Two-phase graceful shutdown manager for coroutine workloads: block new jobs, allow cooldown, then cancel running jobs with configurable cooldown and cancellation timeouts; stop() signals timely completion.
This library provides a single class called GracefulScope. It supports running
coroutine jobs like any CoroutineScope, but it also has a support of graceful
stop. Graceful stop happens in two phases:
First phase: Forbids new jobs to be started on this scope. However,
new jobs still can be started inside already running jobs attached to this
scope. This phase ends after cooldownTimeout or when all children are
completed.
Second phase: Starts cancellation of existing jobs using coroutines
cancellation mechanism. This phase ends after cancellationTimeout or when
all children are cancelled.
Then, stop() function returns true or false depending on whether it could stop
running jobs in cancellationTimeout. If it returns false, cancellation process
never stops, but control flow is returned back to the user of API.
Replace $version with the latest version from Releases Tab.
dependencies {
implementation("me.y9san9.graceful:graceful-scope:$version")
}
See this example to play around with Graceful Scope.
Usual pattern of using such a scope is together with Shutdown Hook:
gracefulScope { notificationsScope ->
runNotificationsActor(notificationsScope)
Runtime.getRuntime().addShutdownHook {
notificationsScope.stop(
cooldownTimeout = 15.seconds,
cancellationTimeout = null,
)
}
}
This will ensure that when you restart service, notifications will have a
15-seconds chance to finish currently running jobs, but no new jobs will be
accepted. It provides the same functionality that server.stop() from ktor
does, but in a more general way.
Surfaced from shared tags and platforms — no rankings paid for.