feat: Multi-Database Writes

[ExoMonk]

Summary

Enables simultaneous PostgreSQL + ClickHouse writes in rindexer (and future ones). Events, checkpoints, reorg cleanup, and factory addresses now write to both backends in parallel when both are configured. Single-backend users see zero behavioral change.

What changes

New abstractions

• Database trait (insert_bulk + backend_name): extensible for future backends (S3, SQLite)
• DatabaseBackends struct Vec<Arc<dyn Database>> for parallel join_all writes + typed postgres/clickhouse fields for backend-specific operations (checkpoints,reorg, DDL)

Core updates

• Config structs (ContractEventProcessingConfig, FactoryEventProcessingConfig) now carry databases: DatabaseBackends instead of separate postgres/clickhouse fields
• EventProcessingConfig enum preserves .postgres() / .clickhouse() accessors + new .databases() accessor
• Write path in no_code.rs collapsed from two sequential if-let blocks to single databases.insert_bulk() call
• Cron scheduler: 10 function signatures migrated
• Code-gen: 3-way conditional (PG-only / CH-only / both) in events + trace bindings

Breaking changes

None. Single-backend configs produce identical behavior and code-gen output. Inherent insert_bulk methods on PostgresClient / ClickhouseClient are preserved: Rust-mode users calling pg.insert_bulk(...) directly still compile unchanged.

Examples

• Minimal dual-write (PG + CH)

  name: myIndexer
  project_type: no-code
  networks:
    - name: ethereum
      chain_id: 1
      rpc: https://mainnet.gateway.tenderly.co
  storage:
    postgres:
      enabled: true
    clickhouse:
      enabled: true

• Production dual-write

  name: myIndexer
  project_type: no-code
  networks:
    - name: ethereum
      chain_id: 1
      rpc: https://mainnet.gateway.tenderly.co
  storage:
    postgres:
      enabled: true
    clickhouse:
      enabled: true
      drop_each_run: false
    write_policy: any
    circuit_breaker:
      enabled: true
      failure_threshold: 5
      cooldown_seconds: 60
    max_batch_size: 1000

If CH goes down, indexer continues writing to PG. CH circuit trips after 5 consecutive failures, probes every 60s. Large batches split into 1000-row chunks.

• PG primary + CH shadow

  storage:
    postgres:
      enabled: true
    clickhouse:
      enabled: true
    write_policy: primary_with_shadow
    circuit_breaker:
      enabled: true
      failure_threshold: 3
      cooldown_seconds: 30

Error only if PG fails. CH failures are logged but tolerated. CH catches up from its stale checkpoint on restart.

• Single-backend (backward compatible, no change needed)

  storage:
    postgres:
      enabled: true

Identical behavior to before. All new fields are optional with safe defaults.

[vercel]

@ExoMonk is attempting to deploy a commit to the joshaavecom's projects Team on Vercel.

A member of the Team first needs to authorize it.