Re: [HTCondor-users] return codes from jobwrapper

[Todd Tannenbaum <tannenba@xxxxxxxxxxx>]

On 3/10/2020 4:04 PM, Beyer, Christoph wrote:
> Hi Greg,
>
> thanks for the superquick reply, it is a user-jobwrapper indeed, are the codes documented somewhere or do I have to dig in the code ?

Hi Christoph,

See the HTCondor Manual for additional details and examples... you can now even enter  _CONDOR_WRAPPER_ERROR_FILE in the 
Search Box and get useful results :)


One blurb of interest from the manual entry on USER_JOB_WRAPPER knob:

"If the wrapper script encounters an error as it runs, and it is unable to run the user job, it is important that the 
wrapper script indicate this to the HTCondor system so that HTCondor does not assign the exit code of the wrapper script 
to the job. To do this, the wrapper script should write a useful error message to the file named in the environment 
variable _CONDOR_WRAPPER_ERROR_FILE, and then the wrapper script should exit with a non-zero value. If this file is 
created by the wrapper script, HTCondor assumes that the wrapper script has failed, and HTCondor will place the job back 
in the queue marking it as Idle, such that the job will again be run. The condor_starter will also copy the contents of 
this error file to the condor_starter log, so the administrator can debug the problem...."

Hope the above helps,
Todd