Fizzled firework - dynamically re-run

Zachary_Ball · March 10, 2016, 6:14am

Hi

Is there a way that based upon a fizzled firework one can instruct the firework to re-run itself? I could always dynamically write cases for a firework to re-run itself like so:

def rerun_firework(current_fw, fw_spec, name):

fw_spec[‘fail_count’] += 1

max_retries = fw_spec[‘max_retries’]

fail_count = fw_spec[‘fail_count’]

if fail_count == max_retries:

Could not succeed in passing this firework with retries

return FWAction(defuse_workflow=True)

failed_workflow = Workflow.from_Firework(Firework(current_fw, spec=fw_spec, name=name))

return FWAction(detours=[failed_workflow])

Although I am not quite sure that this is the best way of achieving what I am trying to achieve?

Basically I am looking for a way for a Firework to fizzle, as caused by an exception or failure, and then for the firework to retry itself a certain amount of times with changed parameters.

Thanks again!

Anubhav_Jain · March 10, 2016, 6:53pm

Hi Zachary,

Currently, you must rerun fizzled FireWorks either through the command line (e.g., by typing the lpad rerun_fws command) or through a Python script (e.g., by programming some conditions to detect fizzled FireWorks that need a rerun and rerunning those through the LaunchPad rerun function). But, neither of those methods will add in the condition of “max_retries” like you stated.

So, basically the solution you wrote is the correct way forward. If you are feeling ambitious, you can write it into Rocket.run() so that the “_max_retries” is a reserved FW spec keyword, and that any Firework with that keyword will have a FIZZLED FWAction modified to add in a detour of that Firework with the number of retries incremented. I could then merge those commits into the main FWS software (but that is up to you).

Good luck!

Anubhav

···

On Wed, Mar 9, 2016 at 10:14 PM, Zachary Ball [email protected] wrote:

Hi