Community
Participate
Working Groups
If I launch a job using the PBS-Generic-Interactive RM, but the command fails, the failure is not connected and the progress bar continues indefinitely. The command fails for me because I don't have a default account, so one must be specified on the qsub command.
It doesn't seem to me that this is a generalized problem. It has to do, once again, with knowing what kind of error messages are sent back by the batch system, and on which of stdout/stderr. This is always, always, a headache. Welcome to the world of HPC.
In this case, qsub returns with exit status 1. This is correctly detected by the PBS-Generic-Batch RM and I get an error message. The interactive RM should do the same thing, but instead it continues as if the command was running.
(In reply to comment #2) > In this case, qsub returns with exit status 1. This is correctly detected by > the PBS-Generic-Batch RM and I get an error message. The interactive RM should > do the same thing, but instead it continues as if the command was running. So it's a problem of the ptty. Another headache.
Except it would become difficult to validate anything in it. Since the submission mechanism is not to use the command-line flags because those are not necessarily supported, but to replace these values on the script, one would also have to rewrite the script so that it has a single -l line. These are not difficult changes if you really think it makes this more usable -- now that the documentation/tutorial exists to demonstrate how to modify the underlying definition to get this. You could even do away with the Advanced Tab altogether, and just provide the "Standard" and the Customized tab controllers.
(In reply to comment #4) > Except it would become difficult to validate anything in it. Since the > submission mechanism is not to use the command-line flags because those are not > necessarily supported, but to replace these values on the script, one would > also have to rewrite the script so that it has a single -l line. > > These are not difficult changes if you really think it makes this more usable > -- now that the documentation/tutorial exists to demonstrate how to modify the > underlying definition to get this. > > You could even do away with the Advanced Tab altogether, and just provide the > "Standard" and the Customized tab controllers. sorry, comment went on the wrong bug.
This was a silly bug. It has been fixed, and I also changed the code so that if an exception like this is thrown while waiting for the jobId, the output stream from the proxy monitor is appended to the error, since the pbs shell script wrapper on lincoln returns the error message, logically, on stdout :-). The fix will be committed when I also commit the validator patch.
Closing this as fixed under the assumption that Al committed the changes.
Closing.
