrastructure.
Hyper-V (Windows): Single Host
The simplest option: the installer creates the Talos VMs on a single Windows Hyper-V host.
Hyper-V enabled on the host. The installer checks that it is enabled, but it does not enable it.
A network adapter that connects the host to the rest of your environment. The cluster reaches your network through an external virtual switch on that adapter. The installer uses an existing external switch that you select, or it creates one. The new switch is named
VAMPublicSwitchby default. If the adapter already belongs to another switch, the installer uses that switch. For an installation with worker nodes, the installer also creates the internal switchVAMPrivateSwitch, or uses it if it already exists.Enough free RAM and disk on the host to satisfy the node sizing above. For a single-node evaluation, budget the full node spec plus headroom for the host OS.
Hyper-V Host Cluster (HA Production)
For a highly-available production deployment, run the VAM cluster across a failover cluster of at least 3 Hyper-V hosts, so the loss of one physical host does not take VAM down. Recommended per Hyper-V host:
Component | Recommended |
|---|---|
CPU | 16-core (or dual 8-core) with VT-x/AMD-V + EPT/NPT (for example, AMD EPYC 7302P / Intel Xeon Silver 4210) |
Memory | 64 GB RAM (to comfortably allocate 8 GB+ per VM) |
System disk | 512 GB NVMe SSD (Hyper-V + host OS) |
VM storage | 2 TB+ NVMe / SAS SSD (local or shared) for VM data |
Networking | 2× 10 GbE NICs (VM traffic + storage replication) |
Host OS | Windows Server 2025 (Core) or equivalent |
The size of the shared or cluster storage on a Hyper-V failover cluster depends on your media volume and retention.
VMware vSphere
The installer can also provision the Talos nodes on VMware vSphere. The vCenter and ESXi versions, the network and the datastore it needs are in VMware vSphere Deployment > Environment Prerequisites, and the permissions in VMware vSphere Deployment > Required vSphere Permissions. That page also has the full walkthrough.
Bring Your Own Kubernetes Cluster
Instead of letting the installer create a cluster, you can point it at an existing Kubernetes cluster (on-prem or cloud). The requirements for that cluster, covering storage (DirectPV), certificates, ingress and identity, are in Existing Kubernetes Cluster > Cluster Prerequisites, and the steps in Existing Kubernetes Cluster > Running the Install.
Network
Item | Notes |
|---|---|
Static IP addresses for the cluster nodes | DHCP reservations by MAC address are acceptable as long as they are stable. |
DNS A record for the VAM endpoint | For example, |
DNS server reachable from the nodes | For in-cluster and external name resolution. |
NTP / time source reachable from the nodes | Talos and Kubernetes require accurate time (for example, |
Virtual switches (Hyper-V) | The installer creates them or uses existing ones, as described in Hyper-V (Windows): Single Host. On VMware vSphere, a network or port group with DHCP instead. |
HTTPS (443) outbound to | Or your internal mirror, for online installs. Not required for offline installs. |
HTTPS (443) outbound to | From the installer host, for a Hyper-V installation. The installer pulls the Talos image builder from it to build the boot images. |
HTTPS (443) outbound to | From the installer host. The installer uses the Talos Image Factory, for example to download the OVA for VMware vSphere. |
TCP 50000 and 6443 from the installer host | To the cluster nodes. The installer manages them through the Talos API on 50000 and the Kubernetes API on 6443. |
Firewall: TCP 30443 (gateway) | Inbound to the control plane node, from every client that uses VAM or the Configuration Service. Clients reach the installation on this port unless you put a load balancer in front that terminates on 443. Then clients leave the port out of the address, and the load balancer still needs TCP 30443 to the nodes. If you changed the gateway port, use that port instead of 30443. |
Firewall: TCP 6443 (Kubernetes API) | Inbound to the control plane node, only for administrators who use |
Firewall: TCP 30442 | Inbound from the installer host, for backup and restore. Inbound from the peer cluster, for active-active replication. |
TCP 22350 outbound to the WIBU license server | From the cluster nodes. Licensing fails without it. See Licensing. |
VAM signs users in with Microsoft Entra ID, see Identity Provider Setup. For TLS, larger deployments often use an internal certificate authority such as Active Directory Certificate Services (ADCS). A certificate authority is not required to install VAM: the installer can self-generate certificates. If you use ADCS, prepare the template described in Certificate Template for ADCS.
Certificate Template for ADCS
When ADCS issues the VAM certificates, VAM requests them through the web enrollment pages of your certification authority against one certificate template. Create the template as a duplicate of the built-in Web Server template with the following settings:
Subject Name: Supply in the request. VAM generates the key and the request inside the cluster, and the request carries the subject and the alternative names.
Extensions: Application Policies with Server Authentication, and Key Usage with digital signature and key encipherment.
Cryptography: RSA with a minimum key size of 2048 and SHA256 as the request hash.
Issuance Requirements: no CA certificate manager approval and no authorized signatures, so the CA issues the certificate immediately. With manager approval, every certificate and every renewal waits for an administrator.
General: a validity period of your choice, for example 1 to 5 years. VAM renews its certificates automatically before they expire.
Security: the user account you enter in the installer needs Read and Enroll on the template.
The private key never leaves the cluster, so the template does not need to allow the private key to be exported.
After you create the template, publish it on the issuing CA under Certification Authority > Certificate Templates > New > Certificate Template to Issue, and make sure the CA runs the Certification Authority Web Enrollment role: VAM requests certificates through its /certsrv pages. In the installer, enter the template name, not its display name, in Certificate template.
Licensing
VAM licenses are floating network licenses. They are held on a WIBU CodeMeter license server that the cluster connects to over the network, and they are never installed on the installer host or on a cluster node. The entitlement orchestrator pod is the only VAM component that talks to the license server. Every other VAM service asks the orchestrator.
What you must have before you start is a running CodeMeter license server, reachable from the cluster, holding the VAM licenses. During the install you tell the installer where that server is.
The License Server
Requirement | Detail |
|---|---|
Product | WIBU CodeMeter Network Server hosted by the CodeMeter Runtime. |
Network server enabled | CodeMeter serves licenses over the network only once Server Access is enabled, under Configuration > Server in CodeMeter WebAdmin on the license server. It is off by default. |
License container | The licenses are activated onto a container held on the license server: a USB dongle (CmDongle), a local software container (CmActLicense local), or a CmActLicense SmartBind Loose software container. |
Reachability | TCP 22350 from the VAM cluster nodes to the license server. The entitlement orchestrator can run on any node, so it must be reachable from all of them. |
Redundancy | Use one license server. A setup with more than one has not been verified, so contact Vizrt support before you list a second server. |
A license server is a small service. It does not need to be a dedicated machine, and it is not part of the VAM cluster. An existing CodeMeter license server already serving other Vizrt products can serve VAM too, once the VAM licenses are activated on it.
Licenses are delivered as a license ticket and activated onto a container through the Vizrt License Portal, online or offline. For the portal, CodeMeter WebAdmin and license server administration in general, see the Vizrt Licensing Administrator Guide.
VAM WIBU Licenses
All VAM licenses are WIBU User Limit licenses. Order them from your Vizrt representative. They arrive as a license ticket to activate on the license server's container.
License | Required | What it does |
|---|---|---|
VAM Core | Yes | Enables VAM. VAM services do not start without it. If it disappears while VAM runs, VAM turns read-only: every |
VAM User Essential | Yes | The user pool, licensed per concurrent user. Requires VAM Core. |
VAM Pilot Data | Only for Viz Pilot data | Licenses the Viz Pilot data capability. Requires VAM Core. |
VAM Replication | Only for active-active replication | Licenses replication between deployments. Requires VAM Core. |
Machine identities, such as a service account or a client-credentials integration, are not licensed as users.
Size the user pool for concurrent users, not named users.
What the Installer Needs
On the Deployment Configuration screen, the License section holds the License servers field. Type the hostname or IP address of the license server.
At least one entry is required. The installer blocks the step while the field is empty.
A hostname only works if the cluster nodes can resolve it. If you are unsure of DNS on the node network, use an IP address.
Before You Start
Have ready:
The hostname or IP address of the license server.
Confirmation that TCP 22350 is open from the cluster node network to that server.
Confirmation that the VAM licenses are present in the container on that server.
If the License Server Cannot Be Reached
VAM needs a reachable license server while it runs. If the license server does not answer, running VAM services turn read-only within about 30 seconds, and a VAM service that restarts during the outage does not start.
Identity Provider
VAM uses OpenID Connect (OIDC) for sign-in, with an identity provider that you already run.
Identity provider | Preparation needed before install |
|---|---|
Microsoft Entra ID | Register the applications, define the app roles, grant admin consent and add a client secret. See Identity Provider Setup > Microsoft Entra ID. |
Complete this setup before you launch the installer. It gives you the tenant ID, the client ID and the client secret that the installer asks for, so the installation runs in a single pass.
Monitoring and Log Storage
VAM does not include a system that keeps metrics, logs and traces over time. To monitor VAM, provide your own monitoring system that accepts the OpenTelemetry Protocol (OTLP). In most organizations this is one central system that serves many applications, or a hosted service such as Grafana Cloud, rather than a separate system for every VAM installation.
You set up and run this system yourself. That includes who can read the logs, how long they are kept and how they are protected, because logs can contain user names and other personal data. Vizrt does not install, run or support it.
To connect VAM to it, have the following ready:
The OTLP endpoint of your monitoring system (OTLP over HTTP, usually port 4318, or over gRPC, usually port 4317), reachable from the VAM cluster over TLS.
The credentials the endpoint requires, for example an API key.
The CA certificate that signed the endpoint's certificate, if your own certificate authority issued it.
Setting Up a Monitoring System
If you do not run such a system yet, set one up by following the documentation of the product you choose. Any system that accepts OTLP works. The products below are examples of compatible systems, not recommendations, and Vizrt does not maintain or support them.
To set up | Start here |
|---|---|
Any system or service that accepts OTLP | |
A hosted service | |
Self-hosted metrics | |
Self-hosted logs | |
Self-hosted traces | |
Dashboards for self-hosted data |
Before You Start the Installer
A short pre-flight checklist:
[ ] Installer host meets the requirements above.
[ ] You have chosen a deployment target (Hyper-V, VMware vSphere, or an existing Kubernetes cluster) and its hardware / hypervisor capacity is ready. See Deployment Target.
[ ] Static IPs, DNS record and a reachable NTP source allocated.
[ ] If ADCS issues the certificates: the certificate template exists and is published, and you have the web enrollment URL of the CA and a user with Read and Enroll on the template. See Certificate Template for ADCS.
[ ] License server running and reachable on TCP 22350, and you have its hostname or IP address. See Licensing.
[ ] You have completed the Identity Provider Setup for Microsoft Entra ID and have the client ID, client secret and tenant ID to hand.
[ ] You have a backup destination in mind (external S3 or MinIO, or a network mount), to configure after install.
[ ] If you want to monitor VAM: the OTLP endpoint of your monitoring system and its credentials. See Monitoring and Log Storage.
Once these are in place, continue to Running the Installer.