fastcpu

Inspired with code from https://github.com/fastai/fastgpu

fastcpu

A queue service for quickly developing scripts that use all your CPUs efficiently

PULL REQUESTS WELCOME!

fastcpu provides a single command, fastcpu_poll, which polls a directory to check for scripts to run, and then runs them on the first available CPU. If no CPUs are available, it waits until one is. If more than one CPU is available, multiple scripts are run in parallel, one per CPU. (Note currently the CPU load checking is not implemented, the scripts are run sequentially at polling interval)

Installation

pip install fastcpu

How to use

--help provides command help:

$ fastgpu_poll --help

optional arguments:

  -h, --help                         show this help message and exit
  --path PATH                        Path containing `to_run` directory (default: .)
  --exit_when_empty EXIT_WHEN_EMPTY  Exit when `to_run` is empty (default: 1)
  --poll_interval POLL_INTERVAL      The duration between polls (default: 0.1)
  --terminate_timout                 The timeout at which point we kill the running script in seconds, -1 to never terminate

If installed via pip there is a handy command line method available

fastcpu_poll --path /path/to/scripts --exit_when_empty 0 --poll_interval 60

If running as a module

python -m fastcpu.cli --path /path/to/scripts --exit_when_empty 0 --poll_interval 60

The above examples will run scrips located in the to_run subdirectory of the directory being monitored The program will not exit when there are no scripts left to run, it will keep polling since we set that to 0 the polling interval is 60 seconds, it can be set as fractions of a second e.g 0.1

once the program starts it creates the following directory structure. you can then your scripts in the to_run folder, and the scrips are run sequentially

.
├── complete
├── fail
├── out
├── running
└── to_run
    ├── script_example1.sh
    └── script_example2.sh

fastcpu_poll will run each script in to_run in sorted order. Each script will be assigned to one CPU (future)

Once a script is selected to be run, it is moved into a directory called running. Once it's finished, it's moved into complete or fail as appropriate. stdout and stderr are captured to files with the same name as the script, plus stdout or stderr appended.

If exit_when_empty is 1 (which is the default), then once all scripts are run, fastcpu_poll will exit. If it is 0 then fastcpu_poll will continue running until it is killed; it will keep polling for any new scripts that are added to to_run.

Development

local testing

python -m fastcpu.cli --path test_scripts --exit_when_empty 1 --poll_interval 5 --terminate_timeout 12

Troubleshooting

If you get permission denied error in stderr logs check that your scripts are executable