Domains: Limit domain's max extrinsic weight to max domain bundle weight

[vedhavyas]

We introduced a regression while setting the maximum domain block limit where in individual domain extrinsic was equal to max domain block limit. This would lead to excessive extrinsic weight allowance on domain leading to transaction being in txpool forever since this value exceeds the bundle weight limit defined on the consensus chain.

This PR ensures to use max bundle limit as the max domain extrinsic weight with bundle slot probability to always set 1. The current domain weights post this change

{
  baseBlock: {
    refTime: 390,584,000
    proofSize: 0
  }
  maxBlock: {
    refTime: 18,446,744,073,709,551,615
    proofSize: 18,446,744,073,709,551,615
  }
  perClass: {
    normal: {
      baseExtrinsic: {
        refTime: 124,414,000
        proofSize: 0
      }
      maxExtrinsic: {
        refTime: 216,666,666,666
        proofSize: 3,407,872
      }
      maxTotal: {
        refTime: 18,446,744,073,709,551,615
        proofSize: 18,446,744,073,709,551,615
      }
      reserved: {
        refTime: 0
        proofSize: 0
      }
    }
    operational: {
      baseExtrinsic: {
        refTime: 124,414,000
        proofSize: 0
      }
      maxExtrinsic: {
        refTime: 216,666,666,666
        proofSize: 3,407,872
      }
      maxTotal: {
        refTime: 18,446,744,073,709,551,615
        proofSize: 18,446,744,073,709,551,615
      }
      reserved: {
        refTime: 0
        proofSize: 0
      }
    }
    mandatory: {
      baseExtrinsic: {
        refTime: 124,414,000
        proofSize: 0
      }
      maxExtrinsic: {
        refTime: 216,666,666,666
        proofSize: 3,407,872
      }
      maxTotal: {
        refTime: 18,446,744,073,709,551,615
        proofSize: 18,446,744,073,709,551,615
      }
      reserved: null
    }
  }
}

If the domain bundle slot probability is lower than 1, the value of the maximum domain block increases but each extrinsic max weight will be limited to 250,000,000,000 which is 250 ms execution time.

TLDR:

On Consensus Chain:
Max allowed proof size for Normal extrinsics: 3.75 MiB
Max bundle weight:
- Execution: 250 ms
- Proof size: 3.25 MiB to have room for additional storage coming as part of Fraud proof
This is the bundle weight that operator uses while proposing the bundles. So each bundle should include one or more transactions that meet this bundle weight.

On Domains:
Max Domain block weight:
- Execution: u64::MAX, this is set to ensure all the extrinsics from bundles are executed irrespective of execution time
- Proof Size: u64::MAX, total proof size for entire domain block will not be used for FP but rather each extrinsic within the domain block

Max Extrinsic block weight:
This is same as max bundle weight on consensus chain.
- Execution: 250 ms
- Proof Size: 3.25 MiB, Single extrinsic can use this size with consideration for Fraud proof size

Closes: #2387

cc: @dariolina

Code contributor checklist:

• I have read, understood and followed contributing guide

[dariolina]

I agree, it's correct that extrinsic limit should be <= bundle limit. I still vouch for the bundle limit be in domain config (#2365) and deprecating calculate_max_bundle_weight. What is the hard limit then on bundle and extrinsic weight, if there was 1 bundle with 1 extrinsic in the block and we need to be able to fit a fraud proof of its invalid state transition?

[NingLin-P]

Make sense

[vedhavyas]

What is the hard limit then on bundle and extrinsic weight, if there was 1 bundle with 1 extrinsic in the block and we need to be able to fit a fraud proof of its invalid state transition?

The hard limit for 1 bundle and 1 big extrinsic would be 250 ms execution time. As per the proof size, this should be more than sufficient to fit that bit extrinsic in the fraud proof. Right now I'm checking with other exisiting para chains for their max POV size to be sure and then will merge once I'm even more confident

[NingLin-P]

Plz resolve the conflict

[vedhavyas]

@dariolina the proof size for each extrinsic is also updated to ensure it fits the consensus block during FP.
Do take a look again PR description for definitive numbers