Version 1.3.1

Changelog

• Renamed smart_dispatch.py to smart-dispatch
• More robust file locking
• Added the ability to add raw PBS flags with --pbsFlags
• Auto detect most recent RAP id on Helios
• Logs are now appended on resume instead of overritten
• Allow resuming from path instead of simply UID
• Allow to add workers to a running batch with resume --expandPool
• Keep a history of the command lines used in command_line.log
• Added support for the Hades cluster

Bug fixes

• Now enforce python2 instead of crashing. (Plan to support python3)
• Non-deterministic behavior with resources ordering
• Updated the way Helios manage queues

Version 1.2

Changelog

• Always use worker.
• Detect command failure and move them to the failed_commands file.
• Resume warns if there is failed commands in the batch.
• Logs are now organized in sub folder.
• The batch folder name now start with the date.
• Nicer output when a job is launched.

Bug fixes

• Failed when generating a file name with Unicode character.
• The -f option was assuming a relative path.
• The file lock was not working on NFS.

Version 1.1

The SMART way of launching jobs on HPC.

Now supporting: Mammouth, Guillimin and Helios.

Includes an improved command parsing, pool of workers, predefined/customizable queues and more!