This document describes how to run a lighthouse node with minimal resources and time on a live network.

This procedure should ONLY be used for testing networks and never in production and never with attached validators. The Lighthouse node described in this state is only a partially functioning node.

We are going to run a single lighthouse node connected to a live network, without syncing and without an execution engine. This should only ever be done for testing.

There two main components needed.

1. A lighthouse node that doesn't sync
2. A fake execution client that does nothing

We will start with the second

This is a service that runs and fakes an execution engine. We firstly need to install the lighthouse lcli tool.

$ make install-lcli

Once installed, run the fake execution client:

$ lcli mock-el --jwt-output-path /tmp/mockel.jwt

This will create a server listening on localhost:8551

To create a lighthouse node that doesn't sync we need to compile it with a special flag.

$ cargo build --release --bin lighthouse --features network/disable-backfill

Once built, it can run via checkpoint sync on any network, making sure we point to our mock-el

Holesky testnet:

$ lighthouse --network holesky bn --execution-jwt /tmp/mockel.jwt --checkpoint-sync-url
https://holesky.checkpoint.sigp.io --execution-endpoint http://localhost:8551

Mainnet:

$ lighthouse --network mainnet bn --execution-jwt /tmp/mockel.jwt --checkpoint-sync-url
https://checkpoint.sigp.io --execution-endpoint http://localhost:8551

Additional flags, such as metrics may be added.

The above is assuming that you have not run the command in the past. If you have a database in existence for the network you are testing, checkpoint sync will not start. You may need to add the --purge-db flag to remove any past database and force checkpoint sync to run.