package current_ocluster

You can search for identifiers within the package.

in-package search v0.2.0

current_ocluster
- Current_ocluster
  - Connection
  - Raw

Legend:
Library
Module
Module type
Parameter
Class
Class type

A connection to the OCluster scheduler. Provides automatic reconnection on failure, queuing with rate limiting, metrics reporting, etc. This can be passed to Current_ocluster.v for simple cases, or it can used directly to implement custom pipeline steps if that isn't flexible enough.

type t

val create : 
  ?max_pipeline:int ->
  Cluster_api.Raw.Client.Submission.t Capnp_rpc_lwt.Sturdy_ref.t ->
  t

create submission_service is a connection that submits jobs to submission_service.

parameter max_pipeline
: how many items to queue up at the scheduler per (pool, urgency).

val pool : 
  job:Current.Job.t ->
  pool:string ->
  action:Cluster_api.Submission.action ->
  cache_hint:string ->
  ?src:(string * string list) ->
  ?urgent:([ `High | `Low ] -> bool) ->
  t ->
  Cluster_api.Raw.Client.Job.t Capnp_rpc_lwt.Capability.t Current.Pool.t

pool ~job ~pool ~action ~cache_hint t is a resource pool, suitable for passing to Current.Job.start_with. Submits action to the pool named pool at t. If t is disconnected, it will keep trying to reconnect until it succeeds.

parameter job
Used to write log messages.

parameter cache_hint
Hint to the scheduler (similar jobs should be assigned to the same machine, if possible).

parameter urgent
Used to calculate the job's urgency, possibly using the OCurrent priority. The default is to mark high priority jobs as urgent.

val run_job : 
  job:Current.Job.t ->
  Cluster_api.Raw.Client.Job.t Capnp_rpc_lwt.Capability.t ->
  (string, [> `Msg of string ]) result Lwt.t

run_job ~job ocluster_job tails the log of ocluster_job to job and then returns the job's result.