-
Type: Improvement
-
Resolution: Duplicate
-
Priority: Major
-
None
-
Affects Version/s: 4.8.0
-
Component/s: Monitord, Pegasus Dashboard
-
None
Currently what dashboard describes as stderr is really the output from pegasus kickstart:
https://sugwg-osg.phy.syr.edu/pegasus/u/patrick.miles/r/4/w/2/j/6466/ji/16140/stderr
and stderr and stdout from the user job are both located in the file called stdout:
https://sugwg-osg.phy.syr.edu/pegasus/u/patrick.miles/r/4/w/2/j/6466/ji/16140/stdout
This is confusing users as they are looking at the stderr link for error messages from the user executable and are not seeing them.
Dashboard should clearly differentiate between what is stderr/stdout from the user's job and what is logging from kickstart so that non-expert users can perform first-line debugging of failed jobs.
Adding to the confusion, users often see error messages from globus when the job fails and the output doesn't exist. They think this is a Globus/Pegasus error and do not realize that the error is in the user's job (see https://jira.isi.edu/browse/PM-1148)