Installing a Data Processor in Proxmox VE (Cluster)

You can deploy the Stellar Cyber data processor (DP) on a dedicated server running Proxmox VE, an open-source KVM-based hypervisor. In this model, you install Stellar Cyber with separate virtual machines for the Data Lake Master (DLm) and Data Analyzer Master (DAm), with cluster support ready for additional DA and DL worker nodes as your deployment grows.

This deployment uses Stellar Cyber's KVM (QCOW2) image and the Proxmox command line. Every step after the Proxmox VE installation itself is a shell command, run either from the built-in console (node > Shell) or over SSH. The web interface is used to confirm results rather than to perform the work — the CLI is deterministic, scriptable, repeatable across sites, and avoids several wizard behaviors that make the graphical path unreliable for this deployment.

Stellar Cyber deployment consists of the following major steps:

Before You Begin

Make sure the target system meets the minimum system requirements for installing a DP. The installation requires:

  • A dedicated server running Proxmox VE 9.x. This document was produced using Proxmox VE 9.2.2.
  • The server must operate as a dedicated hypervisor for a single Stellar Cyber deployment's virtual machines.
  • Management IP addresses:
    • One IP address for the Proxmox host management interface
    • Two IP addresses for each DP node in a cluster (1 management, 1 cluster)

  • Disks: All disks must be SSD.

    Stellar Cyber only supports full-SSD deployments. Spinning-disk-based storage (HDD) and hybrid drives (SSHD) are not supported. All deployments must adhere to this SSD-only policy in order to qualify for performance guarantees and technical support.
    Refer to Stellar Cyber Requires Full SSD Disks for further details.

    Storage must meet or exceed the following sustained minimum performance metrics for 4KB random data operations:

    • Write IOPS: > 30,000

    • Read IOPS: > 60,000

    • Throughput: > 500 MB/s (read/write)

The internal network of the DP uses the 172.17.0.0/16 and 10.244.x.0/24 subnets. Do not use these subnets for the management or cluster networks. The example in this guide uses 192.168.12.0/24 for the cluster network, matching Stellar Cyber's reference examples.

To prepare:

  1. Open the required ports on your firewall. When configuring the DP with separate VMs for the DL and DA, all ports between the nodes must be open.

  2. Contact Stellar Cyber support (support@stellarcyber.ai) for login credentials and a one-time password (also known as a License Key).

  3. Note the QCOW2 image URL. The same image is used to install both the DL and the DA:

    https://acps.stellarcyber.ai/release/6.6.0/dataprocessor/aella-dataprocessor-ubuntu2404-6.6.0.qcow2

    Use the QCOW2 image, not the OVA. Stellar Cyber publishes both an OVA (for VMware) and a QCOW2 (for KVM) at the same release path. The QCOW2 is the correct image for Proxmox — it is KVM-native, contains the required VirtIO drivers, and imports directly. See Troubleshooting Tips if you have already attempted an OVA import.

    Installation links point to the most recent release. To download a different version, simply substitute the version you want for the version specified in the link.

  4. Keep your one-time password (OTP) handy for configuring roles on the DP VMs.

    After license activation, you can find the OTP for your installation in the Licensing page.

Worked Example Values

This guide uses the following values throughout. Substitute your own.

Parameter

Value in this document

Proxmox host

10.20.0.231/24, gw 10.20.0.1

DLm management IP

10.20.0.6/24, gw 10.20.0.1

DAm management IP

10.20.0.7/24, gw 10.20.0.1

DLm cluster IP

192.168.12.25/24

DAm cluster IP

192.168.12.50/24

DNS

8.8.8.8

VM system disk pool

vmstore (ZFS, 1 × 1.92 TB SSD)

Data Lake data pool

datalake (ZFS, 7 × 1.92 TB SSD stripe)

DLm VM ID / name

100 / dl-master

DAm VM ID / name

101 / da-master

Minimum System Requirements

Refer to System Requirements and Capacity Planning for cluster node counts by daily ingestion volume.

Production deployment minimums used in this document:

VM

vCPU

RAM

System Disk

Data Disk

Data Lake Master (DLm)

40

136 GB*

500 GB (from image)

Elasticsearch — sized per capacity planning

Data Analyzer Master (DAm)

44

64 GB*

500 GB (from image)

—

* Adjust RAM to the published system requirements for your ingestion volume.

Reserve host resources. ZFS uses host memory for its ARC cache, so size the Proxmox host with approximately 10% more RAM than the total allocated to the virtual machines. The reference host in this document has 44 physical cores (88 threads) and 256 GB RAM; the two VMs consume 84 of 88 threads and 200 GB of RAM, leaving 32 GB for the ARC plus headroom for Proxmox VE itself.

Each VM disk must be thick-provisioned. On ZFS storage this means the zvol carries a full reservation — see Configure Storage.