Description

This issue outlines a revised approach to handle block production shutdowns in Madara, improving on our previous proposal. The main goal is to eliminate transaction re-execution to preserve receipt consistency and ensure efficient handling of both graceful and ungraceful shutdown scenarios.

The revised approach shifts block sealing logic to the restart phase and introduces configuration-based decision-making, allowing for seamless block production handling even when configurations change.

Key changes in this proposal include:

1. No Block Sealing During Shutdown (whether graceful or ungraceful).
2. On Restart:
   • seal the block irrespective of the config, because closing the block doesn't depend on the configs.

Proposed Solution

Key Changes

• Avoid sealing blocks during both graceful and ungraceful shutdowns.
• On restart, seal the block and make sure we have updated the db for the pending txns and classes etc.

Data Persistence Requirements

• Persist pending transactions and their receipts in the database.
• Do not clear the database on restart to avoid unnecessary transaction re-execution.

Pseudo-Code Overview

async fn handle_restart() {
    seal_the_last_block();
    start_an_empty_block();
}