Wilson Bilkovich
10/05/2021, 1:30 PMDask Performance Report
of a Prefect Flow that is using DaskExecutor
i.e. https://distributed.dask.org/en/latest/api.html#distributed.performance_report
It looks like the dask.distributed.performance_report()
needs to run inside the context of the temporary Dask Cluster that Prefect spins up, but I don’t see how to get into those internals.Kevin Kho
@task(checkpoint=False)
def using_checkpoint_false(filepath):
with worker_client():
with performance_report(filename="dask-report.html"):
return dd.read_csv(filepath)
Kevin Kho
with Flow("example") as flow:
with performance_report(...):
some_task()
other_task()
and this might run if the flow is stored as a scriptWilson Bilkovich
10/05/2021, 1:58 PMBob Colner
10/05/2021, 2:51 PMWilson Bilkovich
10/05/2021, 4:54 PMWilson Bilkovich
10/06/2021, 12:52 PMKevin Kho
executor.submit()
is agnostic to the full workflow already. If you have a mapped task with 5 elements, executor.submit()
calls them one by one. I am not sure there is a good place to inject this then. Will ask the team for ideas.Wilson Bilkovich
10/06/2021, 3:23 PMKevin Kho
Kevin Kho
Bob Colner
10/08/2021, 1:38 PM