Merge pull request #3600 from chapmanb/master

Improved logging for errors during BatchLauncher.stop

This is a logging/debugging improvement that catches and reports errors when shutting down clusters using a BatchLauncher (qdel/bkill). The previous behavior was to silently fail if the external command failed. The patch captures error output and reports potential issues to the log.

IPython/parallel/apps/launcher.py

@@ -1132,7 +1132,15 b' class BatchSystemLauncher(BaseLauncher):'
         return job_id
 
     def stop(self):
-        output = check_output(self.delete_command+[self.job_id], env=os.environ)
+        try:
+            p = Popen(self.delete_command+[self.job_id], env=os.environ,
+                      stdout=PIPE, stderr=PIPE)
+            out, err = p.communicate()
+            output = out + err
+        except:
+            self.log.exception("Problem stopping cluster with command: %s" %
+                               (self.delete_command + [self.job_id]))
+            output = ""
         output = output.decode(DEFAULT_ENCODING, 'replace')
         self.notify_stop(dict(job_id=self.job_id, output=output)) # Pass the output of the kill cmd
         return output