Struct Lifecycle 

pub struct Lifecycle { /* private fields */ }

The registry, its templates and the processes it started.

Implementations

impl Lifecycle

pub fn open( templates: &[ServiceTemplateConfig], dir: PathBuf, network: String, placer: Option<Arc<dyn Placer>>, ) -> Result<Arc<Self>, LifecycleError>

Opens the registry in dir, reading every record once.

network is the Lys network services are placed under; the address a started process connects to is given by Self::listening.

Errors

The directory cannot be made or read, or a record in it is not one.

pub fn listening(&self, address: String)

Records the address the server listens on; a process is started only after it is known. A second call changes nothing.

pub fn list( &self, visible: impl Fn(&ServiceRecord) -> bool, ) -> Result<Value, LifecycleError>

The templates and every service visible keeps, with whether a process is running for it, as the services.list answer.

Errors

The registry lock is poisoned.

pub fn record(&self, id: &str) -> Result<ServiceRecord, LifecycleError>

The record of id.

Errors

No such service, or the lock is poisoned.

pub fn create( &self, template: &str, name: &str, holder: &str, operation: &str, now_ms: u64, ) -> Result<ServiceRecord, LifecycleError>

Creates a service named name from template for holder, asleep, after placing it and its subjects in Lys. The same operation from the same holder answers the same service and places nothing twice.

Errors

By name: the template, the name, the placement or the store.

pub fn deploy( self: &Arc<Self>, id: &str, holder: &str, ) -> Result<ServiceRecord, LifecycleError>

Starts an asleep service’s template for holder; it is recorded live when its worker registers.

Errors

By name: the service, its state, or the start.

pub fn wake( self: &Arc<Self>, id: &str, holder: &str, ) -> Result<ServiceRecord, LifecycleError>

Starts a held service’s template again for holder.

Errors

As Self::deploy.

pub fn hold( &self, id: &str, holder: &str, now_ms: u64, ) -> Result<ServiceRecord, LifecycleError>

Stops a running or live service for holder and records it held.

Errors

By name: the service, its state, the stop or the store.

pub fn retire( &self, id: &str, holder: &str, now_ms: u64, ) -> Result<ServiceRecord, LifecycleError>

Stops a service if it runs and records it retired, for good.

Errors

As Self::hold.

pub fn admit_process(&self, token: &[u8]) -> Option<ProcessGrant>

What token admits, when it is a running service process’s own.

pub fn mark_live(&self, id: &str, now_ms: u64) -> Result<(), LifecycleError>

Records id live when its worker registered, for the holder that started it; nothing when it already is.

Errors

The service is unknown or the store failed.

pub fn restart_live( self: &Arc<Self>, now_ms: u64, ) -> Vec<(String, LifecycleError)>

Starts every service last recorded live, as the holder that last made it so, after a restart. A start that fails is recorded on the service and the rest still start.

pub fn placed_since(&self, subject: &str, since_ms: u64) -> bool

Whether a service serving subject, not retired, was placed in Lys at or after since_ms: a pass issued before then cannot carry its rights, so its offline refusal says nothing and Lys is asked live.

pub fn shutdown_begun(&self)

The server’s shutdown has begun: from here no service starts, and a process that exits (a worker leaves when its connection closes) keeps its service’s recorded state, so a live service is live at the next start. Called when the stop is asked, before any connection drains.

pub fn shutdown(&self)

Stops every running process at server shutdown without changing a state, so a live service is started again by the next start.

Trait Implementations

impl Debug for Lifecycle

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

Formats the value using the given formatter.