Skip to main content

SchedulerPersistence

uni_plugin::scheduler

Trait SchedulerPersistence 

Source 
pub trait SchedulerPersistence:
    Send
    + Sync
    + Debug {
    // Required methods
    fn record_started(
        &self,
        id: &QName,
        started_at: SystemTime,
    ) -> Result<(), SchedulerPersistenceError>;
    fn record_finished(
        &self,
        id: &QName,
        finished_at: SystemTime,
        success: bool,
    ) -> Result<(), SchedulerPersistenceError>;
    fn cancel(&self, id: &QName) -> Result<(), SchedulerPersistenceError>;
    fn load_all(
        &self,
    ) -> Result<Vec<SchedulerJobRecord>, SchedulerPersistenceError>;

    // Provided methods
    fn record_scheduled(
        &self,
        _id: &QName,
        _schedule: &Schedule,
    ) -> Result<(), SchedulerPersistenceError> { ... }
    fn flush_checkpoint(&self) -> Result<(), SchedulerPersistenceError> { ... }
}
Expand description

Persistence backend for Scheduler job state.

Mirrors the meta-plugin’s Persistence trait in shape but scoped to scheduler records. The Tokio driver (crates/uni/src/scheduler.rs) invokes record_started / record_finished / cancel on each lifecycle transition; on startup the driver calls load_all and re-registers persisted jobs (followed by Scheduler::requeue_orphaned_runs for any that were Running at the previous shutdown / crash).

Two impls ship in-tree:

  • MemoryPersistence — no-op tests + as the default before the host wires a system-label backend.
  • SystemLabelPersistence (in uni-query, lands with the M9 cutover): round-trips through uni_system.background_jobs via the write-enabled QueryProcedureHost::execute_inner_query.

Required Methods§

Source

fn record_started( &self, id: &QName, started_at: SystemTime, ) -> Result<(), SchedulerPersistenceError>

Persist a job’s transition into a new run.

§Errors

Returns SchedulerPersistenceError on backend failure.

Source

fn record_finished( &self, id: &QName, finished_at: SystemTime, success: bool, ) -> Result<(), SchedulerPersistenceError>

Persist the outcome of a finished run.

§Errors

Returns SchedulerPersistenceError on backend failure.

Source

fn cancel(&self, id: &QName) -> Result<(), SchedulerPersistenceError>

Persist a cancellation.

§Errors

Returns SchedulerPersistenceError on backend failure.

Source

fn load_all(&self) -> Result<Vec<SchedulerJobRecord>, SchedulerPersistenceError>

Reload all known job records (used on host startup to restore scheduler state across restart). Order is unspecified — the driver re-registers them in any order.

§Errors

Returns SchedulerPersistenceError on backend failure.

Provided Methods§

Source

fn record_scheduled( &self, _id: &QName, _schedule: &Schedule, ) -> Result<(), SchedulerPersistenceError>

Persist a job’s schedule at registration time.

Called by the host wrapper (e.g. SchedulerHost) whenever a caller invokes add_scheduled_job, so the schedule kind (Periodic / Cron / Once / Manual) survives restart and can be round-tripped through Self::load_all. The default no-op suits in-memory backends and pre-existing impls that do not need durability.

§Errors

Returns SchedulerPersistenceError on backend failure.

Source

fn flush_checkpoint(&self) -> Result<(), SchedulerPersistenceError>

Force any in-memory buffers to durable storage.

Invoked by uni.periodic.commit so operators can drive a synchronous checkpoint flush. Backends that write through on every event (the default for the system-label backend) leave this as the default no-op; buffered backends override.

§Errors

Returns SchedulerPersistenceError on backend failure.

Dyn Compatibility§

This trait is dyn compatible.

In older versions of Rust, dyn compatibility was called "object safety".

Implementors§