On 26/06/2024 10:30 am, Farley, Peter wrote:
Just a question of curiosity. In recent days I have been running some ad-hoc BPXBATCH
jobs that executes some "cp" commands to copy a few z/OS data files down to a
Unix directory, then a python script which uses that data.
While the job is running I can use SDSF DA and PS to see the various Unix parts
running, but at the end of the job the shop-local IEFACTRT report in the
JESMSGLG output only seems to account for the actual BPXBATCH CPU time and I/O
count. The much more significant Unix I/O and CPU values are not included in
that report as far as I can tell.
Unix work (i.e. each separate command) typically runs in a separate
address space which produces its own SMF 30 step end accounting records.
They are often in different service/report classes to the parent job.
A job running a script with a loop can potentially produce thousands of
type 30 records for child tasks. The CPU etc accounting is in the child
task records, not the main job.
EasySMF attempts to follow the parent/child relationships from the SMF
30 records and produce cumulative numbers for the parent task. You can
see an example here:
https://www.blackhillsoftware.com/news/2019/08/27/comparing-bash-and-bin-sh-on-z-os/
When the tree is expanded, you can see the accounting for the individual
subtasks. If you collapse the tree, the subtasks are combined into the
parent task figure. This can be pretty significant, e.g. in the example
tested the total CPU time was over 1 minute, but only 2 hundredths of a
second appeared in the main job SMF record.
--
Andrew Rowley
Black Hill Software
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
