Struct Config 

pub struct Config { /* private fields */ }

Configuration for the native io_uring runtime.

A pool of Self::with_worker_threads workers polls ordinary tasks, the first on the thread calling crate::Runner::start. Tasks marked dedicated or blocking each receive a fresh thread and io_uring ring. Shutdown completes retained writes and syncs without a configured time limit.

Implementations

impl Config

pub fn new() -> Self

Return the default runtime configuration with a generated temporary directory.

pub const fn with_worker_threads(self, workers: usize) -> Self

Set the number of pool workers, including the calling thread. Defaults to 2.

Must be between 1 and 64. Each worker has its own io_uring ring. Dedicated and blocking threads are additional to this count.

pub const fn with_global_queue_interval(self, interval: u32) -> Self

Set how often a busy pool worker looks at the global queue first: every intervalth poll. Defaults to 31.

Must be nonzero. A smaller interval starts tasks in the global queue sooner at the cost of more synchronization. At 1, a busy worker checks the global queue before every poll and polls its own queued tasks only while the global queue is empty.

pub const fn with_ring_config(self, config: RingConfig) -> Self

Set io_uring ring capacity and deadline granularity. See RingConfig.

pub const fn with_idle_spinner(self, config: SpinnerConfig) -> Self

Set idle spinning policy. Use SpinnerConfig::disabled to disable it.

pub const fn with_thread_stack_size(self, size: usize) -> Self

Set the stack size of pool, dedicated worker, and Rayon threads. Worker zero runs on the calling thread and keeps its stack.

pub const fn with_catch_panics(self, catch: bool) -> Self

Set whether spawned-task panics and dedicated worker failures are caught.

Caught failures are logged without interrupting the root. Propagated failures are observed only while the root is executing. A failure of a pool worker itself, such as an io_uring error, fails the runner with either setting. It interrupts the root while the root runs, and otherwise the runner resumes it after shutdown.

A worker that fails to start, for example because its io_uring ring cannot be created, reports the error through the panic hook. With panic = "abort", such a failure aborts the process with either setting, like a failure to spawn the worker’s thread.

During worker execution, cancellation-time and task-disposal panics escaping the user-poll wrapper are contained with either setting. An unpublished result may resolve to Error::Closed, while an already-published result remains available. Task factories execute synchronously and propagate panics to their caller.

pub const fn with_connect_timeout(self, timeout: Duration) -> Self

Set the outbound connection timeout, including queueing and retries.

Must be nonzero and no greater than 30 years. The maximum configured network timeout must fit the wheel slot limit in RingConfig.

pub const fn with_read_write_timeout(self, timeout: Duration) -> Self

Set the send and receive timeout, including queueing and partial progress.

Must be nonzero and no greater than 30 years. The maximum configured network timeout must fit the wheel slot limit in RingConfig.

pub const fn with_tcp_nodelay(self, enabled: Option<bool>) -> Self

Set the best-effort TCP_NODELAY override. None leaves the system default.

pub const fn with_zero_linger(self, enabled: bool) -> Self

Set whether sockets request zero linger when configured.

pub const fn with_read_buffer_size(self, size: usize) -> Self

Set the network receive buffer size in bytes.

pub fn with_storage_directory(self, directory: impl Into<PathBuf>) -> Self

Set the storage directory, created and held when the runner starts.

pub fn with_storage_blob_layouts( self, layouts: RangeInclusive<BlobLayout>, ) -> Self

Set the blob layouts accepted by storage.

New blobs use the latest layout in this range. Existing blobs outside it fail to open with crate::Error::BlobLayoutMismatch. Restrict the range to what a rollback target can read before first opening storage.

Panics

Panics if layouts is empty.

pub const fn worker_threads(&self) -> usize

Return the number of pool workers, including the calling thread.

pub const fn global_queue_interval(&self) -> u32

Return how often a busy pool worker looks at the global queue first: every Nth poll.

pub const fn ring_config(&self) -> &RingConfig

Return per-worker io_uring ring configuration.

pub const fn idle_spinner(&self) -> &SpinnerConfig

Return per-worker idle spinning configuration.

pub const fn thread_stack_size(&self) -> usize

Return the configured pool, dedicated worker, and Rayon thread stack size.

pub const fn catch_panics(&self) -> bool

Return whether spawned-task panics and dedicated worker failures are caught.

pub const fn connect_timeout(&self) -> Duration

Return the whole-call outbound connection timeout.

pub const fn read_write_timeout(&self) -> Duration

Return the whole-call send and receive timeout.

pub const fn tcp_nodelay(&self) -> Option<bool>

Return the TCP_NODELAY override.

pub const fn zero_linger(&self) -> bool

Return whether sockets request zero linger.

pub const fn read_buffer_size(&self) -> usize

Return the network receive buffer size.

pub const fn storage_directory(&self) -> &PathBuf

Return the storage directory retained by the runtime’s storage resources.

pub const fn storage_blob_layouts(&self) -> &RangeInclusive<BlobLayout>

Return the accepted blob layout range.

Trait Implementations

impl Clone for Config

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Default for Config

fn default() -> Self

Returns the “default value” for a type.