Make inserts and reorg handling atomic

[iuwqyir]

TL;DR

Implemented atomic block data operations to improve data consistency during reorgs and commits.

What changed?

• Created a new ReplaceBlockData method in ClickHouse storage that atomically handles both deletion and insertion of block data
• Replaced separate DeleteBlockData and InsertBlockData calls in the reorg handler with a single atomic ReplaceBlockData call
• Reordered the commit process to ensure data is saved to main storage before publishing to Kafka to prevent duplicate publishing
• Added new ClickHouse tables and materialized views to support the atomic operations:
  • Created inserts_null_table to store block data in a denormalized format
  • Added materialized views to transform this data into the normalized tables

How to test?

1. Run the application with a chain that experiences frequent reorgs
2. Verify that block data remains consistent during reorgs
3. Check that no data is lost when blocks are replaced during reorgs
4. Verify that Kafka messages are only published after data is successfully committed to storage

Why make this change?

The previous implementation had a race condition where block data could be inconsistent during reorgs. If the process failed between deleting old data and inserting new data, the database would be left in an inconsistent state. This change makes the replace operation atomic, ensuring data consistency even if failures occur.

Additionally, by ensuring data is saved to storage before publishing to Kafka, we prevent situations where consumers might receive notifications about data that isn't yet available in the database.

[iuwqyir]

• Make inserts and reorg handling atomic #192 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.