Failed to convert BEEFY PublicKey to ETH address (westend & kusama)

[alindima]

Trying to upgrade the westend runtime in Versi, we start seeing a lot of errors like this one:
Failed to convert BEEFY PublicKey to ETH address.

As it scales with the number of validators and session frequency, this can pollute the logs of testnets.

CC: @acatangiu

[acatangiu]

Thanks for the report @alindima !

Initial BEEFY keys (ecdsa keys) are some dummy keys that need to be replaced by validators with their actual real keys by using rotate_keys().

So, on testnets, each validator's initial dummy BEEFY key will throw an error that it can't be converted to ETH address.

We might be able to "fix" this by modifying initial dummy BEEFY keys to be some custom sequences of bytes that can be successfully converted to ETH addresses.

But there's a question on whether we should or not.. now this error logs are arguably useful to motivate validators to rotate keys and actually register valid/unique BEEFY key. It indirectly reports that validators are using invalid BEEFY keys.
At the same time, they are annoying in tests..

@andresilva wdyt? should we "fix" it or not?

[bkchr]

to be some custom sequences of bytes that can be successfully converted to ETH addresses

How hard is this? If it is fairly trivial by setting some bits or something, we should do it.

[andresilva]

We are already doing .unwrap_or_default() when attempting the conversion, so I don't really see the need to generate "valid" dummy keys. Maybe just downgrade the logging to debug? We can instead do an error here if we see any of those keys, as this will only issue one log statement per block (and not proportional to the number of validators). We can also include the total number of 0x0 ETH addresses in that log statement and percentage of total.