Struct StaticFileProvider

Source

pub struct StaticFileProvider<N>(/* private fields */);

Available on crate feature provider only.

Expand description

StaticFileProvider manages all existing StaticFileJarProvider.

“Static files” contain immutable chain history data, such as:

transactions
headers
receipts

This provider type is responsible for reading and writing to static files.

Implementations§

Source §

impl<N> StaticFileProvider<N>
where N: NodePrimitives,

Source

pub fn read_only( path: impl AsRef<Path>, watch_directory: bool, ) -> Result<StaticFileProvider<N>, ProviderError>

Creates a new StaticFileProvider with read-only access.

Set watch_directory to true to track the most recent changes in static files. Otherwise, new data won’t be detected or queryable.

Watching is recommended if the read-only provider is used on a directory that an active node instance is modifying.

Source

pub fn read_write( path: impl AsRef<Path>, ) -> Result<StaticFileProvider<N>, ProviderError>

Creates a new StaticFileProvider with read-write access.

Source

pub fn watch_directory(&self)

Watches the directory for changes and updates the in-memory index when modifications are detected.

This may be necessary, since a non-node process that owns a StaticFileProvider does not receive update_index notifications from a node that appends/truncates data.

Source §

impl<N> StaticFileProvider<N>
where N: NodePrimitives,

Source

pub fn with_custom_blocks_per_file( self, blocks_per_file: u64, ) -> StaticFileProvider<N>

Set a custom number of blocks per file.

Source

pub fn with_metrics(self) -> StaticFileProvider<N>

Enables metrics on the StaticFileProvider.

Source

pub fn report_metrics(&self) -> Result<(), ProviderError>

Reports metrics for the static files.

Source

pub fn get_segment_provider_from_block( &self, segment: StaticFileSegment, block: u64, path: Option<&Path>, ) -> Result<StaticFileJarProvider<'_, N>, ProviderError>

Gets the StaticFileJarProvider of the requested segment and block.

Source

pub fn get_segment_provider_from_transaction( &self, segment: StaticFileSegment, tx: u64, path: Option<&Path>, ) -> Result<StaticFileJarProvider<'_, N>, ProviderError>

Gets the StaticFileJarProvider of the requested segment and transaction.

Source

pub fn get_segment_provider( &self, segment: StaticFileSegment, fn_range: impl Fn() -> Option<SegmentRangeInclusive>, path: Option<&Path>, ) -> Result<Option<StaticFileJarProvider<'_, N>>, ProviderError>

Gets the StaticFileJarProvider of the requested segment and block or transaction.

fn_range should make sure the range goes through find_fixed_range.

Source

pub fn remove_cached_provider( &self, segment: StaticFileSegment, fixed_block_range_end: u64, )

Given a segment and block range it removes the cached provider from the map.

CAUTION: cached provider should be dropped before calling this or IT WILL deadlock.

Source

pub fn delete_segment_below_block( &self, segment: StaticFileSegment, block: u64, ) -> Result<Vec<SegmentHeader>, ProviderError>

This handles history expiry by deleting all static files for the given segment below the given block.

For example if block is 1M and the blocks per file are 500K this will delete all individual files below 1M, so 0-499K and 500K-999K.

This will not delete the file that contains the block itself, because files can only be removed entirely.

§Safety

This method will never delete the highest static file for the segment, even if the requested block is higher than the highest block in static files. This ensures we always maintain at least one static file if any exist.

Returns a list of SegmentHeaders from the deleted jars.

Source

pub fn delete_jar( &self, segment: StaticFileSegment, block: u64, ) -> Result<SegmentHeader, ProviderError>

Given a segment and block, it deletes the jar and all files from the respective block range.

CAUTION: destructive. Deletes files on disk.

This will re-initialize the index after deletion, so all files are tracked.

Returns the SegmentHeader of the deleted jar.

Source

pub fn update_index( &self, segment: StaticFileSegment, segment_max_block: Option<u64>, ) -> Result<(), ProviderError>

Updates the inner transaction and block indexes alongside the internal cached providers in self.map.

Any entry higher than segment_max_block will be deleted from the previous structures.

If segment_max_block is None it means there’s no static file for this segment.

Source

pub fn initialize_index(&self) -> Result<(), ProviderError>

Initializes the inner transaction and block index

Source

pub fn check_consistency<Provider>( &self, provider: &Provider, has_receipt_pruning: bool, ) -> Result<Option<PipelineTarget>, ProviderError>
where Provider: DBProvider + BlockReader + StageCheckpointReader + ChainSpecProvider, N: NodePrimitives, <N as NodePrimitives>::Receipt: Value, <N as NodePrimitives>::BlockHeader: Value, <N as NodePrimitives>::SignedTx: Value,

Ensures that any broken invariants which cannot be healed on the spot return a pipeline target to unwind to.

Two types of consistency checks are done for:

When a static file fails to commit but the underlying data was changed.
When a static file was committed, but the required database transaction was not.

For 1) it can self-heal if self.access.is_read_only() is set to false. Otherwise, it will return an error. For 2) the invariants below are checked, and if broken, might require a pipeline unwind to heal.

For each static file segment:

the corresponding database table should overlap or have continuity in their keys (TxNumber or BlockNumber).
its highest block should match the stage checkpoint block number if it’s equal or higher than the corresponding database table last entry.

Returns a Option of [PipelineTarget::Unwind] if any healing is further required.

WARNING: No static file writer should be held before calling this function, otherwise it will deadlock.

Source

pub fn check_segment_consistency( &self, segment: StaticFileSegment, ) -> Result<(), ProviderError>

Checks consistency of the latest static file segment and throws an error if at fault. Read-only.

Source

pub fn earliest_history_height(&self) -> u64

Returns the earliest available block number that has not been expired and is still available.

This means that the highest expired block (or expired block height) is earliest_history_height.saturating_sub(1).

Returns 0 if no history has been expired.

Source