Create an abstract launcher.

This function is useful for inheriting argument documentation in functions that create custom third-party launchers. See @inheritParams crew::crew_launcher in the source code file of crew_launcher_local().

Usage

crew_launcher(
  name = NULL,
  workers = 1L,
  seconds_interval = 0.25,
  seconds_timeout = 60,
  seconds_launch = 30,
  seconds_idle = 300,
  seconds_wall = Inf,
  seconds_exit = NULL,
  tasks_max = Inf,
  tasks_timers = 0L,
  reset_globals = NULL,
  reset_packages = NULL,
  reset_options = NULL,
  garbage_collection = NULL,
  crashes_error = NULL,
  launch_max = NULL,
  tls = crew::crew_tls(),
  processes = NULL,
  r_arguments = c("--no-save", "--no-restore"),
  options_metrics = crew::crew_options_metrics()
)

Arguments

name: Character string, name of the launcher. If the name is NULL, then a name is automatically generated when the launcher starts.
workers: Maximum number of workers to run concurrently when auto-scaling, excluding task retries and manual calls to launch(). Special workers allocated for task retries do not count towards this limit, so the number of workers running at a given time may exceed this maximum. A smaller number of workers may run if the number of executing tasks is smaller than the supplied value of the workers argument.
seconds_interval: Number of seconds between polling intervals waiting for certain internal synchronous operations to complete. In certain cases, exponential backoff is used with this argument passed to seconds_max in a crew_throttle() object.
seconds_timeout: Number of seconds until timing out while waiting for certain synchronous operations to complete, such as checking mirai::info().
seconds_launch: Seconds of startup time to allow. A worker is unconditionally assumed to be alive from the moment of its launch until seconds_launch seconds later. After seconds_launch seconds, the worker is only considered alive if it is actively connected to its assign websocket.
seconds_idle: Maximum number of seconds that a worker can idle since the completion of the last task. If exceeded, the worker exits. But the timer does not launch until tasks_timers tasks have completed. See the idletime argument of mirai::daemon(). crew does not excel with perfectly transient workers because it does not micromanage the assignment of tasks to workers, so please allow enough idle time for a new worker to be delegated a new task.
seconds_wall: Soft wall time in seconds. The timer does not launch until tasks_timers tasks have completed. See the walltime argument of mirai::daemon().
seconds_exit: Deprecated on 2023-09-21 in version 0.5.0.9002. No longer necessary.
tasks_max: Maximum number of tasks that a worker will do before exiting. Also determines how often the controller auto-scales. See the Auto-scaling section for details.
tasks_timers: Number of tasks to do before activating the timers for seconds_idle and seconds_wall. See the timerstart argument of mirai::daemon().
reset_globals: Deprecated on 2025-05-30 (crew version 1.1.2.9004). Please use the reset_globals option of crew_controller() instead.
reset_packages: Deprecated on 2025-05-30 (crew version 1.1.2.9004). Please use the reset_packages option of crew_controller() instead.
reset_options: Deprecated on 2025-05-30 (crew version 1.1.2.9004). Please use the reset_options option of crew_controller() instead.
garbage_collection: Deprecated on 2025-05-30 (crew version 1.1.2.9004). Please use the garbage_collection option of crew_controller() instead.
crashes_error: Deprecated on 2025-01-13 (crew version 0.10.2.9002).
launch_max: Deprecated on 2024-11-04 (crew version 0.10.2.9002).
tls: A TLS configuration object from crew_tls().
processes: Deprecated on 2025-08-27 (crew version 1.2.1.9009).
r_arguments: Optional character vector of command line arguments to pass to Rscript (non-Windows) or Rscript.exe (Windows) when starting a worker. Example: r_arguments = c("--vanilla", "--max-connections=32").
options_metrics: Either NULL to opt out of resource metric logging for workers, or an object from crew_options_metrics() to enable and configure resource metric logging for workers. For resource logging to run, the autometric R package version 0.1.0 or higher must be installed.

Auto-scaling

crew launchers implement auto-scaling in the scale() method. When the task load increases, the number of workers increases in response to demand. When the task load decreases, the workers start to exit. This behavior happens dynamically over the course of a workflow, and it can be tuned with arguments seconds_interval, seconds_wall, and tasks_max.

tasks_max is special: it determines not only the number of tasks a worker runs before exiting, it also determines how often auto-scaling runs. If tasks_max is finite, then crew uses an aggressive deterministic exponential backoff algorithm to determine how frequently to auto-scale (see crew_throttle()). But if tasks_max is Inf, then crew only scales at equally-spaced time intervals of seconds_interval to allow enough pending tasks to accumulate for job arrays. This last part is important because auto-scaling too frequently could lead to hundreds of separate job arrays with only job per array (as opposed to the desired outcome of 1 or 2 arrays with many jobs each).

Examples

if (identical(Sys.getenv("CREW_EXAMPLES"), "true")) {
client <- crew_client()
client$start()
launcher <- crew_launcher_local()
launcher$start(url = client$url, profile = client$profile)
launcher$launch()
task <- mirai::mirai("result", .compute = client$profile)
mirai::call_mirai(task)
task$data
client$terminate()
}

Usage

Arguments

Auto-scaling

See also

Examples