The list of hacks:
- We search for the file specified under
-xrslrecursively in all subfolders of a given folder.
- The option
-I(idle time) was added
- The job is considered to be successfully downloaded when it is indeed successfully downloaded, i.e., we check for the presence of
Successfully cleaned: 1etc. message rather than the absence of
ERRORetc. which comes in handy when the
line 3188error happens:)
arcproxycommand is automatically called on every iteration
- The current job ids of the run with
-xrsl <name>.<extension>are saved to
- The program no longer calls
arcstat -ato get job statuses, it instead calls
arcstatonly for jobs that it is managing. This prevents seeing unrelated warnings if the program is not managing all your current jobs (multiple arcrunners running or some manual submissions).
- The parameters
Sare set to infinity in the case when the number of clusters is 1. This prevents canceling the job and resubmitting it to the same cluster again, which makes no sense.
Tintroduced, which is useful, if we want to stop processing the jobs even though they are not all finished. More or less, this was the easiest way to refresh the job list ...
- We check whether the output files were downloaded successfully. The reason for this is that sometimes the first byte of the file is lost on the way from grid to the computer.
- Arccat option