Before the first step
Do the whole path once on the testnet, with the same remote signer, sentries, and monitoring you will use on mainnet. Testnet mistakes cost nothing. Mainnet mistakes cost 0.01 percent of stake for downtime and 5 percent plus permanent removal for a double sign. Hardware comes first: read Hardware and system requirements and plan for the validator host, at least two sentry hosts in different locations, and a signer host.What each host runs
Three kinds of host, three different jobs:- Validator host: gaiad with
pexoff and peers limited to your sentries, no public ports, andpriv_validator_laddrpointing at the signer. It holds the node key and nothing else of value. - Sentry hosts: gaiad with the public peer-to-peer port open, the validator listed as a private and unconditional peer, and no keys of value. Two or more, in different networks.
- Signer host: the remote signer process and the consensus key (or the HSM that holds it), reachable only from the validator host on the privval port.
The path
1
Join mainnet with a full node
Follow Join mainnet to install gaiad, restore from a snapshot or state sync, and catch up to the chain head. Do not create the validator until
catching_up is false. This node becomes your validator. Your sentries are built the same way.2
Run gaiad as a service
Follow Run gaiad as a service so the node restarts on failure and on reboot, has a file descriptor limit high enough for the peer count, and logs to journald. Everything that follows assumes the service is in place.
3
Create the validator
Follow Create and manage a validator: fund the operator key, write
validator.json, send create-validator, and confirm the validator appears on chain. The same page covers editing the description and commission, unjailing, rewards, and a clean exit.4
Move the consensus key to a remote signer
Follow Remote signing. The consensus key should never be on the host that peers can reach. This is also the one step where a mistake (two signers with one key) ends the validator permanently, so read the operational rules before switching.
5
Put the validator behind sentries
Follow Validator security: sentry topology, the
config.toml keys on each side, the firewall rules per host, and where each key goes.6
Add monitoring and alerts
Follow Monitoring: enable the CometBFT Prometheus endpoint, scrape it, and alert on the signals that matter (not signing, height stalled, no peers, voting power 0). Add a second signal from another machine that queries the chain.
7
Rehearse an upgrade
Follow Upgrade with Cosmovisor. Hub upgrades are governance proposals with a fixed height. Missing one means jail. Run Cosmovisor on the testnet first. Know how to roll back a failed binary swap.
After onboarding
- Vote on every governance proposal. See Governance process.
- Watch the validator announcement channels on the Interchain Discord and the gaia releases page.
- Re-read the live parameters on the Validator overview after each governance round. Slashing and staking parameters change by proposal.
- Keep the Validator FAQ handy for the questions delegators ask.
Verify
Your validator is onboarded when all of these are true:- The query
gaiad query staking validator <cosmosvaloper>showsstatus: BOND_STATUS_BONDEDandjailed: false. - On the validator host,
curl -s localhost:26657/status | jq .result.validator_infoshows the public key held by your remote signer. That host has nopriv_validator_key.jsonwith a live key. - The validator host accepts peer connections only from your sentries, and the sentries hold the public address.
- A missed-blocks alert fires when you stop the signer on the testnet, and clears when you start it again.
- You have upgraded the testnet node through at least one governance upgrade height with Cosmovisor.