Skip to content
GitLab
Explore
Sign in
Michael Krause
tardis-doc
Compare revisions
4bfb276a16fb41b7fdb7ac378c1bd52e3d43c7b7 to 173d6289e64966007c6829adb3c52a233d60a270
Commits on Source (3)
Slurm: more sacct examples
· b846e5e5
Michael Krause
authored
Dec 07, 2022
b846e5e5
Change width a little
· 7579219a
Michael Krause
authored
Dec 07, 2022
7579219a
Merge remote-tracking branch 'origin/main'
· 173d6289
Michael Krause
authored
Dec 07, 2022
173d6289
Hide whitespace changes
Inline
Side-by-side
_static/.gitignore
deleted
100644 → 0
View file @
4bfb276a
*
!.gitignore
_static/custom.css
0 → 100644
View file @
173d6289
.wy-nav-content
{
max-width
:
1000px
!important
;
}
conf.py
View file @
173d6289
...
...
@@ -285,3 +285,7 @@ texinfo_documents = [
# If true, do not generate a @detailmenu in the "Top" node's menu.
#texinfo_no_detailmenu = False
html_css_files
=
[
'
custom.css
'
,
]
rm/slurm/commands.inc
View file @
173d6289
...
...
@@ -76,11 +76,17 @@ The ``sacct`` command has many filtering options to help you wade through large
numbers
of
jobs
.
Notable
examples
are
`--starttime`
,
`--endtime`
,
`--state`
,
or
`--name`
(
does
not
support
wildcards
unfortunately
,
but
you
can
use
grep
)
.
Example
:
Look
up
all
(
historical
)
jobs
for
user
`moneta`
on
Nikolaustag
,
that
started
at
8
am
or
later
and
finished
at
11
pm
or
earlier
with
the
state
completed
(
`CD`
)
.
Limit
output
to
the
fields
listed
with
`-o`
.
Because
output
data
will
be
truncated
this
way
and
in
this
case
the
job
name
is
longer
than
10
characters
,
we
add
a
format
hint
`%30`
to
leave
enough
room
for
the
complete
name
.
Example
:
Look
up
all
(
historical
)
jobs
for
user
`moneta`
on
Nikolaustag
,
that
started
at
8
am
or
later
and
finished
at
11
pm
or
earlier
with
the
state
completed
(
`CD`
)
.
Limit
output
to
the
fields
listed
with
`-o`
.
Because
output
data
will
be
truncated
this
way
and
in
this
case
the
job
name
is
longer
than
10
characters
,
we
add
a
format
hint
`%30`
to
leave
enough
room
for
the
complete
name
.
..
code
-
block
::
bash
[
krause
@
master
~
]
sacct
--
starttime
12
/
06
-
08
:
00
--
endtime
12
/
06
-
23
:
00
--
state
CD
--
parsable
--
user
moneta
-
o
JobID
,
JobName
%
30
,
TotalCPU
,
Elapsed
,
ExitCode
|
grep
MagicM41_3_3_1
[
krause
@
master
~
]
sacct
--
starttime
12
/
06
-
08
:
00
--
endtime
12
/
06
-
23
:
00
--
state
CD
--
parsable
--
user
moneta
\
-
o
JobID
,
JobName
%
30
,
TotalCPU
,
Elapsed
,
ExitCode
|
grep
MagicM41_3_3_1
2392298
MagicM41_3_3_1_1_1
00
:
35.463
00
:
00
:
36
0
:
0
2392299
MagicM41_3_3_1_1_2
00
:
27.970
00
:
00
:
29
0
:
0
...
...
@@ -101,15 +107,19 @@ the `--parsable` option. All data will be printed in full and with a pipe (`|`)
symbol
as
a
separator
(
configurable
)
.
Using
`--units`
,
we
can
get
uniform
scaling
of
the
values
.
Example:
Find job ids of jobs with a specific name and other parameters.
In
a sec
ond step, query all job "batch" *steps* with that job id for
their maximum memory usage (this not an aggregated value
unfortunately), filter values that are
exactly 0 (ended to quickly
for sampling) and plot the data in the terminal
using bashplotlib
(must be installed with pip).
Example
:
Find
job
ids
of
jobs
with
a
s
p
ec
ific
name
and
other
parameters
.
In
a
second
step
,
query
all
job
"batch"
*
steps
*
with
that
job
id
for
their
maximum
memory
usage
(
this
not
an
aggregated
value
unfortunately
),
filter
values
that
are
exactly
0
(
ended
to
quickly
for
sampling
)
and
plot
the
data
in
the
terminal
using
bashplotlib
(
must
be
installed
with
pip
)
.
[krause@master ~] ids=$(sacct --starttime 12/06-08:00 --endtime 12/06-23:00 --state CD --parsable --user moneta -o JobID,JobName%30.batch | grep MagicM41 | cut -d"|" -f1)
[krause@master ~] for id in $ids ; do sacct --parsable2 --noheader --units M -o MaxRSS -j $id.batch; done | tr -d "M" | grep -v "^0$" | hist -b 50 -x
..
code
-
block
::
bash
[
krause
@
master
~
]
ids
=
$
(
sacct
--
starttime
12
/
06
-
08
:
00
--
endtime
12
/
06
-
23
:
00
--
state
CD
--
parsable
\
--
user
moneta
-
o
JobID
,
JobName
%
30.
batch
|
grep
MagicM41
|
cut
-
d
"|"
-
f1
)
[
krause
@
master
~
]
for
id
in
$ids
;
do
sacct
--
parsable2
--
noheader
--
units
M
-
o
MaxRSS
\
-
j
$id
.
batch
;
done
|
tr
-
d
"M"
|
grep
-
v
"^0$"
|
hist
-
b
50
-
x
10
|
o
9
|
o
...
...