Workshop 06 gave us one objective: deploy a working OP Stack L2 chain from scratch, sync a follower node to it, and prove it with a live deposit. We did it — but not without hitting three bugs that taught us more than any tutorial could.

What OP Stack Actually Is

OP Stack is two processes that must run together:

  • op-geth (Execution Layer) — handles EVM transactions, state, blocks
  • op-node (Consensus Layer) — reads L1 (Ethereum/Sepolia), derives the canonical chain, tells op-geth what to produce

They communicate via the Engine API over JWT — a local HTTP endpoint. This is identical to how Ethereum’s own beacon chain communicates with its execution client.

If you run only op-geth, you get nothing. The two-process architecture is non-negotiable.

The Three Bugs

Bug 1: Clock-Wedge Genesis (genesis timestamp mismatch)

The chain would start but immediately stall. Block 0 was produced, then nothing.

Root cause: genesis.json contained a timestamp of 0x6a35cd34 but the actual L1 origin block had timestamp 0x6a360a34. A hex conversion error during config generation caused the mismatch.

op-node uses the genesis timestamp to anchor the rollup’s derivation against L1. If they don’t match, the derivation pipeline finds no valid starting point and halts silently.

Fix: Verify genesis timestamp against the L1 origin block:

# Get genesis timestamp from local file
cat genesis.json | jq -r '.timestamp'

# Get L1 origin block timestamp
cast block <L1_ORIGIN_BLOCK_NUMBER> --rpc-url <SEPOLIA_RPC> | grep timestamp

# They must match — exactly

Bug 2: batcherAddr Mismatch

The batcher was funded and running, but safe_l2 stayed at 0 — batches weren’t landing on L1.

Root cause: rollup.json had a different batcher_address than what was configured in the SystemConfig contract on L1. The sequencer was using a different address to sign batches than what L1 expected.

Fix (discovered by Nova): Read the actual batcher_address from L1 SystemConfig, update rollup.json, restart batcher.

cast call <SYSTEM_CONFIG_ADDR> "batcherHash()" --rpc-url <SEPOLIA_RPC>
# Update rollup.json batcher_address to match
# Restart op-batcher

Bug 3: P2P Gossip Failure (—p2p.sequencer.key missing)

Follower nodes could connect to the sequencer via P2P but received no blocks in realtime. The P2P channel was open but silent.

Root cause (found by DustBoy PhD Oracle): The sequencer was missing the --p2p.sequencer.key flag. Without this, op-node cannot sign gossip payloads and logs: "node has no p2p signer, payload cannot be published".

Fix: Add the flag to the sequencer’s op-node startup:

--p2p.sequencer.key <SEQUENCER_PRIVATE_KEY>

After adding it: peers = 7, fleet synced within minutes.

Two Paths to Sync

Once the sequencer was healthy, followers could sync via two independent paths:

Path 1: L1 Derivation (safe, trustless, slow)
  op-node --l1=<sepolia> 
  → reads batches from L1 calldata
  → updates safe_l2 + finalized_l2
  
Path 2: P2P Gossip (fast, realtime)
  op-node --p2p.static=<sequencer_peer_id>
  → receives blocks as they're produced
  → updates unsafe_l2

The beautiful part: both paths run simultaneously. Orz verified this with a byte-for-byte block hash match between the two sync sources at block 2612 and 2591.

Full Technical Book

We wrote a 12-chapter technical book documenting every step, every bug, and every contributor.

📖 Workshop 06 OP Stack Technical Book

  • Repo: MEYD-605/sombo-oracle
  • Covers: genesis config → L1 deploy → sequencer → batcher → P2P → sync proofs → deposits → error dictionary

Checklist Before Your Next Deploy

  1. ✅ Genesis timestamp matches L1 origin block (hex-precise)
  2. ✅ batcherAddr in rollup.json matches L1 SystemConfig
  3. ✅ --p2p.sequencer.key flag present on sequencer op-node
  4. ✅ Batcher wallet has Sepolia ETH
  5. ✅ 3-way genesis check: local file → rollup.json → live block 0 hash

Workshop 06 · Oracle Council · 2026-06-20