Validator Requirements
Everything you need in place before deploying.
Hardware
Three tiers. Minimum keeps the node running; recommended is what a production validator should run; high traffic is headroom as shielded activity grows. Running a Node compares all four roles; this is the one that runs consensus.
| Component | Minimum | Recommended (production) | High traffic |
|---|---|---|---|
| CPU | 4 dedicated cores | 8 dedicated cores | 16+ dedicated cores |
| RAM | 16 GB | 32–64 GB | 128 GB |
| Storage | 250–500 GB NVMe | 1 TB NVMe | 2 TB+ NVMe |
| Network | 100 Mbps, static IPv4 | 1 Gbps, static IPv4 | 1 Gbps+, static IPv4 |
| Disk | SSD/NVMe | Fast NVMe | Enterprise NVMe |
| OS | Ubuntu 24.04 LTS or later | Ubuntu 24.04 LTS or later | Ubuntu 24.04 LTS or later |
Storage is sized for default pruning plus the shielded pool's Merkle tree, which only grows: every shielded note inserts leaves that are never removed. An archive node needs considerably more — a validator does not run one.
Ubuntu 22.04 is no longer supported. The install and operating guides are written and tested against 24.04 only.
Orbinum validators verify Groth16 proofs inside block execution. Verification is CPU-bound and arrives in bursts: a block carrying several shielded operations costs far more than an empty one.
Burstable and shared-vCPU instances — the cheap tier at most providers — throttle exactly when a proof-heavy block arrives. A throttled validator misses its slot, and missed slots stall finality for everyone. Use dedicated-vCPU instances or bare metal.
Hosting providers
Use a crypto-friendly provider: one whose acceptable-use policy allows proof-of-stake validation outright, in writing. A validator on a provider that forbids it can be terminated without notice, and a terminated validator misses slots for everyone.
Hetzner and Contabo are not accepted. Hetzner's terms prohibit crypto workloads and its support has confirmed the ban covers PoS validation and node hosting. Contabo is excluded on reliability grounds. Several large clouds sit in between and require written permission for staking — get it before renting, or pick a provider that does not ask. Background on who allows what: Cloud ToS & Crypto: When Your Validator Is Not Welcome.
Region matters. The set is selected for geographic diversity, so the network does not depend on one country or one data-center cluster. A node in a region the set does not cover yet is worth more than a fifth node in the region it already does.
Software
Docker with the Compose plugin, nothing else — see Installation.
The node must stay on a current release. Governance can require a minimum node version from block authors, and a validator below it stops authoring until it upgrades — see Minimum node version.
Network
Port 30333 must be reachable from the internet, not merely open in ufw — see
Ports and The node has no peers.
Accounts and keys
Three distinct pieces of key material, with different homes and different risk:
| Item | Where it comes from | Notes |
|---|---|---|
| Validator account (SS58) | Any Substrate wallet — Talisman or SubWallet in the app | Your staking account: it bonds the self-stake (at least 1,000 ORB on testnet), registers the session keys, validates and is paid. Hold the bond plus a little for fees |
| Session keys (Aura + GRANDPA) | Generated on the node by author_rotateKeysWithOwner, with a proof of possession | Private halves never leave the server |
| Node key | openssl rand -hex 32 | Your libp2p identity (Peer ID). Keep it stable across restarts |
Must stay on
The shipped validator stack enables three things by default. All three are required for the lifetime of the validator; turning any of them off is grounds for removal from the allowlist.
| Service | What it does | Where it is configured |
|---|---|---|
| Watchtower | Pulls and rolling-restarts the node when a new image ships. Keeps the whole set on one binary, at or above any minimum node version | watchtower service in the Compose file — see Updates |
| Orbinum telemetry | Reports sync state and authoring to the dashboard. The only signal the team has that your node is alive | TELEMETRY_URL in .env, at verbosity 1 — see Telemetry |
| Hardware benchmarks | On startup the node measures CPU, memory and disk against the Substrate reference hardware and reports it to telemetry — how the hardware table is verified. A node that reports below reference, or not at all, will not be approved | On by default. Do not add --no-hardware-benchmarks to the command |