Struct CheckpointService

pub struct CheckpointService { /* private fields */ }

Implementations

impl CheckpointService

pub fn build( state: Arc<AuthorityState>, checkpoint_store: Arc<CheckpointStore>, epoch_store: Arc<AuthorityPerEpochStore>, effects_store: Arc<dyn TransactionCacheRead>, global_state_hasher: Weak<GlobalStateHasher>, checkpoint_output: Box<dyn CheckpointOutput>, certified_checkpoint_output: Box<dyn CertifiedCheckpointOutput>, metrics: Arc<CheckpointMetrics>, max_transactions_per_checkpoint: usize, max_checkpoint_size_bytes: usize, ) -> Arc<Self>

Constructs a new CheckpointService in an un-started state.

pub async fn spawn( &self, epoch_store: Arc<AuthorityPerEpochStore>, consensus_replay_waiter: Option<ReplayWaiter>, )

Starts the CheckpointService.

This function blocks until the CheckpointBuilder re-builds all checkpoints that had been built before the most recent restart. You can think of this as a WAL replay operation. Upon startup, we may have a number of consensus commits and resulting checkpoints that were built but not committed to disk. We want to reprocess the commits and rebuild the checkpoints before starting normal operation.

impl CheckpointService

pub async fn wait_for_rebuilt_checkpoints(&self)

Waits until all checkpoints had been built before the node restarted are rebuilt. This is required to preserve the invariant that all checkpoints (and their transactions) below the highest_synced_checkpoint watermark are available. Once the checkpoints are constructed, we can be sure that the transactions have also been executed.

Trait Implementations

impl CheckpointServiceNotify for CheckpointService

fn notify_checkpoint_signature( &self, epoch_store: &AuthorityPerEpochStore, info: &CheckpointSignatureMessage, ) -> SuiResult

fn notify_checkpoint(&self) -> SuiResult