Operator Checklist

Official Monad: 16-core CPU with 4.5 GHz+ base (e.g. Ryzen 9950X / 7950X / EPYC 4584PX). Prefer bare metal — VMs are not officially supported. Use Susano catalog + Fit Checker to avoid overloaded ASNs/providers.

Official Monad minimum: 32 GB+ RAM. Susano often targets ~128 GB for operational headroom (Susano recommendation, not a Foundation minimum). Confirm the SKU RAM on the provider order page.

Official: 2 TB dedicated TrieDB + 500 GB for MonadBFT/OS, PCIe Gen4x4 NVMe or better. Prefer proven drives (e.g. Samsung 980/990 Pro, PM9A1); avoid known unstable consumer SSDs under sustained load.

Official bandwidth: ≥300 Mbit/s for validators, ≥100 Mbit/s for full nodes. Also pick geography wisely — Gaps + Deployment Planner show thin countries/providers so you don’t stack another node on the densest hubs.

Follow current Monad install docs for supported OS/kernel (Ubuntu 24.04+ is common in node ops). Keep the host patched, but don’t surprise-upgrade mid-epoch without a plan.

Tight consensus timing needs accurate clocks. Configure chrony/NTP and verify offset stays near zero before joining the set.

Open only the ports required by the current Monad node docs; firewall the rest. Don’t expose admin SSH/RPC to the world.

Install host metrics + process checks before mainnet stake. Wire alerts into a channel you actually watch. Susano Alerts + Stress Lab help think about failure domains.

Define snapshot/backup + restore drill before go-live. Know RPO/RTO. Keys stay offline — this checklist never asks for them.

Read the official release notes end-to-end (breaking flags, DB migrations, config changes). Don’t upgrade from a Discord rumor alone.

Snapshot / backup state and keep the previous binary handy before you touch production.

Validate the upgrade on testnet/staging with the same config pattern. Confirm sync + liveness before touching the active validator.

Document exact rollback steps (binary path, config, restart order). If you can’t roll back in minutes, you’re not ready to upgrade.

Pick a maintenance window, notify stakeholders, and raise alert sensitivity. Avoid upgrading during known network stress if you can wait.

Alert on process down, sync lag, disk fill, and peer loss — not only CPU pretty charts. Route pages to humans who can act.

One dashboard for CPU/RAM/disk/net + node-specific sync/height. Susano Overview/Heatmaps help see network context after you’re live.

Centralize logs with retention long enough to debug an incident overnight. Redact secrets; never paste keys into tickets.

External liveness checks catch “host up, validator stuck”. Pair with Susano Explore/Alerts for stake-set visibility.

TrieDB growth will fill disks — alert before 80–85% full and know the expand/migrate path. Official split: TrieDB vs OS volumes.