Struct mz_storage_client::controller::persist_handles::PersistReadWorker

pub struct PersistReadWorker<T: Timestamp + Lattice + Codec64> {
    tx: UnboundedSender<(Span, PersistReadWorkerCmd<T>)>,
}

A wrapper that holds on to backing persist shards/collections that the storage controller is aware of. The handles hold back the since frontier and we need to downgrade them when the read capabilities change.

Internally, this has an async task and the methods for registering a handle and downgrading sinces add commands to a queue that this task is working off. This makes the methods non-blocking and moves the work outside the main coordinator task, meaning the coordinator is spending less time waiting on persist calls.

Fields

tx: UnboundedSender<(Span, PersistReadWorkerCmd<T>)>

Implementations

impl<T: Timestamp + Lattice + Codec64> PersistReadWorker<T>

pub(crate) fn new() -> Self

pub(crate) fn register(
    &self,
    id: GlobalId,
    since_handle: SinceHandle<SourceData, (), T, Diff, PersistEpoch>
)

pub(crate) fn update(
    &self,
    id: GlobalId,
    since_handle: SinceHandle<SourceData, (), T, Diff, PersistEpoch>
)

Update the existing since handle associated with id to since_handle.

Note that this should only be called when updating a since handle; to initially associate an id to a since handle, use Self::register.

Panics

• If id is not currently associated with any since handle.

pub(crate) fn downgrade(&self, frontiers: BTreeMap<GlobalId, Antichain<T>>)

fn send(&self, cmd: PersistReadWorkerCmd<T>)

Trait Implementations

impl<T: Debug + Timestamp + Lattice + Codec64> Debug for PersistReadWorker<T>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.