2. Prerequisites¶
2.1 Hardware — per node¶
| Item | Requirement |
|---|---|
| CPU | x86_64, 4+ vCPU/cores recommended |
| Memory | 8 GB RAM recommended minimum (more improves ZFS ARC cache hit rate) |
| Firmware | UEFI (GPT), Secure Boot OFF. Legacy BIOS is not supported. |
| OS / boot disk | A dedicated disk, ≥ 20 GB usable after the fixed 45 GB RAUC A/B layout (see Step-by-Step Installation). Only ONE disk should be visible to the installer at install time — attach any data/storage disks afterward. |
| Storage disk(s) | One or more SPARE disks (any bus), never the boot disk, for the ZFS pool. Each must expose a unique serial/WWN (needed for stable /dev/disk/by-id resolution). |
| Optional: SLOG device | A fast SSD/NVMe device, separate from the data disks, if accelerated synchronous writes are wanted (cluster role only). |
| Network interfaces | Cluster: at least 2 NICs (one for the data/client network, one for the private heartbeat network). Standalone: 1 NIC is sufficient. |
2.2 Network requirements¶
Plan the following values before starting — the installer asks for all of them and validates their format, but does not choose them for you:
| Value | Applies to | Notes |
|---|---|---|
| Node IP address(es) | both roles | Plain IPv4, no /prefix — the data network address of each node |
| Data network prefix / gateway / DNS | both roles | Standard IPv4 network settings for the client-facing network |
| Heartbeat IP + subnet | cluster only | Private, node-to-node only network (e.g. 10.10.10.0/24); never routed to clients |
| Virtual IP (VIP) + prefix | cluster only | Floats to whichever node is active; this is the address clients and the GUI actually use |
| S3 DNS name | both roles | FQDN (e.g. s3.example.lab) that resolves to the node IP (standalone) or the VIP (cluster); goes into the S3 TLS certificate |
| NTP server(s) | both roles, optional | Hostnames or IPs; can be set at install time or later in the GUI. Use the same server(s) on both cluster nodes. |
| Timezone | both roles, optional | Set later in the GUI; use the same zone on both cluster nodes for consistent file timestamps |
| Quorum device host | cluster only | A separate, already-configured qnetd arbitrator (its IP, hostname and hacluster password) |
2.3 Installation media¶
- The NearZero installer ISO (bootable, offline-capable — no internet access is required during installation).
- A way to boot it on the target: USB flash media, IPMI/iDRAC/iLO virtual media, or a virtual CD-ROM on a hypervisor.
2.4 Access¶
- Console access to each node (physical, IPMI/KVM, or hypervisor console) is required at least for the first boot.
- Remote root SSH login is disabled automatically at the end of installation, by design. Plan to use the console, or the restricted local "support" operator account, for anything needed afterward outside the GUI.
- The bootstrap root console password is provided separately by NearZero for each ISO build — it is not published in this guide.
2.5 Cluster-only: quorum device¶
A two-node cluster needs a third, independent vote to avoid a split-brain when the two nodes can't see each other. This is a small, separate host (can be a VM) running the quorum-device role, set up once and shared by one or more clusters. It must be reachable from both nodes before the cluster role's Phase B (primary) step runs.