Struct GroupCommitMaster 

pub struct GroupCommitMaster { /* private fields */ }

Group-commit implementation for the Master role.

When a transaction arrives it is added to a pending queue. A leader thread waits for up to interval_ms milliseconds (or until max_count transactions are queued) before issuing a single fsync. After the fsync the queued transactions are acknowledged.

Threshold semantics

buffer_commit() returns false (caller must fsync) every max_count calls — the count-based threshold. The caller (Txn::commit_with_durability) treats a false return as a signal to call LogManager::flush_sync(), which then handles the time-based coalescing via FSyncManager. This correctly separates concerns: GroupCommit enforces the batch-size policy; FSyncManager enforces the time-window and leader/waiter coalescing.

Implementations

impl GroupCommitMaster

pub fn new(max_count: usize, interval_ms: u64) -> Self

Creates a new GroupCommitMaster.

Arguments

• max_count — maximum batch size.
• interval_ms — batch window in milliseconds MASTER_GROUP_COMMIT_INTERVAL).

pub fn interval_ms(&self) -> u64

Returns the batch window in milliseconds.

pub fn flush_count(&self) -> usize

Returns the number of times the count threshold has fired.

Used in tests to verify durability threshold enforcement.

Trait Implementations

impl Default for GroupCommitMaster

fn default() -> Self

Returns the “default value” for a type.

impl GroupCommit for GroupCommitMaster

fn buffer_commit(&self, _commit_vlsn: i64) -> bool

Buffer a commit and enforce the count-based threshold.

Returns false (caller must fsync) on every max_countth call. Returns true (commit is buffered, skip fsync) otherwise.

Count-threshold path. The time-window threshold is handled by FSyncManager when the caller proceeds to LogManager::flush_sync() on a false return.

fn is_enabled(&self) -> bool

Returns true if group commit is currently enabled.

fn shutdown(&self)

Shuts down the group-commit background machinery.