Cuted

Cuted is a very simple, single-machine, batch scheduler written in Ruby. I often want to queue up some commands that will each take a few hours to run, but I don't want to keep track of which ones are already running and don't want them all running simultaneously. Bash jobs can't let me do both of those things.

And it's a pun on queue.

• Ruby 1.9.0+
• A few Gems:
  • Trollop
  • Parseconfig
  • Daemons
  • Prowler (optional; for Prowl notifications)

There are two commands cuted, the scheduler daemon, and cute the queueing command.

You can start the scheduler by running cuted start. This will start the scheduling daemon in the background.

Then you can queue a command to run in the current working directory with cute ANY COMMAND. By default, the command's output and error are appended to a file cuted.log in that working directory.

You can stop the schdeuler with cuted stop and you can run it in the foreground, so it shows what tasks are in progress, with cuted run.

You can configure the scheduler by setting options in ~/.cuterc file. Here are the defaults:

• max_concurrent = 4 How many commands are allowed to run at once.

• log = cuted.log Wheither to append the output and error of the command to cuted.log.

• uri = drbunix:///tmp/cuted.socket Where the server opens a pipe to recieve commands.

• prowl_key = nil What Prowl key should be notified upon command completion. If the environment variable PROWLKEY is set, that is used.

There are some command line options for cute.

• --dir Which working directory this command will run in. Defaults to the current.

• --weight How many commands does this command count for. Defaults to 1. Since the scheduler defaults to a maximum of 4 simultaneous commands, if you submit some with --weight 2 only two will run at a time.

• --log A custom log to use for just this command.

• --prowl-key A custom Prowl key to notify upon completion of just this command.

• --uri Which pipe (and server) to write commands to. This option defaults to drbunix:///tmp/cuted.sock but will use the value of the environment variable cuted_uri if set (which the scheduler will write), or will use the path passed on the command line.

This is a pretty simple beast and does not currently:

• Reorder commands so as close to max_concurrent are running simultaneously. If you queue three 1-count tasks and one 4-count task with max_concurrent = 4, the first three will run and then the 4-count task will begin once the others have completed.

• This is a single-user system; all commands are run as the user who started cuted.