runjob is a program for managing a group of related jobs running on a compute cluster localhost, Sun Grid Engine, BatchCompute, slurm . It provides a convenient method for specifying dependencies between jobs and the resource requirements for each job (e.g. memory, CPU cores). It monitors the status of the jobs so you can tell when the whole group is done. Litter cpu or memory resource is used in the login compute node.
python >=3.5
The development version can be installed with (for recommend)
pip install git+https://github.com/yodeng/runjob.git
The stable release (maybe not latest) can be installed with
pypi:
pip install runjob -U
conda:
conda install -c bioconda runjob
All manual can be found here.
You can get the quick help like this:
$ runjob --help
Usage: runjob [-h] [-v] [-j [<jobfile>]] [-n <int>] [-s <int>] [-e <int>]
[-w <workdir>] [-d] [-l <file>] [-r <int>] [-f] [-R <int>] [-C]
[-M {local,localhost,sge,slurm,batchcompute}] [--ini <configfile>]
[--dag] [--dag-extend] [--strict] [--quiet] [--max-check <float>]
[--max-submit <float>] [--max-queue-time <float/str>]
[--max-run-time <float/str>] [--max-wait-time <float/str>]
[--max-timeout-retry <int>]
[--local | --localhost | --sge | --slurm | --batchcompute]
[-i [<str> ...]] [-L <logdir>]
runjob is a tool for managing parallel tasks from a specific job file running
in localhost, sge, slurm, batchcompute.
Optional Arguments:
-h, --help show this help message and exit
--local submit your jobs to local, same as '--mode local'.
--localhost submit your jobs to localhost, same as '--mode localhost'.
--sge submit your jobs to sge, same as '--mode sge'.
--slurm submit your jobs to slurm, same as '--mode slurm'.
--batchcompute submit your jobs to batchcompute, same as '--mode
batchcompute'.
-i, --injname [<str> ...]
job names you need to run. (default: all job names of the
jobfile)
-L, --logdir <logdir>
the output log dir. (default: join(workdir, "logs"))
Base Arguments:
-v, --version show program's version number and exit
-j, --jobfile [<jobfile>]
input jobfile, if empty, stdin is used. (required)
-n, --num <int> the max job number runing at the same time. (default: all
of the jobfile, max 1000)
-s, --start <int> which line number(1-base) be used for the first job.
(default: 1)
-e, --end <int> which line number (include) be used for the last job.
(default: last line of the jobfile)
-w, --workdir <workdir>
work directory. (default: /home/dengyong/soft/git/runjob)
-d, --debug log debug info.
-l, --log <file> append log info to file. (default: stdout)
-r, --retry <int> retry N times of the error job, 0 or minus means do not
re-submit. (default: 0)
-f, --force force to submit jobs even already successed.
-R, --retry-sec <int>
retry the error job after N seconds. (default: 2)
-C, --config show configurations and exit.
-M, --mode {local,localhost,sge,slurm,batchcompute}
the mode to submit your jobs, if no sge installed, always
localhost. (default: sge)
--ini <configfile> input configfile for configurations search.
--dag do not execute anything and print the directed acyclic
graph of jobs in the dot language.
--dag-extend do not execute anything and print the extend directed
acyclic graph of jobs in the dot language.
--strict use strict to run, means if any errors, clean all jobs and
exit.
--quiet suppress all output and logging.
--max-check <float> maximal number of job status checks per second, fractions
allowed. (default: 5)
--max-submit <float> maximal number of jobs submited per second, fractions
allowed. (default: 20)
Time Control Arguments:
--max-queue-time <float/str>
maximal time (d/h/m/s) between submit and running per job.
(default: no-limiting)
--max-run-time <float/str>
maximal time (d/h/m/s) start from running per job.
(default: no-limiting)
--max-wait-time <float/str>
maximal time (d/h/m/s) start from submit per job. (default:
no-limiting)
--max-timeout-retry <int>
retry N times for the timeout error job, 0 or minus means
do not re-submit. (default: 0)
$ runsge --help
Usage: runsge [-h] [-v] [-j [<jobfile>]] [-n <int>] [-s <int>] [-e <int>]
[-w <workdir>] [-d] [-l <file>] [-r <int>] [-f] [-R <int>] [-C]
[-M {local,localhost,sge,slurm,batchcompute}] [--ini <configfile>]
[--dag] [--dag-extend] [--strict] [--quiet] [--max-check <float>]
[--max-submit <float>] [--max-queue-time <float/str>]
[--max-run-time <float/str>] [--max-wait-time <float/str>]
[--max-timeout-retry <int>]
[--local | --localhost | --sge | --slurm | --batchcompute]
[-N <jobname>] [-L <logdir>] [-g <int>] [--init <cmd>]
[--call-back <cmd>] [-q [<queue> ...]] [-m <int>] [-c <int>]
[--out-maping <dir>] [--access-key-id <str>]
[--access-key-secret <str>]
[--region {beijing,hangzhou,huhehaote,shanghai,zhangjiakou,chengdu,hongkong,qingdao,shenzhen}]
runsge is a tool for managing parallel tasks from a specific shell file
runing in localhost, sge, slurm, batchcompute.
Optional Arguments:
-h, --help show this help message and exit
--local submit your jobs to local, same as '--mode local'.
--localhost submit your jobs to localhost, same as '--mode localhost'.
--sge submit your jobs to sge, same as '--mode sge'.
--slurm submit your jobs to slurm, same as '--mode slurm'.
--batchcompute submit your jobs to batchcompute, same as '--mode
batchcompute'.
-N, --jobname <jobname>
job name. (default: basename of the jobfile)
-L, --logdir <logdir>
the output log dir. (default:
"/home/dengyong/soft/git/runjob/runsge_*_log_dir")
-g, --groups <int> N lines to consume a new job group. (default: 1)
--init <cmd> command before all jobs, will be running in localhost.
--call-back <cmd> command after all jobs finished, will be running in
localhost.
Base Arguments:
-v, --version show program's version number and exit
-j, --jobfile [<jobfile>]
input jobfile, if empty, stdin is used. (required)
-n, --num <int> the max job number runing at the same time. (default: all
of the jobfile, max 1000)
-s, --start <int> which line number(1-base) be used for the first job.
(default: 1)
-e, --end <int> which line number (include) be used for the last job.
(default: last line of the jobfile)
-w, --workdir <workdir>
work directory. (default: /home/dengyong/soft/git/runjob)
-d, --debug log debug info.
-l, --log <file> append log info to file. (default: stdout)
-r, --retry <int> retry N times of the error job, 0 or minus means do not
re-submit. (default: 0)
-f, --force force to submit jobs even already successed.
-R, --retry-sec <int>
retry the error job after N seconds. (default: 2)
-C, --config show configurations and exit.
-M, --mode {local,localhost,sge,slurm,batchcompute}
the mode to submit your jobs, if no sge installed, always
localhost. (default: sge)
--ini <configfile> input configfile for configurations search.
--dag do not execute anything and print the directed acyclic
graph of jobs in the dot language.
--dag-extend do not execute anything and print the extend directed
acyclic graph of jobs in the dot language.
--strict use strict to run, means if any errors, clean all jobs and
exit.
--quiet suppress all output and logging.
--max-check <float> maximal number of job status checks per second, fractions
allowed. (default: 5)
--max-submit <float> maximal number of jobs submited per second, fractions
allowed. (default: 20)
Time Control Arguments:
--max-queue-time <float/str>
maximal time (d/h/m/s) between submit and running per job.
(default: no-limiting)
--max-run-time <float/str>
maximal time (d/h/m/s) start from running per job.
(default: no-limiting)
--max-wait-time <float/str>
maximal time (d/h/m/s) start from submit per job. (default:
no-limiting)
--max-timeout-retry <int>
retry N times for the timeout error job, 0 or minus means
do not re-submit. (default: 0)
Sge/Slurm Arguments:
-q, --queue [<queue> ...]
the queue/partition your job running, multi queue can be
sepreated by whitespace. (default: all accessed queue)
-m, --memory <int> the memory used per command (GB). (default: 1)
-c, --cpu <int> the cpu numbers you job used. (default: 1)
Batchcompute Arguments:
--out-maping <dir> the oss output directory if your mode is "batchcompute",
all output file will be mapping to you OSS://BUCKET-NAME.
if not set, any output will be reserved.
--access-key-id <str>
AccessKeyID while access oss.
--access-key-secret <str>
AccessKeySecret while access oss.
--region {beijing,hangzhou,huhehaote,shanghai,zhangjiakou,chengdu,hongkong,qingdao,shenzhen}
batch compute region. (default: beijing)
$ qs --help
For summary all jobs
Usage: qs [jobfile|logdir|logfile]
qcs --help
qslurm
runjob is distributed under the MIT license.
Please send comments, suggestions, bug reports and bug fixes to yodeng@tju.edu.cn.
$ cat test.flow
logs: ./logs
envs:
samples: A B C
qc:
echo "qc $samples"
bwa:
echo "bwa $samples"
depends: qc.samples ## each sample-bwa depends each sample-qc
# depends: qc ## each sample-bwa depends all sample-qc
sort:
echo "sort $samples"
depends: bwa.samples
index:
echo "index $samples"
depends: sort.samples
calling:
echo "calling allsample"
depends: index
stats:
echo "stats allsample"
depends: calling
command runflow -j test.flow --dag | dot -Tsvg > test.svg will get the job graph:
More functions will be improved in the future.