Struct mz_persist_client::async_runtime::IsolatedRuntime

pub struct IsolatedRuntime {
    inner: Option<Runtime>,
}

An isolated runtime for asynchronous tasks, particularly work that may be CPU intensive such as encoding/decoding and shard maintenance.

Using a separate runtime allows Persist to isolate its expensive workloads on its own OS threads as an insurance policy against tasks that erroneously fail to yield for a long time. By using separate OS threads, the scheduler is able to context switch out of any problematic tasks, preserving liveness for the rest of the process.

Note: Even though the work done by this runtime might be “blocking” or CPU bound we should not use the tokio::task::spawn_blocking API. There can be issues during shutdown if tasks are currently running on the blocking thread pool 1, and the blocking thread pool generally creates many more threads than are physically available. This can pin CPU usage to 100% starving other important threads like the Coordinator.

Fields

inner: Option<Runtime>

Implementations

impl IsolatedRuntime

pub fn new(worker_threads: usize) -> IsolatedRuntime

Creates a new isolated runtime.

pub fn spawn_named<N, S, F>(&self, name: N, fut: F) -> JoinHandle<F::Output>

where S: AsRef<str>, N: FnOnce() -> S, F: Future + Send + 'static, F::Output: Send + 'static,

Spawns a task onto this runtime.

Note: We purposefully do not use the tokio::task::spawn_blocking API here, see the doc comment on IsolatedRuntime for explanation.

Trait Implementations

impl Debug for IsolatedRuntime

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

Formats the value using the given formatter.

impl Default for IsolatedRuntime

fn default() -> Self

Returns the “default value” for a type.

impl Drop for IsolatedRuntime

fn drop(&mut self)

Executes the destructor for this type.