NewStatefulInterceptor returns an initialized Interceptor.

SetBatch stores multiple index→position mappings.

AppendColumnTxn appends the Columns update type Txn to wal for the provided rowKey. Update/DeleteKV Ops for column is decided by the Log Operation typ

DeleteRange deletes logs in [min, max] inclusive. Supports two scenarios required by Raft: 1. PREFIX deletion: Log compaction after snapshot (min=fir

FlushedIndex returns the last Raft log index flushed to B-tree.

GetBounds returns the first and last available log indices. Returns (0, 0) when no indexed entries exist.

LeaderWithID returns the address and ID of the current leader.

NewGrpcStreamer returns an initialized GrpcStreamer that implements grpc-based WalStreamerService.

NewOutboundConn creates a new OutboundConn that will write the namespace header on the first Write() call.

NewReplicator returns an initialized Replicator that uses LSN-based replication.

RegisterCluster adds a ClusterDiscover listener that will receive membership events. This allows multiple Raft clusters (one per namespace) to react t

Restore performs a full restore operation based on the provided options. It holds an exclusive lock during the entire operation to prevent the server

WithClearIndexOnFlush enables clearing segment's in-memory index after it's flushed to disk. This is useful when an external index is maintained.

AutoScope returns a tally.Scope scoped by the caller's package name.

BatchAppend writes multiple records to the active WAL segment in a single batch operation. This is more efficient than calling Append multiple times a

DeleteRange deletes a range of log entries, but limits the max to the safe-to-GC index to preserve logs needed for pending segment processing.

DeserializeRowUpdateEntry decodes a FlatBuffer-encoded RowUpdateEntry.

Encode serializes a raft.Log into a pre-sized buffer.

GetLatestLSN reads the latest committed LSN from partitionlog catalog head.

IsCurrentEntry checks if the given position is the current one for an index. Returns true if the index exists and the position matches exactly.

MarkSegmentsForDeletion identifies and queues WAL segments for deletion based on their age and segment count retention constraints.

NewReaderWithStart returns a new Reader that begins reading from the specified position. If SegmentID is 0, the reader will begin from the very start

NewSafeGCLogStore creates a new SafeGCLogStore wrapping the given LogStore. The provider supplies the safe-to-GC index (typically FlushedIndexHolder o

ParseKeyKind returns the KeyKind associated with the key.

ReadHeader reads the namespace header from r. Returns the namespace string or an error.

Set stores the safe-to-GC index.

StreamNamespace starts replicating WAL records for the given namespace into object storage. It blocks until the context is cancelled or the streamer i

Format: [RecordProcessed:8][Pos:12][RaftIndex:8][RaftTerm:8] = 36 bytes.

WithAutoCleanupPolicy configures the automatic segment cleanup policy for the WAL. maxAge: Segments older than this duration are eligible for deletion

Commit advances the committed position to the given position. When readerCommitCheck is enabled, readers cannot advance beyond this position. In Raft

DefaultBlobStoreStreamerConfig returns defaults for partitionlog-backed blob streaming.

Enqueue adds an inbound connection to the accept queue. Returns false if the stream layer is closed or the queue is full.

HasActiveReaders returns true if there are any currently active readers on the segment.

NewRateLimitedWalIO constructs a RateLimitedWalIO.

NewTransaction returns a transaction implementation based on the current engine mode. In Raft mode, it returns a Raft-backed transaction; otherwise it

ParseKeyKind returns the KeyKind associated with the key.

Range iterates over all entries and calls fn for each. If fn returns false, iteration stops.

Replicate reads wal record from the underlying engine, and sends the WalRecords when batchSize/maxBatchDuration is reached.

SetWALCommitCallback sets the callback that is invoked when a committed Raft entry is applied. The callback receives the WAL position of the committed

WithBytesPerSync sets the threshold in bytes after which a msync is triggered. Useful for batching writes. 0 disable this feature.

persistKeyValue writes a key-value pair to WAL and MemTable, ensuring durability. multistep process to persist the key-value pair: 1. Encodes the reco

BackupSegmentsAfter copies every sealed segment with ID > afterID into backupDir. Returns a map of SegmentID to the backup file path.