Delete full db directory with purge-chain subcommand

[skunert]

Closes #1767

Until now the purge-chain command would only remove the full subfolder of the db folder. However there is also the parachains db that currently remains and can cause problems on node restart.

Example wiht old code:

polkadot purge-chain --database paritydb --base-path /tmp/some-folder
Are you sure to remove "/tmp/some-folder/chains/polkadot/paritydb/full"? [y/N]: y
"/tmp/some-folder/chains/polkadot/paritydb/full" removed.

In this case /tmp/some-folder/chains/polkadot/paritydb/parachains would remain and might cause problem on node restart because of version conflicts as described in #1767. After this PR the whole /tmp/some-folder/chains/polkadot/paritydb folder will be deleted.

[arkpar]

It would be better to pass config.data_path here. It specifically points to a root where all databases are located.

[skunert]

It would be better to pass config.data_path here. It specifically points to a root where all databases are located.

If we nuke the entire data_path folder both rocksdb and paritydb would be removed in contrast to current behaviour of just removing the specified one. Is this what you are proposing?

[arkpar]

That's a good point. Still, in addition to the main database and the parachain db, there is other stuff that needs to be deleted. The statement store database added in paritytech/substrate#13701 in particular.
I suggest we deprecate --database option for the purge command and just delete everything in data_path.
@bkchr what do you think?

[bkchr]

and just delete everything in data_path.

Hmm, this would then also mean that we delete the keystore. I don't think we should do this. We could maybe introduce a new cli command purge-data-dir or whatever. That makes clear that it deletes everything, but I would not assume that delete database kills the keys.

[arkpar]

Alright, then this is good for now.

[skunert]

There was actually a test that was testing that parachains db survived. It mentioned one should use purge-chain --parachains which does not exist. I am not sure there is a use case to delete these separately, cc @cheme.

[cheme]

There was actually a test that was testing that parachains db survived. It mentioned one should use purge-chain --parachains which does not exist. I am not sure there is a use case to delete these separately, cc @cheme.

I remember deleting the parachain db being useful in some cases, at the time the point was that this could be safely deleted and would be restored from peers while keeping the substrate db.
I am not exactly sure if it should be a command on itself, or manual deletion is enough (I only expect this action to be done to fix some specific issues). I really don't know if it make sense to a have this command as a maintenance/regular action CC\ @ordian ?

[ordian]

In general, it's not safe for more than 1/3 of validators to delete parachains db folder (or at least not used to be), but deleting the substrate db should be ok as long as there's one (full) node with the latest block. This is probably the reason why we had this --parachains flag previously for an expert mode or testnets.