Please add to the faq if running a pruned consensus node is required for a full archival consensus client node.

sigp / lighthouse

Ethereum consensus client in Rust

https://lighthouse.sigmaprime.io/

Apache License 2.0

2.87k stars 724 forks source link

Please add to the faq if running a pruned consensus node is required for a full archival consensus client node. #3549

Open ytrezq opened 2 years ago

ytrezq commented 2 years ago

I couldn’t find the answer myself. I just need my own node enpoint parity_tracecallmany for evading Tornado Cash rpc censorship so I would like to reduce database size for the consensus layer since I just need the execution layer.

I also failed to find how to run lighthouse as a light client.

michaelsproul commented 1 year ago

I just need my own node enpoint parity_tracecallmany for evading Tornado Cash rpc censorship so I would like to reduce database size for the consensus layer since I just need the execution layer.

A consensus node with state history is not required for an archival execution node. Lighthouse will store all historic blocks even when running in "pruned" mode, which is all that's required to fetch execution node history.

I also failed to find how to run lighthouse as a light client.

We don't have a light client.

I'll leave this issue open for someone to update the docs

ytrezq commented 1 year ago

@michaelsproul sorry for the wording but how to run lighthouse in pruned mode with only storing the last 10 epochs then?

michaelsproul commented 1 year ago

@ytrezq There's no mode that just stores the last N epochs. Your options currently are:

Store all blocks and some occasional states since the point at which you checkpoint synced. This is the default and the closest we have to a "pruned" mode until changes like https://github.com/sigp/lighthouse/issues/3211 and https://github.com/sigp/lighthouse/issues/3209 are implemented.
"Archive" mode, where all blocks and all historic states are available. Depending on how you set the --slots-per-restore-point flag, this will use anywhere from 50GB-3TB more disk space. More info here: https://lighthouse-book.sigmaprime.io/advanced_database.html.

chong-he commented 1 month ago

It has been a while since the issue is open, at this moment, the pruning in Lighthouse is pretty much auto-prune for states, blobs and payloads, see: https://discord.com/channels/605577013327167508/746918775919738893/1262720320465539112

The database for a freshly synced beacon node is ~100-120 GB

If you want to run as archive node (which is not required unless you want access to full states information), refer: https://lighthouse-book.sigmaprime.io/checkpoint-sync.html#reconstructing-states