[9.0] JobWrapper: call jobReport.commit() at job finalize()

[fstagni]

BEGINRELEASENOTES

*WMS
FIX: call jobReport.commit() at job finalize()

ENDRELEASENOTES

[chaen]

Can it be called multiple time (in case it ends up being called already further down) ?

[fstagni]

Yes.

[aldbr]

What happens if self.jobID is not defined (if the job wrapper is executed locally)? Not entirely sure it would work.

[aldbr]

What kind of issue are you trying to solve here?
Because there is already a call to commit() in the JobWrapperTemplate:

DIRAC/src/DIRAC/WorkloadManagementSystem/JobWrapper/JobWrapperTemplate.py

Line 104 in 741ef31

jobReport.commit()

[aldbr]

What happens if self.jobID is not defined (if the job wrapper is executed locally)? Not entirely sure it would work.

[fstagni]

The trigger for this are few jobs, ran in the certification setup, where some parameters were not added. Specifically, the parameter not added was "OutputSandboxLFN". Such parameter is added by the JobWrapper:

DIRAC/src/DIRAC/WorkloadManagementSystem/JobWrapper/JobWrapper.py

Line 948 in 741ef31

"OutputSandboxLFN", self.__getLFNfromOutputFile(outputSandboxData)[0], sendFlag=False

In the case when the output sandbox is too large and therefor uploaded as an LFN.

This parameter is set after the execution of jobReport.commit() that you mention above.

[aldbr]

The trigger for this are few jobs, ran in the certification setup, where some parameters were not added. Specifically, the parameter not added was "OutputSandboxLFN". Such parameter is added by the JobWrapper:

We indeed have the same issue in LHCb.
It's not only the OutputSandboxLFN from what I can see but all the parameters coming from the JobWrapper.
I am trying to debug it but I haven't managed to set up a reproducer yet, and I don't see any problematic DiracX log.

#8399 could potentially help

[aldbr]

I am closing this PR now that I know what's the problem.