questions about state.scheme=path and path-based storage

[waynercheung]

Hi, from https://github.com/ethereum/go-ethereum/releases/tag/v1.13.0, it mentioned that The new database model is optional for now (you need to enable it via --state.scheme=path) and does require resyncing the state, , what does the require resyncing the state means? Resyncing from the genesis block or I just need to modify the configuration to set --state.scheme=path and the node will do the resycing work automatically?

What about the old data before I set --state.scheme=path?

[s1na]

If I remember correctly setting --state.scheme=path will fix up the data but if the node failed to start you will need to do the following steps:

geth --datadir PATH removedb

It will ask you whether to remove ancient data and state. Only ask it to remove the state data. But keep the ancient you don't need to redownload all of that again.

Then start geth with --state.scheme=path.

[rjl493456442]

If you already have an existing database using the legacy hash mode, setting
state.scheme=path will be rejected at startup.

To switch to path mode, you must enable it with a fresh key-value store
(you can still reuse the existing ancient chain data).

[waynercheung]

Hi @rjl493456442 @s1na , thanks for your reply, but I still can't understand it.

If I have run the geth as a fullnode for a while, and stop it, then restart it by setting state.scheme=path, the node will not run successfully, right?

What does the command geth --datadir PATH removedb mean? It will ask me remove the whole DB or just the state data?
what's the differences between removing ancient data and state or just removing state?

Another question, about the path-based storage model, what's the differences between traditional hash-based storage model, especially for the MPT or the state MPT?

And for the same block data, how do they calculate the stateRoot?
For the path-based model, if the content modifies, but the path keeps the same as before, what about the MPT structure? can you explain it?

[waynercheung]

Another question, what's the differences between fullnode and archive node by using path-based storage model?

[waynercheung]

Hi @rjl493456442 , what is the PR to implement path-based storage model, especially the modification for MPT.
#25963?
#26274?

[s1na]

Another question, what's the differences between fullnode and archive node by using path-based storage model?

Archive node is not possible with path-based model. Archive in hash mode will need around 20Tb of storage. It is only useful for particular cases.

What does the command geth --datadir PATH removedb mean? It will ask me remove the whole DB or just the state data?
what's the differences between removing ancient data and state or just removing state?

Ancient contains the whole history of blocks, txes and receipts. State is account balances and such. When you run that command ask it to remove the state but keep freezer/ancient.

[waynercheung]

@s1na Thanks for your reply.

I still can't understand the differences of MPT between path-based and hash-based storage, especially how to calculate the stateRoot and the structure of the MPT for path-based model.

Looking for your explain.