/trapperkeeper-scheduler

Primary LanguageClojure

Trapperkeeper Scheduler Service

Build Status

Clojars Project

A Trapperkeeper service that provides a simple API for scheduling background tasks.

Other Trapperkeeper services may specify a dependency on the Scheduler service, and then use its functions to schedule and cancel jobs to be run on background worker threads.

What Does This Service Do?

The SchedulerService provides some simple API for scheduling (potentially recurring) background tasks. The service manages the lifecycle of the underlying scheduling subsystem so that other services don't need to (and avoids potential issues around multiple services attempting to initialize the same scheduling subystem in a single JVM process).

This is a very early release, and provides only the most basic API. The functions that are currently available are as follows:

  • interspaced [interval-ms f]: schedules a job that will call f, block until that call completes, sleep for interval-ms milliseconds, and then repeat. Returns an identifier that can be used to reference this scheduled job (e.g., for cancellation) later.
  • after [interval-ms f]: schedules a job that will call f a single time, after a delay of interval-ms milliseconds. Returns an identifier that can be used to reference this scheduled job (e.g. for cancellation) later.
  • stop-job [job-id]: Given a job-id returned by one of the previous functions, cancels the job. If the job is currently executing it will be allowed to complete, but will not be invoked again afterward. Returns true if the job was successfully stopped, false otherwise.

Implementation Details

The current implementation of the SchedulerService is a fairly thin wrapper around the overtone/at-at library. This approach was chosen for a couple of reasons:

  • A few existing PL projects already use this library, so we thought it'd be better not to introduce a different scheduling subsystem until all of the PL TK projects are aligned to use the SchedulerService.
  • The at-at API seems like a pretty reasonable Clojure API for scheduling tasks.

It would probably be a good idea to switch the implementation out to use a different backend in the future; without having done too much investigation yet, I'd be interested in looking into Quartz/Quartzite, simply because Quartz is very widely-used and battle-tested in the Java world. at-at does not appear to be being maintained any longer. We've had a few minor issues with it (especially w/rt shutting down the system), and haven't gotten any responses from the maintainer on the github issues we've opened. Also, the source repo for at-at contains no tests :(

What's Next?

  • Add additional scheduling API functions, e.g. every.
  • Add API for introspecting the state of currently scheduled jobs
  • Consider porting the backend to something more sophisticated (and maintained) than at-at; if we do this, the intent would be to maintain the existing API.

License

Copyright © 2015 Puppet Labs