Struct RocksDBProvider 

pub struct RocksDBProvider(/* private fields */);

RocksDB provider for auxiliary storage layer beside main database MDBX.

Implementations

impl RocksDBProvider

pub fn check_consistency<Provider>( &self, provider: &Provider, ) -> ProviderResult<Option<BlockNumber>>

where Provider: DBProvider + StageCheckpointReader + StorageSettingsCache + StaticFileProviderFactory + TransactionsProvider<Transaction: Encodable2718>,

Checks consistency of RocksDB tables against MDBX stage checkpoints.

Returns an unwind target block number if the pipeline needs to unwind to rebuild RocksDB data. Returns None if all invariants pass or if inconsistencies were healed.

Invariants checked

For TransactionHashNumbers:

• The maximum TxNumber value should not exceed what the TransactionLookup stage checkpoint indicates has been processed.
• If RocksDB is ahead, excess entries are pruned (healed).
• If RocksDB is behind, an unwind is required.

For StoragesHistory:

• The maximum block number in shards should not exceed the IndexStorageHistory stage checkpoint.
• Similar healing/unwind logic applies.

Requirements

For pruning TransactionHashNumbers, the provider must be able to supply transaction data (typically from static files) so that transaction hashes can be computed. This implies that static files should be ahead of or in sync with RocksDB.

impl RocksDBProvider

pub fn new(path: impl AsRef<Path>) -> ProviderResult<Self>

Creates a new RocksDB provider.

pub fn builder(path: impl AsRef<Path>) -> RocksDBBuilder

Creates a new RocksDB provider builder.

pub fn is_read_only(&self) -> bool

Returns true if this provider is in read-only mode.

pub fn tx(&self) -> RocksTx<'_>

Creates a new transaction with MDBX-like semantics (read-your-writes, rollback).

Note: With OptimisticTransactionDB, commits may fail if there are conflicts. Conflict detection happens at commit time, not at write time.

Panics

Panics if the provider is in read-only mode.

pub fn batch(&self) -> RocksDBBatch<'_>

Creates a new batch for atomic writes.

Use Self::write_batch for closure-based atomic writes. Use this method when the batch needs to be held by crate::EitherWriter.

Panics

Panics if the provider is in read-only mode when attempting to commit.

pub fn batch_with_auto_commit(&self) -> RocksDBBatch<'_>

Creates a new batch with auto-commit enabled.

When the batch size exceeds the threshold (4 GiB), the batch is automatically committed and reset. This prevents OOM during large bulk writes while maintaining crash-safety via the consistency check on startup.

pub fn get<T: Table>(&self, key: T::Key) -> ProviderResult<Option<T::Value>>

Gets a value from the specified table.

pub fn get_encoded<T: Table>( &self, key: &<T::Key as Encode>::Encoded, ) -> ProviderResult<Option<T::Value>>

Gets a value from the specified table using pre-encoded key.

pub fn put<T: Table>(&self, key: T::Key, value: &T::Value) -> ProviderResult<()>

Puts upsert a value into the specified table with the given key.

Panics

Panics if the provider is in read-only mode.

pub fn put_encoded<T: Table>( &self, key: &<T::Key as Encode>::Encoded, value: &T::Value, ) -> ProviderResult<()>

Puts a value into the specified table using pre-encoded key.

Panics

Panics if the provider is in read-only mode.

pub fn delete<T: Table>(&self, key: T::Key) -> ProviderResult<()>

Deletes a value from the specified table.

Panics

Panics if the provider is in read-only mode.

pub fn clear<T: Table>(&self) -> ProviderResult<()>

Clears all entries from the specified table.

Uses delete_range_cf from empty key to a max key (256 bytes of 0xFF). This end key must exceed the maximum encoded key size for any table. Current max is ~60 bytes (StorageShardedKey = 20 + 32 + 8).

pub fn first<T: Table>(&self) -> ProviderResult<Option<(T::Key, T::Value)>>

Gets the first (smallest key) entry from the specified table.

pub fn last<T: Table>(&self) -> ProviderResult<Option<(T::Key, T::Value)>>

Gets the last (largest key) entry from the specified table.

pub fn iter<T: Table>(&self) -> ProviderResult<RocksDBIter<'_, T>>

Creates an iterator over all entries in the specified table.

Returns decoded (Key, Value) pairs in key order.

pub fn table_stats(&self) -> Vec<RocksDBTableStats>

Returns statistics for all column families in the database.

Returns a vector of (table_name, estimated_keys, estimated_size_bytes) tuples.

pub fn flush(&self, tables: &[&'static str]) -> ProviderResult<()>

Flushes pending writes for the specified tables to disk.

This performs a flush of:

1. The Write-Ahead Log (WAL) with sync
2. The column family memtables for the specified table names to SST files

After this call completes, all data for the specified tables is durably persisted to disk.

Panics

Panics if the provider is in read-only mode.

pub fn raw_iter<T: Table>(&self) -> ProviderResult<RocksDBRawIter<'_>>

Creates a raw iterator over all entries in the specified table.

Returns raw (key_bytes, value_bytes) pairs without decoding.

pub fn account_history_shards( &self, address: Address, ) -> ProviderResult<Vec<(ShardedKey<Address>, BlockNumberList)>>

Returns all account history shards for the given address in ascending key order.

This is used for unwind operations where we need to scan all shards for an address and potentially delete or truncate them.

pub fn storage_history_shards( &self, address: Address, storage_key: B256, ) -> ProviderResult<Vec<(StorageShardedKey, BlockNumberList)>>

Returns all storage history shards for the given (address, storage_key) pair.

Iterates through all shards in ascending highest_block_number order until a different (address, storage_key) is encountered.

pub fn unwind_account_history_indices( &self, last_indices: &[(Address, BlockNumber)], ) -> ProviderResult<WriteBatchWithTransaction<true>>

Unwinds account history indices for the given (address, block_number) pairs.

Groups addresses by their minimum block number and calls the appropriate unwind operations. For each address, keeps only blocks less than the minimum block (i.e., removes the minimum block and all higher blocks).

Returns a WriteBatchWithTransaction that can be committed later.

pub fn unwind_storage_history_indices( &self, storage_changesets: &[(Address, B256, BlockNumber)], ) -> ProviderResult<WriteBatchWithTransaction<true>>

Unwinds storage history indices for the given (address, storage_key, block_number) tuples.

Groups by (address, storage_key) and finds the minimum block number for each. For each key, keeps only blocks less than the minimum block (i.e., removes the minimum block and all higher blocks).

Returns a WriteBatchWithTransaction that can be committed later.

pub fn write_batch<F>(&self, f: F) -> ProviderResult<()>

where F: FnOnce(&mut RocksDBBatch<'_>) -> ProviderResult<()>,

Writes a batch of operations atomically.

pub fn commit_batch( &self, batch: WriteBatchWithTransaction<true>, ) -> ProviderResult<()>

Commits a raw WriteBatchWithTransaction to RocksDB.

This is used when the batch was extracted via RocksDBBatch::into_inner and needs to be committed at a later point (e.g., at provider commit time).

Panics

Panics if the provider is in read-only mode.

Trait Implementations

impl Clone for RocksDBProvider

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl DatabaseMetrics for RocksDBProvider

fn gauge_metrics(&self) -> Vec<(&'static str, f64, Vec<Label>)>

Returns a list of Gauge metrics for the database.

fn report_metrics(&self)

Reports metrics for the database.

fn counter_metrics(&self) -> Vec<(&'static str, u64, Vec<Label>)>

Returns a list of Counter metrics for the database.

fn histogram_metrics(&self) -> Vec<(&'static str, f64, Vec<Label>)>

Returns a list of Histogram metrics for the database.

impl Debug for RocksDBProvider

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

Layout

Note: Most layout information is completely unstable and may even differ between compilations. The only exception is types with certain repr(...) attributes. Please see the Rust Reference's “Type Layout” chapter for details on type layout guarantees.

Size: 8 bytes