Peer-to-Peer Sync

SyncEngine lets nodes exchange signed events without a central coordinator.

File-based protocol

Each peer pair uses:

• outbound_dir
• inbound_dir

In production, these directories are usually mounted from shared transport or exchange channels.

SyncBatchFile structure

A batch file includes:

• sender metadata (node_id, cursor)
• list of proof envelopes
• watermark or cursor metadata for resume

Publish flow

1. Query unsynced ledger events.
2. Build deterministic sync batch.
3. Write sync batch file to peer inbound path.
4. Mark publish attempt in runtime metrics.

Receive flow

1. Read inbound batch file.
2. Verify each event signature.
3. Re-apply replay and dedupe checks.
4. Append new events to ledger.
5. Advance peer watermark on success.

Peer state tracking

Watermarks are stored so each peer can resume from last confirmed sync position.

Operational guidance

• Keep peer clocks synchronized to reduce replay false positives.
• Monitor sync_published vs sync_received drift.
• Alert when watermark staleness exceeds SLA.
• Rotate stuck inbound files to quarantine for manual inspection.