Swapped two colours in the status plot, to make RUNNING
status more distinct.
Added type hints.
Added support for TOML specification files.
Fixed failing test.
Removed kaleido
from the requirements in setup.py
. No idea how that got there.
Reversed order of status categories displayed in bin/slurm-pipeline-status-plot.py
plots.
Added "pandas" to test requirements in setup.py
.
Small tweaks to improve the appearance of bin/slurm-pipeline-status-plot.py
plots.
Added bin/slurm-pipeline-status-plot.py
.
Make loading the JSON specification file give a better error message, including the specification filename.
Added missing <
in command line providing input from *.bz2
files to the
command being executed by sbatch.py
.
Added --verbose
option to sbatch.py
.
Added --compressLevel
and --randomSleep
options to sbatch.py
.
Added compression of input files to sbatch.py
and an --uncompress
option to turn this off. Thanks to
@holtgrewe for this suggestion.
Updated .travis.yml
and build status image URL on README.md
.
Added conda install instructions to README.md
.
Improved README.md
description of sbatch.py
.
Added JSON output of all SLURM job numbers and added to README.md
to
describe how to easily use the output with
jq to pass job ids to sacct
, squeue
,
scancel
or sbatch.py --afterOk
. Fixed silly bug causing --then
,
--else
, and --finally
commands not to be wrapped in a script for
submission to sbatch
.
Improved README.md
.
Moved test
directory back to top level.
Moved test
directory under slurm_pipeline
.
Use getpass.getuser()
instead of pwd.getpwuid(os.getuid()).pw_name
to
get the user's name (since the import pwd
fails on windows).
Added LICENSE.txt
for conda forge distribution.
Removed support for Python 2.7. Added conda building.
Added sbatch.py
utility script (and remove-repeated-headers.py
helper).
Added --printOutput
option to tell slurm-pipeline.py
to print the
output of running each step.
The distinction between simulated and skipped steps has been removed as it
was too subtle and complicated. Only step skipping remains now, and step
scripts need to decide what to do based on the value of SP_SKIP
. For
backwards compatibility, the SP_SIMULATE
variable is still set in the
environment, but its value is always identical to SP_SKIP
.
Removed use of now-outdated ugly mocking code for open
. This might break
Python 2 compatibility, but at this point I don't care.
The --nice
argument put into the SP_NICE_ARG
environment variable
changed to be e.g., --nice=50
(used to be --nice 50
which causes
sbatch
to complain "sbatch: error: Unable to open file 50").
Removed unused specification
argument to SAcct
class.
Added missing --scriptArgs
to README.md
. Thanks
@avatar-lavventura!
Added env PYTHONPATH=../..
to all example Makefiles.
- Added
--scriptArgs
toslurm-pipeline.py
to formalize giving arguments on the command line that should be passed to initial step scripts (i.e., those that have no dependencies). This prevents the possibility of putting unrecognized arguments (like--startStep
) on the command line which are then passed to the script steps.
Added Travis CI build messaging to IRC.
Updated funding acknowledgements in README.md
.
Added "error step" setting to per-step information printed by
slurm-pipeline-status.py
.
Added JobName to list of default job status fields printed by
slurm-pipeline-status.py
.
Added printing of SP_* environment variables to per-step output of
slurm-pipeline-status.py
.
Added "Collect step:" to per-step output of slurm-pipeline-status.py
.
This change is not backward-compatible with earlier versions, hence the major version number change.
- SLURM job status reporting was changed to use
sacct
rather thansqueue
as a result of a pull request submitted by healther. This has the advantage of being able to show status information for jobs that have completed. - The
squeueArgs
command line option toslurm-pipeline.py
was removed. - A
fieldNames
argument was added toslurm-pipeline-status.py
. This can be used to specify what job fields you want to see in the summary output. The default isState
,Elapsed
, andNodelist
. You can setSP_STATUS_FIELD_NAMES
to a comma-separated list of job fields to set your own default preference. Seeman sacct
for the full list of field names.
Other changes
- It is now possible to specify
error step: true
for a step that has no dependencies. This used to be flagged as an error, but it makes sense to allow it as the user might use--startAfter
on the command line, and the jobs which the no-dependency step is suppose to start after may fail. - The
Makefile
now usespycodestyle
instead ofpep8
to check Python style (this is just a new name for thepep8
tool). - An
--output
option was added toslurm-pipeline.py
to allow the specification of an output file for the JSON status. The default is to print it to standard output. SP_DEPENDENCY_ARG
is now set for all steps that don't have dependencies if--startAfter
is used when invokingslurm-pipeline.py
.- The all-step summary section of the
slurm-pipeline-status.py
output is slightly improved, to list an overall count of jobs started and finished.