Ticket 1954

Summary:	correlating sacct and sreport output
Product:	Slurm	Reporter:	Susan Chacko <susanc>
Component:	Accounting	Assignee:	David Bigagli <david>
Status:	RESOLVED INFOGIVEN	QA Contact:
Severity:	3 - Medium Impact
Priority:	---	CC:	brian, da
Version:	14.11.9
Hardware:	Linux
OS:	Linux
Site:	NIH	Slinky Site:	---
Alineos Sites:	---	Atos/Eviden Sites:	---
Confidential Site:	---	Coreweave sites:	---
Cray Sites:	---	DS9 clusters:	---
Google sites:	---	HPCnow Sites:	---
HPE Sites:	---	IBM Sites:	---
NOAA SIte:	---	NoveTech Sites:	---
Nvidia HWinf-CS Sites:	---	OCF Sites:	---
Recursion Pharma Sites:	---	SFW Sites:	---
SNIC sites:	---	Tzag Elita Sites:	---
Linux Distro:	---	Machine Name:
CLE Version:		Version Fixed:	14.11.8
Target Release:	---	DevPrio:	---
Emory-Cloud Sites:	---

Description Susan Chacko 2015-09-17 02:06:18 MDT

I have a question about the output of sacct and sreport. Here's an example for a single hour:

sacct -pna --format User,Account,Partition,JobID,NCPUS,CPUTime,CPUTimeRaw,State,Start,End --start=2015-04-26T11:00:00 --end=2015-04-26T12:00:00
susanc|sb|interactive|16990|8|2-16:02:00|230520|TIMEOUT|2015-04-26T10:12:20|2015-04-26T18:12:35|
|sb||16990.0|8|2-16:06:24|230784|CANCELLED|2015-04-26T10:12:20|2015-04-26T18:13:08|

[susanc@biowulf2 ~]$ sreport Cluster AccountUtilizationByUser start=2015-04-26T11:00:00 end=2015-04-26T12:00:00
--------------------------------------------------------------------------------
Cluster/Account/User Utilization 2015-04-26T11:00:00 - 2015-04-26T11:59:59 (3600 secs)
Time reported in CPU Minutes
--------------------------------------------------------------------------------
  Cluster         Account     Login     Proper Name       Used     Energy
--------- --------------- --------- --------------- ---------- ----------
  biowulf            root                                  480          0
  biowulf              sb                                  480          0
  biowulf              sb    susanc    Susan Chacko        480          0


As you see, sacct reports CPU Time of  230520 seconds =~ 64 hrs. sreport reports 480 minutes = 8 hrs. It looks like sreport is reporting the actual utilization of CPU during that one hour time period (8 cpus * 1 hr = 8 hrs), while sacct is reporting the total CPU used by any jobs that ran during that time period, including the CPU which was utilized outside that hour? Do I have that right?

And, related, are there sacct flags I could use to get the same output as I'm getting from sreport?

Thanks!
Susan.

Comment 1 David Bigagli 2015-09-18 00:50:07 MDT

Hi,
   the answer to your first question is yes. If you want to get the same output
from sacct as sreport you have to use the -T option of sacct with the start and
end time given.

David

Comment 2 Susan Chacko 2015-09-23 05:04:19 MDT

Thanks, that answers and solves my problem.

Comment 3 Danny Auble 2015-09-23 05:05:13 MDT

Excellent! thanks!