Ticket 9528

Summary:	mpiexec does not propagate exit code
Product:	Slurm	Reporter:	Simon Byrne <simonbyrne>
Component:	User Commands	Assignee:	Tim Wickberg <tim>
Status:	RESOLVED FIXED	QA Contact:
Severity:	C - Contributions
Priority:	---	CC:	jess, rberger
Version:	21.08.x
Hardware:	Linux
OS:	Linux
Site:	-Other-	Alineos Sites:	---
Atos/Eviden Sites:	---	Confidential Site:	---
Coreweave sites:	---	Cray Sites:	---
DS9 clusters:	---	HPCnow Sites:	---
HPE Sites:	---	IBM Sites:	---
NOAA SIte:	---	OCF Sites:	---
Recursion Pharma Sites:	---	SFW Sites:	---
SNIC sites:	---	Linux Distro:	---
Machine Name:		CLE Version:
Version Fixed:	23.11.0rc1	Target Release:	---
DevPrio:	---	Emory-Cloud Sites:	---
Attachments:	Patch to correctly propagate exit code

Description Simon Byrne 2020-08-06 11:20:05 MDT

$ mpiexec bash -c "exit 1" 

$ echo $? 
0

I'm not that familiar with perl, but it looks like the return value for the system command here:
https://github.com/SchedMD/slurm/blob/b1656169f48a73c1229560d8d19a902e1956aac7/contribs/torque/mpiexec.pl#L165
should be captured (https://perldoc.perl.org/functions/system.html) so something like

my $exit_code = system($command);

system("rm -f $new_config") if($new_config);

exit($exit_code >> 8);

Comment 3 Simon Byrne 2020-08-07 12:14:32 MDT

Created attachment 15353 [details]
Patch to correctly propagate exit code

Comment 4 Richard Berger 2023-06-26 17:41:17 MDT

Are there any plans to fix this issue? We've just replicated that workaround for us locally. But it would be nice if it gets into a future release.

Comment 5 Tim Wickberg 2023-06-27 10:51:43 MDT

Simon -

Thanks for the submission. It's finally upstream, and will be included in the 23.11 release later this fall. Commit details follow for reference.

- Tim

commit 5ab128ca079102305be383a1ae23c7a87543150b
Author:     Simon Byrne <simonbyrne@gmail.com>
AuthorDate: Tue Jun 27 10:43:52 2023 -0600

    torque/mpiexec - Propagate exit code from launched process.
    
    Bug 9528.