sc_cli `key insert` gives wrong name to _gran_ key file created in local keystorage (BREAKS blocks finalization)

[agryaznov]

(found while passing this Tutorial)

Summary

key insert gives a wrong name to file containing gran (ed25519) key, which effectively leads to broken GRANDPA finalization in node

Details

Performing the key insert command operation with pre-generated Alice' key

# insert Alice's sr25519 aura key into storage
$ ./target/release/node-template key insert --base-path /tmp/node01 --chain local --key-type aura --suri "clip organ olive upper oak void inject side suit toilet stick narrow"

# insert Alice's ed25519 gran key into storage
$ ./target/release/node-template key insert --base-path /tmp/node01 --chain local --key-type gran --suri "clip organ olive upper oak void inject side suit toilet stick narrow"

leads to creating these two files :

$ find /tmp/node01/chains/local_testnet/keystore -type f -printf "%f\n"
617572619effc1668ca381c242885516ec9fa2b19c67b6684c02a8a3237b6862e5c8cd7e
6772616e9effc1668ca381c242885516ec9fa2b19c67b6684c02a8a3237b6862e5c8cd7e

one of which is possibly wrong-named, as in their names:

• 61757261 encodes aura key type
• 6772616e encodes gran key type

and the other 9effc1668ca381c242885516ec9fa2b19c67b6684c02a8a3237b6862e5c8cd7e part derives from public key, which should be different for aura (sr25519) and gran (ed25519), but it turns out to be just the same for both

how it (supposedly) should work

if we just insert the same keys using JSON RPC (e.g. with this script), we get the gran key file named properly (and nodes finalizing blocks successfully)

$ find /tmp/node01/chains/local_testnet/keystore -type f -printf "%f\n"
617572619effc1668ca381c242885516ec9fa2b19c67b6684c02a8a3237b6862e5c8cd7e
6772616eb48004c6e1625282313b07d1c9950935e86894a2e4f21fb1ffee9854d180c781

[agryaznov]

found in version:

[dependencies.sc-cli]
features = ['wasmtime']
git = 'https://github.com/paritytech/substrate.git'
tag = 'monthly-2021-09+1'
version = '0.10.0-dev'

[agryaznov]

Figured out that:

InsertKeyCmd run() function just assumes the default Sr25519 crypto scheme when no --scheme parameter set to node-template key insert command, no matter which --key-type was set.

But shouldn't it implicitly derive the Ed25519 crypto scheme in case when --key-type=gran ?

[nuke-web3]

Thanks for the docs correction to specify scheme. I think this is OK to close, but possibly there should be a fix in the CLI's behavior to try and reason about if your key is correct or not. I will let the core team comment on that idea 🙏

[bkchr]

But shouldn't it implicitly derive the Ed25519 crypto scheme in case when --key-type=gran ?

It doesn't know this.

However, with this: #9909 the scheme will be required to be passed and not assumed anymore.