Ticket 4621

Summary:	srun with multiprog option cannot find executable in subdir
Product:	Slurm	Reporter:	Pierre Trespeuch <pierre-externe.trespeuch>
Component:	User Commands	Assignee:	Felip Moll <felip.moll>
Status:	RESOLVED FIXED	QA Contact:
Severity:	4 - Minor Issue
Priority:	---	CC:	felip.moll
Version:	17.02.2
Hardware:	Linux
OS:	Linux
Site:	EDF - Electricite de France	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:	18.08	Target Release:	---
DevPrio:	---	Emory-Cloud Sites:	---
Attachments:	fix for 4621 add test_exec to multi-prog

Description Pierre Trespeuch 2018-01-12 03:02:41 MST

Hello,

srun send an error with multiprog option when the binary is in a subdirectory and path is not prefixed with "./". This problem can be observed with the following setup :


multiprog.conf :
  0-27 subdir/test.exe

command :
  srun --multi-prog multiprog.conf

error:
  srun: error: Could not find executable subdir/test.exe


But subdir/test.exe is executed as it should.

Thanks,

Pierre

Comment 5 Felip Moll 2018-01-15 03:37:29 MST

Created attachment 5920 [details]
fix for 4621

Comment 6 Felip Moll 2018-01-15 03:37:51 MST

Created attachment 5921 [details]
add test_exec to multi-prog

Comment 7 Felip Moll 2018-01-15 03:41:42 MST

Pierre,

I created this two patches to solve your problem. This patches are still not commited but will be soon. Feel free to download them, apply and try.

Attachment 5920 [details] will fix your problem.

I took also into account the --chdir case.

Attachment 5921 [details] adds test_exec feature to --multi-prog

Using test_exec feature will now work with multi-prog. It is not necessary for your issue but I recommend it.

Tell me how it goes.

Comment 8 Pierre Trespeuch 2018-01-15 06:26:33 MST

Hello,

Thank you for your quick answer.
We are going to update Slurm from version 17.02.2 to version 17.11.2 next week.
Do you know if the --multiprog behavior is the same with version 17.11.12 ?
If there is the same problem, would it be possible for you to provide us a patch for Slurm 17.11.2 ?

Thanks,

Pierre

Comment 9 Felip Moll 2018-01-15 08:03:34 MST

(In reply to Pierre Trespeuch from comment #8)
> Hello,
> 
> Thank you for your quick answer.
> We are going to update Slurm from version 17.02.2 to version 17.11.2 next
> week.
> Do you know if the --multiprog behavior is the same with version 17.11.12 ?
> If there is the same problem, would it be possible for you to provide us a
> patch for Slurm 17.11.2 ?
> 
> Thanks,
> 
> Pierre

Sorry I forgot to mention this patch is coded already for 17.11, so you can apply it next week when upgrading.

If it was needed for 17.02 I would backport it but since you are upgrading I am not going to start it for now.

Comment 13 Felip Moll 2018-01-18 03:08:04 MST

Hi Pierre,

We've finally committed the patches in commits:

Your issue:
https://github.com/SchedMD/slurm/commit/d3d302b0eda0

Improvement test_exec with --multi-prog:
https://github.com/SchedMD/slurm/commit/cf4d2145cd485
https://github.com/SchedMD/slurm/commit/1ab46fddca7b6

This will be available in next 18.08 release. It should be safe to backport to 17.11.3 but it will not be there officially.

I am closing this bug now, if you need further assistance don't hesitate to reopen the bug.

Regards
Felip M

Comment 14 Pierre Trespeuch 2018-01-18 03:43:48 MST

Hi Felip,

Thank you for your assistance !

Regards,

Pierre