def _qstat(jobid=None, username=getlogin(), full=False):
"""Return the stdout of ``qstat`` minus the header lines.
By default, 'username' is set to the current user.
'full' is the '-f' option
'jobid' is a string or list of strings of job ids
Returns:
str: the text of qstat, minus the header lines
"""
# -u and -f contradict in earlier versions of Torque
if full and username is not None and int(
torque_version.split(".")[0]) < 5 and jobid is None:
# First get all jobs by the user
qopt = ["qselect"]
qopt += ["-u", username]
# Call 'qselect' using subprocess
stdout = run(qopt)[0]
# Get the jobids
jobid = []
for line in StringIO(stdout):
jobid += [line.rstrip("\n")]
opt = ["qstat"]
# If there are jobid(s), you don't need a username
if username is not None and jobid is None:
opt += ["-u", username]
# But if there are jobid(s) and a username, you need -a to get full output
elif username is not None and jobid is not None and not full:
opt += ["-a"]
# By this point we're guaranteed torque ver >= 5.0, so -u and -f are safe together
if full:
opt += ["-f"]
if jobid is not None:
if isinstance(jobid, string_types):
jobid = [jobid]
elif isinstance(jobid, list):
pass
else:
print(
"Error in prisms_jobs.interface.torque.qstat(). type(jobid):",
type(jobid))
sys.exit()
opt += jobid
# call 'qstat' using subprocess
stdout, stderr, returncode = run(opt) #pylint: disable=unused-variable
sout = StringIO(stdout)
# strip the header lines
if full is False:
for line in sout:
if line[0] == "-":
break
# return the remaining text
return sout.read()
def submit(substr, write_submit_script=None):
"""Submit a job using ``sbatch``.
Args:
substr (str): The submit script string
write_submit_script (bool, optional): If true, submit via file skipping
lines containing '#SBATCH -J'; otherwise, submit via commandline. If
not specified, uses ``prisms_jobs.config['write_submit_script']``.
Returns:
str: ID of submitted job
Raises:
JobsError: If a submission error occurs
"""
m = re.search(r"#SBATCH\s+-J\s+(.*)\s", substr) #pylint: disable=invalid-name
if m:
jobname = m.group(1) #pylint: disable=unused-variable
else:
raise JobsError(
None,
r"""Error in pbs.misc.submit(). Jobname ("#SBATCH\s+-J\s+(.*)\s") not found in submit string."""
)
# if write_submit_script is None:
# write_submit_script = prisms_jobs.config(['write_submit_script'])
if write_submit_script:
if os.path.exists(jobname):
index = 0
while os.path.exists(jobname + ".bak." + str(index)):
index += 1
print("Backing up existing submit script:", jobname, "->",
jobname + ".bak." + str(index))
os.rename(jobname, jobname + ".bak." + str(index))
# write submit script, without -N line
with open(jobname, 'w') as f:
for line in substr.splitlines():
if not re.search(r"SBATCH\s+-J\s+(.*)", line):
f.write(line + '\n')
stdout, stderr, returncode = run(["sbatch", jobname]) #pylint: disable=unused-variable
else:
stdout, stderr, returncode = run(["sbatch"],
input=substr,
stdin=subprocess.PIPE) #pylint: disable=unused-variable
print(stdout[:-1])
if re.search("error", stdout):
print(stdout)
raise JobsError(0, "Submission error.\n")
else:
jobid = stdout.rstrip().split()[-1]
return jobid
def alter(jobid, arg):
"""``scontrol`` update job.
Args:
jobid (str): ID of job to alter
arg (str): 'arg' is a scontrol command option string. For instance, "-a 201403152300.19"
Returns:
int: ``scontrol`` returncode
"""
return run(["scontrol", "update", "JobId=", jobid] + arg.split())[2]
def release(jobid):
"""``scontrol`` un-delay a job.
Args:
jobid (str): ID of job to release
Returns:
int: ``scontrol`` returncode
"""
return run(["scontrol", "update", "JobId=", jobid, "StartTime=", "now"])[2]
def delete(jobid):
"""``scancel`` a job.
Args:
jobid (str): ID of job to cancel
Returns:
int: ``scancel`` returncode
"""
return run(["scancel", jobid])[2]
def _getversion():
"""Returns the torque version as string or None if no ``qstat`` """
if find_executable("qstat") is None:
return None
opt = ["qstat", "--version"]
# call 'qstat' using subprocess
stdout = run(opt)[0]
# return the version number
return stdout.rstrip("\n").lower().lstrip("version: ")
def hold(jobid):
"""``scontrol`` delay a job.
Args:
jobid (str): ID of job to delay (for 30days)
Returns:
int: ``scontrol`` returncode
"""
return run(
["scontrol", "update", "JobId=", jobid, "StartTime=", "now+30days"])[2]
def alter(jobid, arg):
"""``qalter`` a job.
Args:
jobid (str): ID of job to alter
arg (str): 'arg' is a scontrol command option string. For instance, "-a 201403152300.19"
Returns:
int: ``qalter`` returncode
"""
stdout, stderr, returncode = run(["qalter"] + arg.split() + [jobid]) #pylint: disable=unused-variable
return returncode
def release(jobid):
"""``qrls`` a job.
Args:
jobid (str): ID of job to release
Returns:
int: ``qrls`` returncode
"""
stdout, stderr, returncode = run(["qrls", jobid]) #pylint: disable=unused-variable
return returncode
def hold(jobid):
"""``qhold`` a job.
Args:
jobid (str): ID of job to hold
Returns:
int: ``qhold`` returncode
"""
stdout, stderr, returncode = run(["qhold", jobid]) #pylint: disable=unused-variable
return returncode
def delete(jobid):
"""``qdel`` a PBS job.
Args:
jobid (str): ID of job to delete
Returns:
int: ``qdel`` returncode
"""
stdout, stderr, returncode = run(["qdel", jobid]) #pylint: disable=unused-variable
return returncode
def _squeue(jobid=None, username=getlogin(), full=False, sformat=None): #pylint: disable=unused-argument
"""Return the stdout of squeue minus the header lines.
By default, 'username' is set to the current user.
'full' is the '-f' option
'jobid' is a string or list of strings of job ids
'sformat' is a squeue format string (e.g., "%A %i %j %c")
Returns:
str: the text of squeue, minus the header lines
"""
# If Full is true, we need to use scontrol:
if full is True:
if jobid is None:
if username is None:
# Clearly we want ALL THE JOBS
sopt = ["scontrol", "show", "job"]
# Submit the command
# Nothing to strip, as scontrol provides no headers
return run(sopt)[0]
else:
# First, get jobids that belong to that username using
# squeue (-h strips the header)
sopt = ["squeue", "-h", "-u", username]
qsout = run(sopt)[0]
# Get the jobids
jobid = []
for line in StringIO(qsout):
jobid += [line.rstrip("\n")]
# Great, now we have some jobids to pass along
# Ensure the jobids are a list, even if they're a list of 1...
if not isinstance(jobid, list) and jobid is not None:
jobid = [jobid]
if isinstance(jobid, list):
opt = ["scontrol", "show", "job"]
sreturn = ""
for my_id in jobid:
sopt = opt + [str(my_id)]
sreturn = sreturn + run(sopt)[0] + "\n"
return sreturn
else:
sopt = ["squeue", "-h"]
if username is not None:
sopt += ["-u", username]
if jobid is not None:
sopt += ["--job="]
if isinstance(jobid, list):
sopt += ["'" + ",".join([str(i) for i in jobid]) + "'"]
else:
sopt += [str(jobid)]
if sformat is not None:
sopt += ["-o", "'" + sformat + "'"]
else:
if jobid is None and username is None:
sopt += ["-o", "'%i %u %P %j %U %D %C %m %l %t %M'"]
else:
sopt += ["-o", "'%i %j %u %M %t %P'"]
return run(sopt)[0]
