Struct GroupCommit 

pub struct GroupCommit { /* private fields */ }

Group-commit coordinator (spec §9.3). The commit sequencer appends a txn’s records under the WAL mutex but does not fsync there; instead each committer calls Self::await_durable with its commit record’s WAL seq. Exactly one waiter becomes the leader and issues a single group_sync (fsync), which makes durable every record appended up to that point; the others are followers that simply wait until durable_seq reaches their commit seq. One fsync therefore covers a whole batch of concurrent commits.

Implementations

impl GroupCommit

pub fn new(durable_seq: u64) -> Self

pub fn await_durable( &self, wal: &PlMutex<SharedWal>, commit_seq: u64, ) -> Result<()>

Block until commit_seq is durable. The first eligible caller fsyncs on behalf of the batch; the rest wait on the condvar. On fsync error the coordinator is poisoned and every waiter (current and future) returns Err (spec §9.3e). wal is the same SharedWal the sequencer appended to — locked here only for the brief fsync, never across the wait.