Skip to content

Running a node

This section takes a geth you have installed to a node that syncs, keeps running, and stops cleanly. The steps are the same on every platform while the commands differ, so each platform has its own guide.

Choose your guide

You run the node on Guide How the guide keeps it running
Linux Linux users guide A systemd service
macOS Mac users guide A launchd agent
Windows Windows users guide A window that opens when you sign in
Docker Docker users guide A container restart policy

Every guide takes the same steps: install geth, choose the network, start the node, check that it syncs, use it, keep it running, stop it cleanly, and update it.

Choose your network

The network flag decides which chain the node follows. Pass it every time you start the node. On Linux, macOS and Docker, pass it to geth attach as well: attach looks for the node inside that network’s data directory.

Network Use it for Flag Chain ID
Ethereum Classic The main network --classic, or no flag 61
Mordor Testing: trying the client, testing contracts, rehearsing a deployment --mordor 63

MintMe.com Coin runs with --mintme, and a private network runs from its own genesis (Private network).

Each network keeps its own data directory:

Platform Ethereum Classic Mordor
Linux ~/.ethereum/classic ~/.ethereum/mordor
macOS ~/Library/Ethereum/classic ~/Library/Ethereum/mordor
Windows %LOCALAPPDATA%\Ethereum\classic %LOCALAPPDATA%\Ethereum\mordor
Docker /root/.ethereum/classic, inside the container /root/.ethereum/mordor, inside the container

To move a node to the other network, stop it and start it with the other flag. It syncs that network into that network’s directory and leaves the first network’s data where it was. --datadir puts the data somewhere else; give each network its own. Run a Mordor node covers running both networks on one machine.

Choose your configuration

A configuration is the set of flags that suits what the node is for. Add them to the start command, after the network flag. Your platform guide shows where they go, both when you start the node yourself and in the service that keeps it running. The flags below name Ethereum Classic; on Mordor, use --mordor in place of --classic.

The node is for Configuration Full guide
Yourself, or a wallet on the same machine A personal node This section
Crediting deposits and sending withdrawals An exchange or custodian Production operations
Serving JSON-RPC to other people A public RPC endpoint Public RPC endpoint
Explorers, indexers and analytics An archive node Archive node
Mining with your own hardware A solo mining node Mining
Supplying work to a mining pool A mining pool node Mining pool node
Helping other nodes find peers A bootnode Bootnodes and peer discovery

A personal node

1
--classic

The platform guide’s commands, unchanged. For a wallet on the same machine that connects over HTTP, add --http. The node then serves JSON-RPC on 127.0.0.1:8545, and to nothing beyond the machine.

An exchange or custodian

1
--classic --http --http.api eth,net,web3
  • Keep the HTTP listener on 127.0.0.1, or on a private network address set with --http.addr, where only your own systems reach it (RPC exposure).
  • Run at least two nodes on separate hosts, on the same release and the same MESS setting (Redundancy).
  • Credit a deposit only from a node that is in sync, after the number of confirmations the MESS confirmation calculator gives (Crediting deposits).

A public RPC endpoint

1
--classic --http --http.api eth,net,web3 --http.vhosts <hostname>
  • The node listens on 127.0.0.1, and a reverse proxy or RPC gateway faces the internet in front of it (What to put in front of the node).
  • --http.vhosts names the public hostname the proxy passes on (Hostnames).
  • Serve only eth, net and web3, and keep account keys off the node. The flags that cap gas, call time and batch size are under Limits the node enforces.

An archive node

1
--classic --syncmode full --gcmode archive
  • Start it on a new data directory. Full sync executes every block from genesis, and archive mode writes the state of every block to disk, so it needs more disk and time than the default (Hardware requirements).
  • Keep the hash scheme, which a new data directory gets by default. Under --state.scheme path, an archive node keeps no older state (With the path scheme).
  • Serve it to your applications with --http --http.api eth,net,web3. Tracing calls are in the debug namespace, which also holds methods that change the node (Serve it to others).

A solo mining node

1
--classic --mine --miner.etherbase <your address> --http --http.api eth,net,web3
  • Your mining software fetches work from the node with eth_getWork, on 127.0.0.1:8545 or a private network address. --mine does not mine on the node’s own CPU unless --miner.threads asks it to.
  • The node needs no key for the reward address: the address only goes into the blocks it builds.
  • Mine with a node that has finished syncing. To practice first, mine on Mordor.

A mining pool node

1
--classic --mine --miner.etherbase <pool address> --http --http.api eth,net,web3 --miner.notify http://127.0.0.1:<port>/
  • Only the pool software should reach the HTTP endpoint: anyone who reaches it can request work and submit results.
  • --miner.notify posts each new work package to the pool software as soon as the node has it.
  • Run at least two pool nodes, peered with each other, on the same release and MESS setting (Run more than one node).

A bootnode

1
--classic --nat extip:<public IP> --identity <your name>
  • Open port 30303 on TCP and UDP. --nat extip advertises the address other nodes can reach.
  • Keep geth/nodekey. It sets the enode ID other operators list, and a new key is a new ID.
  • The bootnode tool runs discovery without a chain (Run your own bootnode).

The same on every platform