Ev Node Explainer

Package Overview

Block Package Deep Dive

The block package is the most complex part of ev-node. See block-architecture.md for the complete breakdown.

Component Summary

Components struct:
├── Executor    - Block production (Aggregator only)
├── Reaper      - Transaction scraping (Aggregator only)
├── Syncer      - Block synchronization
├── Submitter   - DA submission and inclusion
└── Cache       - Unified state caching

Entry Points

• NewAggregatorComponents() - Full node that produces and syncs blocks
• NewSyncComponents() - Non-aggregator that only syncs

Key Data Types

Header - Block metadata (height, time, hashes, proposer) Data - Transaction list with metadata SignedHeader - Header with proposer signature State - Chain state (last block, app hash, DA height)

Block Production Flow (Aggregator)

Sequencer.GetNextBatch()
    │
    ▼
Executor.ExecuteTxs()
    │
    ├──► SignedHeader + Data
    │
    ├──► P2P Broadcast
    │
    └──► Submitter Queue
            │
            ▼
        DA Layer

Block Sync Flow (Non-Aggregator)

┌─────────────────────────────────────┐
│           Syncer                     │
├─────────────┬─────────────┬─────────┤
│ DA Worker   │ P2P Worker  │ Forced  │
│             │             │ Incl.   │
└──────┬──────┴──────┬──────┴────┬────┘
       │             │           │
       └─────────────┴───────────┘
                  │
                  ▼
          processHeightEvent()
                  │
                  ▼
          ExecuteTxs → Update State

Data Availability Layer

The DA layer abstracts blob storage. ev-node uses Celestia but the interface is pluggable. See da-sequencing.md for full details.

Namespaces

DA uses 29-byte namespaces (1 byte version + 28 byte ID). Three namespaces are used:

DA Client Interface

type Client interface {
    Submit(ctx, data [][]byte, gasPrice, namespace, options) ResultSubmit
    Retrieve(ctx, height uint64, namespace) ResultRetrieve
    Get(ctx, ids []ID, namespace) ([]Blob, error)
}

Key Files

Sequencing

Sequencers order transactions for block production. See da-sequencing.md for full details.

Two Modes

Sequencer Interface

type Sequencer interface {
    SubmitBatchTxs(ctx, req) (*SubmitBatchTxsResponse, error)
    GetNextBatch(ctx, req) (*GetNextBatchResponse, error)
    VerifyBatch(ctx, req) (*VerifyBatchResponse, error)
    SetDAHeight(height uint64)
    GetDAHeight() uint64
}

ForceIncludedMask

Batches include a mask distinguishing tx sources:

type Batch struct {
    Transactions      [][]byte
    ForceIncludedMask []bool  // true = from DA (must validate)
}

This allows the execution layer to skip validation for already-validated mempool txs.

Key Files

Forced Inclusion

Forced inclusion prevents sequencer censorship:

1. User submits tx directly to DA layer
2. Syncer detects tx in forced-inclusion namespace
3. Grace period starts (adjusts based on block fullness)
4. If not included by sequencer within grace period → sequencer marked malicious
5. Tx gets included regardless

Key Files

Common Questions

How does block production work?

The Executor runs executionLoop():

1. Wait for block time or new transactions
2. Get batch from sequencer
3. Execute via execution layer
4. Create SignedHeader + Data
5. Broadcast to P2P
6. Queue for DA submission

How does syncing work?

The Syncer coordinates three workers:

• DA Worker - Fetches confirmed blocks from DA
• P2P Worker - Receives gossiped blocks
• Forced Inclusion - Monitors for censored txs

All feed into processHeightEvent() which validates and executes.

What happens if DA submission fails?

Submitter has retry logic with exponential backoff. Status codes:

• TooBig - Splits blob into chunks
• AlreadyInMempool - Skips (duplicate)
• NotIncludedInBlock - Retries with backoff
• ContextCanceled - Request canceled

How is state recovered after crash?

The Replayer syncs execution layer from disk:

1. Load last committed height from store
2. Check execution layer height
3. Replay any missing blocks
4. Ensure consistency before starting

Architecture Diagrams

For detailed component diagrams and state machines, see block-architecture.md.