Fix: Batcher User Balance Checks

[PatStiles]

Fix: Batcher User Balance Checker

Description

Replaces the hardcoded constant in https://github.com/yetanotherco/aligned_layer/blob/testnet/batcher/aligned-batcher/src/lib.rs#L607 with the user's min_fee. Addresses the first point of #1383 .

This fix is done by adding total_fees_in_queue variable in the UserState struct, and comparing if user has enough balance to pay for the needed amount.

This PR also refactors the infamous min_fee variable, to be called max_fee_limit / last_max_fee_limit

This PR also adds some metrics of this. dont forget to visualize them in the following tests running

make run_metrics

To Test:

test 1

• Run a local network and ensure proofs are being validated
• Submit a single proof with no balance in the BatcherPaymentService Contract :
  1.) Start a local testnet with an no balance:
  2.) Create a wallet with no funds

cast wallet new <KEYSTORE_PATH>

3.)

echo"Sending SP1 fibonacci task to Batcher..."cd batcher/aligned/ && cargo run --release -- submit \
--proving_system SP1 \
--proof ../../scripts/test_files/sp1/sp1_fibonacci.proof \
--vm_program ../../scripts/test_files/sp1/sp1_fibonacci.elf \
--repetitions 1 \
--proof_generator_addr <WALLET_ADDRESS> \
--keystore_path <KEYSTORE_PATH> \
--rpc_url http://localhost:8545 \
--network devnet

ensure the proof submission errors within the batcher with InsufficientBalance.

test 2

• Submit multiple proofs and not have enough to pay.
  1.) Fund the wallet by sending eth from a prefunded anvil account:

cast send <WALLET_ADDRESS --rpc-url http://localhost:8545 --private-key 2a871d0798f97d79848a013d4936a73bf4cc922c825d33c1cf7073dff6d409c6 --value 1ether

2.) Fund the batcher with an insufficient balance for multiple proofs.

aligned deposit-to-batcher \
--rpc_url http://localhost:8545 \  --network devnet \ 
--keystore_path <KEYSTORE_PATH> \
--amount 0.0026ether

3.) Send one proof, see it succeeds.

cargo run --release -- submit \
--proving_system SP1 \
--proof ../../scripts/test_files/sp1/sp1_fibonacci.proof \
--vm_program ../../scripts/test_files/sp1/sp1_fibonacci.elf \
--repetitions 1 \
--proof_generator_addr <WALLET_ADDRESS \
--keystore_path <KEYSTORE_PATH> \
--rpc_url http://localhost:8545 \
--network devnet

4.) Send a burst of 10 or more proofs, observe it submission fails due to insufficient balance.

cargo run --release -- submit \
--proving_system SP1 \
--proof ../../scripts/test_files/sp1/sp1_fibonacci.proof \
--vm_program ../../scripts/test_files/sp1/sp1_fibonacci.elf \
--repetitions 10 \
--proof_generator_addr <WALLET_ADDRESS \
--keystore_path <KEYSTORE_PATH> \
--rpc_url http://localhost:8545 \
--network devnet

Type of change

Please delete options that are not relevant.

• Refactor
• Fix

Checklist

• Linked to Github Issue
• This change requires new documentation.
  • Documentation has been added/updated.
• This change is an Optimization
  • Benchmarks added/run

[MauroToscano]

Code looks good, may need some extra testing

[MarcosNicolau]

Maybe I am wrong here but why don't we move the min_balance check after the msg_max_fee check? This way, in the min_balance check we'll verify against the msg_max_fee directly (as it will become the new min_fee if everything goes ok), we'll save ourselves this if:

if user_min_fee == U256::max_value(){
min_fee = user_max_fee
}

It is hard to understand that U256::max_value() is the default value.

[PatStiles]

My thoughts are to leave this check in as it handles an explicit case that a user's proof has been submitted for the first time not the replacement max_fee invariant. I we move this check behind the msg_max_fee check we still need to have logic to check and specify that if a proof is submitted for the first time we use it's submitted max_fee instead of the user_max_fee. I think it's worth considering verifying that these checks are not redundant, but lets address that in another issue.

[MarcosNicolau]

I tested on macos, everything went fine.

[uri-99]

the removed var, MIN_FEE_PER_PROOF, is no longer used in the codebase.
We should remove it from the constants file as well

[uri-99]

the check is not accurate, it is multiplying user's amt_of_proofs x min_fee_per_proof. This is a big wrong estimation, see pic attached:

[uri-99]

Its working now but I can't approve it because i made this pr

[MarcosNicolau]

Tested on macos, everything worked fine. Code looks good, I leave some comments that you can address if you want.

[uri-99]

1 sec testing something