Scientific workflow engine designed for simplicity & scalability. Trivially transition between one off use cases to massive scale production environments
Based on a conversation with the Epigenomics group that is using Cromwell 34 on PAPI v2 --
it seems that a task can fail, not delocalize an expected output and the workflow will continue to start a downstream task but fail at running the command (as an expected input is missing). It's been confirmed that Cromwell was running in a Fail fast mode.
There seem to be two main requirements here:
Cromwell should be evaluating if all expected outputs for a task exist before marking the task as a success.
Cromwell should fail a job if it failed localize a specific input file.
Cromwell should fail a job if it failed to delocalize any expected output file.
AC: Ensure that the PAPI v2 backend fulfills requirements #1-3.
Based on a conversation with the Epigenomics group that is using Cromwell 34 on PAPI v2 -- it seems that a task can fail, not delocalize an expected output and the workflow will continue to start a downstream task but fail at running the command (as an expected input is missing). It's been confirmed that Cromwell was running in a Fail fast mode.
There seem to be two main requirements here:
AC: Ensure that the PAPI v2 backend fulfills requirements #1-3.