Struct PeerLogScanner 

pub struct PeerLogScanner { /* private fields */ }

A LogScanner backed by an in-memory queue of (vlsn, type, payload) entries.

Entries are pushed by the ReplicaReceiver as they arrive from the master (or from another peer). The PeerFeederRunner consumes entries from this queue and streams them to a downstream replica.

Bounded memory (F10)

The queue has two configurable bounds:

• max_entries: maximum entry count (default DEFAULT_PEER_SCANNER_MAX_ENTRIES).
• max_bytes: maximum cumulative payload size in bytes (default DEFAULT_PEER_SCANNER_MAX_BYTES).

On push, if either bound is exceeded the oldest entries are evicted from the front of the queue until both bounds are satisfied. The evicted entries are no longer available for peer streaming through this scanner; downstream peers that fall behind the eviction window must catch up via the on-disk EnvironmentLogScanner or via network restore. This matches HA semantics where peer-to-peer log distribution is best-effort and the on-disk log is the durable source.

Closes finding F10 of docs/src/internal/api-audit-2026-05-rep.md.

Thread safety: the queue is protected by a Mutex so that the receiver thread (writer) and the feeder thread (reader) can operate concurrently.

Implementations

impl PeerLogScanner

pub fn new() -> Self

Create an empty scanner with the default F10 bounds.

pub fn with_capacity(max_entries: usize, max_bytes: usize) -> Self

Create an empty scanner with explicit bounds.

max_entries and max_bytes are both honoured; whichever is breached first triggers oldest-evicting on subsequent push calls. Passing usize::MAX disables the corresponding bound (not recommended in production).

pub fn push(&self, vlsn: u64, entry_type: u8, payload: Vec<u8>)

Push a log entry into the scanner’s queue.

Called by the ReplicaReceiver each time an entry is applied. Entries are expected to be pushed in VLSN order, but this method is not enforcing: every entry is appended to the queue unconditionally and the cached (first_vlsn, last_vlsn) range is widened to cover the new VLSN. Out-of-order or duplicate entries are filtered later by LogScanner::next_entry, which skips entries with vlsn < from_vlsn.

F10 bound: after the new entry is appended, if the queue exceeds either max_entries or max_bytes, the oldest entries are evicted from the front until both bounds are satisfied. The retained first_vlsn is updated to the new front-of-queue VLSN so downstream peers that ask for an evicted VLSN range observe log_range().first > from_vlsn and know they must catch up via the durable log.

pub fn evicted_count(&self) -> u64

Cumulative number of entries dropped by the F10 bound since scanner construction. Useful for monitoring whether downstream peers are keeping up.

pub fn current_bytes(&self) -> usize

Current cumulative payload size in bytes (live snapshot).

pub fn log_range(&self) -> Option<(u64, u64)>

Return the VLSN range currently held in this scanner.

Returns None if the scanner is empty (no entries pushed yet).

pub fn len(&self) -> usize

Return the number of entries currently queued.

pub fn is_empty(&self) -> bool

Returns true if no entries are queued.

Trait Implementations

impl Default for PeerLogScanner

fn default() -> Self

Returns the “default value” for a type.

impl LogScanner for PeerLogScanner

fn next_entry(&mut self, from_vlsn: u64) -> Option<(u64, u8, Vec<u8>)>

Return the next available entry with VLSN >= from_vlsn, or None if no new entry is available at this moment.