def getListOfJobStates(jobName, username=None, detailed=True):
command = "squeue --clusters=serial -u $(whoami) -l -h"
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("squeue failed (stderr: '" +
stderr + "')")
jobList = []
jobStates = []
for line in stdout.split('\n'):
if line.startswith("CLUSTER: serial"):
continue
line = line.rstrip('\n')
lineSplit = line.split()
try:
currentJobId = int(lineSplit[0])
currentJobStatus = JobStatus(currentJobId)
# name
name = lineSplit[2]
if name == jobName or jobName == None:
jobList.append(currentJobId)
jobStates.append(currentJobStatus)
# status
status = lineSplit[4]
currentJobStatus.setStatus(JobStatus.kUnknown, name=status)
if status == 'RUNNING':
currentJobStatus.setStatus(JobStatus.kRunning)
elif status == 'PENDING' or status == 'SUSPENDED' or status == 'COMPLETING' or status == 'COMPLETED' or status == 'COMPLETI':
currentJobStatus.setStatus(JobStatus.kWaiting)
elif status == 'CANCELLED' or status == 'FAILED' or status == 'TIMEOUT' or status == 'NODE_FAIL':
currentJobStatus.setStatus(JobStatus.kError)
else:
print "Unknown job status", status
# time
time_str = lineSplit[5]
try:
hours = 0.0
if '-' in time_str:
time_str = time_str.split('-')
hours += float(time_str[0]) * 24
time_str = time_str[1].split(':')
else:
time_str = time_str.split(':')
seconds = float(time_str[-1])
minutes = float(time_str[-2])
if (len(time_str) > 2):
hours += float(time_str[-3])
total_time = hours + minutes / 60.0 + seconds / 3600.0
currentJobStatus.setCpuTime(total_time, 0)
except ValueError:
raise batchelor.BatchelorException(
"parsing of squeue output to get time information failed. ({0})"
.format(lineSplit[5]))
except ValueError:
raise batchelor.BatchelorException(
"parsing of squeue output to get job id failed.")
return jobStates
def submitJob(config, command, outputFile, jobName, wd=None):
# check if only a certain amount of active jobs is allowd
if config.has_option(submoduleIdentifier(), "max_active_jobs"):
max_active_jobs = int(
config.get(submoduleIdentifier(), "max_active_jobs"))
i = 0
waitTime = 90
while len(getListOfActiveJobs(None)) >= max_active_jobs:
if i == 0:
sys.stdout.write("Waiting for free slots")
sys.stdout.flush()
time.sleep(waitTime)
# wait 1.5 min
i += 1
if i > 0:
sys.stdout.write("\r")
if wd == None:
wd = os.getcwd()
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
headerFileName = batchelor._getRealPath(
config.get(submoduleIdentifier(), "header_file"))
with open(fileName, 'w') as tempFile:
tempFile.write("#!/bin/bash\n\n")
tempFile.write("#SBATCH -D " + wd + "\n")
tempFile.write("#SBATCH -o " + outputFile + "\n")
tempFile.write("#SBATCH --time=" +
config.get(submoduleIdentifier(), "wall_clock_limit") +
"\n")
tempFile.write("#SBATCH --mem=" +
config.get(submoduleIdentifier(), "memory") + "\n")
if jobName is not None:
tempFile.write("#SBATCH -J " + jobName + "\n")
tempFile.write("#SBATCH --get-user-env \n")
tempFile.write("#SBATCH --export=NONE \n")
tempFile.write("#SBATCH --clusters=serial \n\n\n")
with open(headerFileName, 'r') as headerFile:
for line in headerFile:
if line.startswith("#!"):
continue
tempFile.write(line)
tempFile.write("\n\n")
tempFile.write(command)
cmnd = "sbatch " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(cmnd)
batchelor.runCommand("rm -f " + fileName)
if returncode != 0:
raise batchelor.BatchelorException("sbatch failed (stderr: '" +
stderr + "')")
jobId = stdout.split()[3]
try:
jobId = int(jobId)
except ValueError:
raise batchelor.BatchelorException(
'parsing output of sbatch to get job id failed.')
return jobId
def submitJob(config,
command,
outputFile,
jobName,
wd=None,
arrayStart=None,
arrayEnd=None,
arrayStep=None):
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
batchelor.runCommand("cp " + batchelor._getRealPath(
config.get(submoduleIdentifier(), "header_file")) + " " + fileName)
with open(fileName, 'a') as scriptFile:
scriptFile.write(command)
if arrayStart is not None:
if (jobName is None) or (len(jobName) is 0):
jobName = ''.join(random.sample(string.lowercase, 7))
jobName = jobName + "[" + str(arrayStart) + "-" + str(
arrayEnd) + ":" + str(arrayStep) + "]"
cmnd = "bsub "
cmnd += "" if jobName is None else ("-J " + jobName + " ")
cmnd += "-o " + outputFile + " "
cmnd += "-q " + config.get(submoduleIdentifier(), "queue") + " "
cmnd += "-R '"
cmnd += "-cwd '{0}'".format(wd) if wd else ""
cmnd += " select[type=" + config.get(submoduleIdentifier(), "type") + "]"
cmnd += " rusage[pool=" + config.get(submoduleIdentifier(), "pool") + "]"
try:
cmnd += " rusage[mem=" + config.get(submoduleIdentifier(),
"memory") + "]"
cmnd += " select[maxmem>" + config.get(submoduleIdentifier(),
"memory") + "]"
except ConfigParser.NoOptionError:
pass
cmnd += _getExcludedHostsString(config)
cmnd += "' "
cmnd += "< " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(cmnd)
if returncode != 0:
raise batchelor.BatchelorException("bsub failed (stderr: '" + stderr +
"')")
# example output: Job <533476534> is submitted to queue <1nd>.
jobId = stdout.lstrip("Job <")
jobId = jobId[:jobId.find(">")]
try:
jobId = int(jobId)
except ValueError:
raise batchelor.BatchelorException(
'parsing of bsub output to get job id failed.')
batchelor.runCommand('rm -f ' + fileName)
return jobId
def getListOfActiveJobs(jobName):
if jobName is None:
command = "llq -u `whoami`"
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("llq failed (stderr: '" +
stderr + "')")
if stdout == "llq: There is currently no job status to report.":
return []
stringList = [job.split()[0] for job in stdout.split('\n')[2:-2]]
jobList = []
try:
for item in stringList:
jobId = int(item[item.find(".") + 1:item.rfind(".")])
if jobId not in jobList:
jobList.append(jobId)
except ValueError:
raise batchelor.BatchelorException(
"parsing of llq output to get job id failed.")
return jobList
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
command = "llq -u `whoami` -m &> " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
batchelor.runCommand("rm -f " + fileName)
raise batchelor.BatchelorException("llq failed (stderr: '" + stderr +
"')")
jobList = []
currentJobId = -1
with open(fileName, 'r') as llqOutput:
for line in llqOutput:
line = line[:-1]
if line.startswith("===== Job Step mgmt."):
try:
currentJobId = int(line[line.find(".") +
1:line.rfind(".")])
except ValueError:
batchelor.runCommand("rm -f " + fileName)
raise batchelor.BatchelorException(
"parsing of llq output to get job id failed.")
line = ' '.join(line.split())
if line.startswith("Job Name: "):
if currentJobId < 0:
batchelor.runCommand("rm -f " + fileName)
raise batchelor.BatchelorException(
"parsing of llq output failed, got job name before job id."
)
name = line[10:]
if name == jobName:
jobList.append(currentJobId)
batchelor.runCommand("rm -f " + fileName)
return jobList
def submitJob(config, command, outputFile, jobName):
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
headerFileName = batchelor._getRealPath(config.get(submoduleIdentifier(), "header_file"))
with open(fileName, 'w') as tempFile:
tempFile.write("#!/bin/bash\n\n")
tempFile.write("#@ group = " + config.get(submoduleIdentifier(), "group") + "\n")
tempFile.write("#@ output = " + outputFile + "\n")
tempFile.write("#@ notification = " + config.get(submoduleIdentifier(), "notification") + "\n")
tempFile.write("#@ notify_user = "******"notify_user") + "\n")
tempFile.write("#@ node_usage = " + config.get(submoduleIdentifier(), "node_usage") + "\n")
tempFile.write("#@ wall_clock_limit = " + config.get(submoduleIdentifier(), "wall_clock_limit") + "\n")
tempFile.write("#@ resources = " + config.get(submoduleIdentifier(), "resources") + "\n")
tempFile.write("#@ job_type = " + config.get(submoduleIdentifier(), "job_type") + "\n")
tempFile.write("#@ class = " + config.get(submoduleIdentifier(), "job_type") + "\n")
if jobName is not None:
tempFile.write("#@ job_name = " + jobName + "\n")
tempFile.write("#@ queue\n\n\n")
with open(headerFileName, 'r') as headerFile:
for line in headerFile:
if line.startswith("#!"):
continue
tempFile.write(line)
tempFile.write("\n\n")
tempFile.write(command)
cmnd = "llsubmit - < " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(cmnd)
if returncode != 0:
batchelor.runCommand("rm -f " + fileName)
raise batchelor.BatchelorException("llsubmit failed (stderr: '" + stderr + "')")
# example output stdout:
# llsubmit: The job "mgmt.12309" has been submitted.
#
# example output stderr:
#
# llsubmit: Stdin job command file written to "/tmp/loadlx_stdin.27558.CdoVxX".
#
# INFO: Project: pr83mo
# INFO: Project's Expiration Date: 2015-01-31
# INFO: Budget: Total [cpuh] Used [cpuh] Credit [cpuh]
# INFO: 1350000 1011028 (75%) 338972 (25%)
#
# llsubmit: Processed command file through Submit Filter: "/lrz/loadl/filter/submit_filter_c2pap.pl".
jobId = stdout.split("\n")[0]
jobId = jobId[jobId.find('"mgmt.')+6:jobId.rfind('"')]
try:
jobId = int(jobId)
except ValueError:
batchelor.runCommand("rm -f " + fileName)
raise batchelor.BatchelorException('parsing of qsub output to get job id failed.')
batchelor.runCommand("rm -f " + fileName)
return jobId
def submitJob(config,
command,
outputFile,
jobName,
wd=None,
arrayStart=None,
arrayEnd=None,
arrayStep=None):
if wd:
raise batchelor.BatchelorException(
"Choosing the working directory is not jet implemented for {0}".
format(submoduleIdentifier()))
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
batchelor.runCommand("cp " + batchelor._getRealPath(
config.get(submoduleIdentifier(), "header_file")) + " " + fileName)
with open(fileName, 'a') as scriptFile:
scriptFile.write(command)
cmnd = "qsub "
cmnd += "-j y "
cmnd += "" if jobName is None else ("-N " + jobName + " ")
if arrayStart is not None:
cmnd += "-t " + str(arrayStart) + "-" + str(arrayEnd) + ":" + str(
arrayStep) + " "
cmnd += "-o " + outputFile + " "
cmnd += "-P " + config.get(submoduleIdentifier(), "project") + " "
cmnd += "-q " + config.get(submoduleIdentifier(), "queue") + " "
cmnd += "-l h_vmem=" + config.get(submoduleIdentifier(), "memory") + " "
cmnd += _getExcludedHostsString(config)
cmnd += "< " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(cmnd)
if returncode != 0:
raise batchelor.BatchelorException("qsub failed (stderr: '" + stderr +
"')")
# example output: "Your job 1601905 ("J2415c980b8") has been submitted"
if arrayStart is not None:
jobId = stdout.lstrip("Your job-array ")
jobId = jobId[:jobId.find('.')]
else:
jobId = stdout.lstrip("Your job ")
jobId = jobId[:jobId.find(' ')]
try:
jobId = int(jobId)
except ValueError:
raise batchelor.BatchelorException(
'parsing of qsub output to get job id failed.')
batchelor.runCommand("rm -f " + fileName)
return jobId
def getListOfActiveJobs(jobName):
command = "bjobs"
if not jobName is None:
command = command + " -J " + jobName
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("bjobs failed (stderr: '" + stderr +
"')")
if stdout == "":
return []
jobList = stdout.split('\n')[1:]
try:
return [int(job.split()[0]) for job in jobList]
except ValueError:
raise batchelor.BatchelorException(
"parsing of bjobs output to get job id failed.")
def resetErrorJobs(jobName):
for id in getListOfErrorJobs(jobName):
command = "qmod -cj " + str(id)
(returncode, stdout, stderr) = batchelor.runCommand(command)
if stdout.find('cleared error state of job') is -1:
raise batchelor.BatchelorException("qmod failed (stderr: '" +
stderr + "')")
return True
def deleteJobs(jobIds):
if not jobIds:
return True
command = "llcancel"
for jobId in jobIds:
command += " mgmt." + str(jobId)
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("llcancel failed (stderr: '" + stderr + "')")
return True
def deleteJobs(jobIds):
if not jobIds:
return True
command = "bkill"
for jobId in jobIds:
command += ' ' + str(jobId)
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
if not 'Job has already finished' in stderr:
raise batchelor.BatchelorException("bkill failed (stderr: '" +
stderr + "')")
return True
def deleteJobs(jobIds):
if not jobIds:
return True
command = "condor_rm"
for jobId in jobIds:
command += ' ' + str(jobId)
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
if not 'Couldn\'t find/remove all jobs matching constraint' in stderr:
raise batchelor.BatchelorException("condor_rm failed (stderr: '" +
stderr + "')")
return True
def getListOfErrorJobs(jobName):
listOfActiveJobs = getListOfActiveJobs(jobName)
command = "qstat"
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("qstat failed (stderr: '" + stderr +
"')")
qstatLines = stdout.split('\n')[2:]
listOfErrorJobs = []
for line in qstatLines:
lineList = line.split()
jobId = -1
try:
jobId = int(lineList[0])
except ValueError:
raise batchelor.BatchelorException(
"parsing of qstat output to get job id failed.")
if jobId not in listOfActiveJobs:
continue
if lineList[4] == "Eqw":
listOfErrorJobs.append(jobId)
return listOfErrorJobs
def getListOfActiveJobs(jobName):
command = "condor_q -format \"%d.\" ClusterId -format \"%d\n\" ProcId "
if jobName:
command += "-constraint 'JobBatchName == \"{0}\"' ".format(jobName)
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("condor_q failed (stderr: '" +
stderr + "')")
if stdout == "":
return []
jobList = stdout.split('\n')
jobs = []
for job in jobList:
job = job.split()
if len(job) > 0:
try:
jobID = int(job[0].rstrip(".0"))
jobs.append(jobID)
except ValueError:
raise batchelor.BatchelorException(
"Cannot parse return of condor_q (stdout: '" + stdout +
"')")
return jobs
def getListOfActiveJobs(jobName):
if jobName is None:
command = "qstat"
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("qstat failed (stderr: '" +
stderr + "')")
if stdout == "":
return []
jobList = stdout.split('\n')[2:]
try:
return [int(job.split()[0]) for job in jobList]
except ValueError:
raise batchelor.BatchelorException(
"parsing of qstat output to get job id failed.")
command = "qstat -j " + jobName
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
if stderr.split(
'\n'
)[0][:
-1] == "Following jobs do not exist or permissions are not sufficient:":
return []
raise batchelor.BatchelorException("qstat failed (stderr: '" + stderr +
"')")
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
command = "qstat -xml -j " + jobName + " > " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("qstat failed (stderr: '" + stderr +
"')")
batchelor.runCommand(
"awk '/<\?xml version='\"'\"'1.0'\"'\"'\?>/{n++}{print >\"" +
fileName + "\" n \".awkOut\" }' " + fileName)
batchelor.runCommand("rm -f " + fileName)
xmlFiles = glob.glob(fileName + "*.awkOut")
jobIds = []
for xmlFile in xmlFiles:
tree = ElementTree.parse(xmlFile)
root = tree.getroot()
batchelor.runCommand("rm -f " + xmlFile)
for child in root[0]:
jobIdList = child.findall("JB_job_number")
if len(jobIdList) != 1:
raise batchelor.BatchelorException(
"parsing xml from qstat failed")
try:
jobId = int(jobIdList[0].text)
except ValueError:
raise batchelor.BatchelorException(
"parsing int from xml from qstat failed")
jobIds.append(jobId)
return jobIds
def run(self):
while True:
try:
jobId = queue.get(timeout=2)
except Queue.Empty:
with guard:
if aux[1]:
break
else:
continue
with guard:
knownJobIds = [job.jobId for job in jobs]
if not jobId in knownJobIds:
# might actually happen if a job is deleted
continue
i = knownJobIds.index(jobId)
jobs[i].running = True
outputFile = jobs[i].outputFile
command = jobs[i].command
cmdFile = tempfile.NamedTemporaryFile(delete=False)
for line in command:
cmdFile.write(line)
cmdFile.close()
logFile = open(outputFile, "w")
p = subprocess.Popen([self.shell, cmdFile.name],
stdout=logFile,
stderr=subprocess.STDOUT,
preexec_fn=lambda: os.setpgid(0, 0))
jobs[i].runningProcess = p
p.wait()
logFile.close()
os.unlink(cmdFile.name)
with guard:
knownJobIds = [job.jobId for job in jobs]
if not jobId in knownJobIds:
raise batchelor.BatchelorException(
"Job ID {0} finished, but already removed from list of jobs."
.format(jobId))
i = knownJobIds.index(jobId)
del jobs[i]
queue.task_done()
def run(self):
while True:
try:
jobId = queue.get(timeout=2)
except Queue.Empty:
with guard:
if aux[1]:
break
else:
continue
with guard:
for i in range(len(jobs)):
if jobs[i].jobId == jobId:
break
if jobs[i].jobId != jobId:
continue # might actually happen if a
# job is deleted
jobs[i].running = True
outputFile = jobs[i].outputFile
command = jobs[i].command
cmdFile = tempfile.NamedTemporaryFile(delete=False)
for line in command:
cmdFile.write(line)
cmdFile.close()
with open(outputFile, "w") as logFile:
subprocess.call([self.shell, cmdFile.name],
stdout=logFile,
stderr=subprocess.STDOUT,
preexec_fn=lambda: os.setpgid(0, 0))
os.unlink(cmdFile.name)
with guard:
for i in range(len(jobs)):
if jobs[i].jobId == jobId:
break
if jobs[i].jobId != jobId:
raise batchelor.BatchelorException(
"Job ID {0} finished, but already removed from list of jobs."
.format(jobId))
del jobs[i]
queue.task_done()
def submitJobs(config, newJobs):
if len(newJobs) == 0:
return []
poolJobsArgs = []
for job in newJobs:
poolJobsArgs.append([config] + job)
try:
nParallelSubmissions = int(config.get(submoduleIdentifier(), "parallel_submissions"))
except ValueError:
raise batchelor.BatchelorException('option \'parallel_submissions\' in config file does not seem to be an int.')
if nParallelSubmissions < 0:
nParallelSubmissions = len(newJobs)
nParallelSubmissions = min(nParallelSubmissions, len(newJobs))
pool = multiprocessing.Pool(processes = nParallelSubmissions)
jobIds = pool.map(_wrapSubmitJob, poolJobsArgs, 1)
pool.close()
pool.join()
return jobIds
def _submitJob(config, command, outputFile, jobName, wd=None, nTasks=None):
# check if only a certain amount of active jobs is allowd
if config.has_option(submoduleIdentifier(), "max_active_jobs"):
max_active_jobs = int(
config.get(submoduleIdentifier(), "max_active_jobs"))
i = 0
waitTime = 90
while True:
try:
nRunningJobs = len(getListOfActiveJobs(None))
except batchelor.BatchelorException:
nRunningJobs = max_active_jobs
if nRunningJobs < max_active_jobs:
break
if i == 0:
sys.stdout.write("Waiting for free slots")
sys.stdout.flush()
time.sleep(waitTime)
# wait 1.5 min
i += 1
if i > 0:
sys.stdout.write("\r")
if wd == None:
wd = os.getcwd()
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
headerFileName = batchelor._getRealPath(
config.get(submoduleIdentifier(), "header_file"))
with open(fileName, 'w') as tempFile:
tempFile.write("#!/bin/bash\n\n")
tempFile.write("#SBATCH -D " + wd + "\n")
tempFile.write("#SBATCH -o " + outputFile + "\n")
tempFile.write("#SBATCH --time=" +
config.get(submoduleIdentifier(), "wall_clock_limit") +
"\n")
if config.get(submoduleIdentifier(), "clusters") != 'mpp3':
tempFile.write("#SBATCH --mem-per-cpu=" +
config.get(submoduleIdentifier(), "memory") + "\n")
if jobName is not None:
tempFile.write("#SBATCH -J " + jobName + "\n")
tempFile.write("#SBATCH --get-user-env \n")
tempFile.write("#SBATCH --export=NONE \n")
if nTasks is not None:
if config.get(submoduleIdentifier(), "clusters") != 'mpp3':
tempFile.write("#SBATCH --ntasks={0:d} \n".format(nTasks))
else:
tempFile.write("#SBATCH --nodes={0:d} \n".format(
(nTasks + 63) // 64))
tempFile.write("#SBATCH --ntasks-per-node={0} \n".format(
config.get(submoduleIdentifier(), "n_tasks_per_node")))
tempFile.write("#SBATCH --clusters={0}\n".format(
config.get(submoduleIdentifier(), "clusters")))
if config.get(submoduleIdentifier(),
"clusters") not in ['cm2_tiny', 'mpp3']:
tempFile.write("#SBATCH --partition={0}\n\n".format(
config.get(submoduleIdentifier(), "partition")))
if config.get(submoduleIdentifier(),
"clusters") == 'cm2' or config.get(
submoduleIdentifier(), "clusters") == 'c2pap':
tempFile.write("#SBATCH --qos={0}\n\n".format(
config.get(submoduleIdentifier(), "partition")))
tempFile.write("module load slurm_setup \n\n\n")
with open(headerFileName, 'r') as headerFile:
for line in headerFile:
if line.startswith("#!"):
continue
tempFile.write(line)
tempFile.write("\n\n")
tempFile.write(command)
cmnd = "sbatch " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(cmnd)
batchelor.runCommand("rm -f " + fileName)
if returncode != 0:
raise batchelor.BatchelorException("sbatch failed (stderr: '" +
stderr + "')")
jobId = stdout.split()[3]
try:
jobId = int(jobId)
except ValueError:
raise batchelor.BatchelorException(
'parsing output of sbatch to get job id failed.')
return jobId
def submitJob(config,
command,
outputFile,
jobName,
wd=None,
arrayStart=None,
arrayEnd=None,
arrayStep=None):
if arrayStart is not None or arrayEnd is not None or arrayStep is not None:
raise BatchelorException(
"Array jobs are not (yet) implementet for CERNs HTCondor system")
filesDir = os.path.join(os.getcwd(), '.log')
if " " in filesDir:
raise BatchelorException(
"Cannot handle submit directories with whitespaces")
if not os.path.exists(filesDir):
os.makedirs(filesDir)
(fileDescriptor, submitFileName) = tempfile.mkstemp(dir=filesDir,
prefix='submitFiles_',
suffix='.submit')
os.close(fileDescriptor)
atexit.register(lambda: os.remove(submitFileName))
(fileDescriptor, scriptFileName) = tempfile.mkstemp(dir=filesDir,
prefix='scriptFiles_',
suffix='.sh')
os.close(fileDescriptor)
atexit.register(lambda: os.remove(scriptFileName))
os.chmod(scriptFileName, 0755)
batchelor.runCommand("cp " + batchelor._getRealPath(
config.get(submoduleIdentifier(), "header_file")) + " " +
scriptFileName)
with open(scriptFileName, 'a') as scriptFile:
scriptFile.write(command)
with open(submitFileName, 'w') as submitFile:
outputFile = os.path.abspath(outputFile)
submitFile.write("executable = {0}\n".format(scriptFileName))
if outputFile:
submitFile.write("output = {0}\n".format(outputFile))
submitFile.write("log = {0}.condor\n".format(outputFile))
submitFile.write("error = {0}.err\n".format(outputFile))
submitFile.write(
"should_transfer_files = NO\n") # Disable file transport
submitFile.write("request_cpus = 1\n")
submitFile.write("request_memory = {0}\n".format(
config.get(submoduleIdentifier(), "memory")))
submitFile.write("request_disk = {0}\n".format(
config.get(submoduleIdentifier(), "disk")))
submitFile.write("+JobFlavour = \"{0}\"\n".format(
config.get(submoduleIdentifier(), "flavour")))
submitFile.write("queue 1\n")
cmnd = "condor_submit '{0}'".format(submitFileName)
if jobName:
cmnd += " -batch-name {0} ".format(jobName)
kwargs = {}
if wd:
kwargs['wd'] = wd
(returncode, stdout, stderr) = batchelor.runCommand(cmnd, **kwargs)
if returncode != 0:
raise batchelor.BatchelorException("condor_submit failed (stderr: '" +
stderr + "')")
jobId = stdout.split('\n')[1].split()[5].rstrip(".")
try:
jobId = int(jobId)
except ValueError:
raise batchelor.BatchelorException(
'parsing of condor_submit output to get job id failed.')
return jobId
def getListOfJobStates(select_jobIDs, username):
# get list of all jobs
if username == None:
command = "qstat"
else:
command = "qstat -u {0}".format(username)
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("qstat failed (stderr: '" + stderr +
"')")
if stdout == "":
return []
jobList = stdout.split('\n')[2:]
try:
jobIDs = [int(job.split()[0]) for job in jobList]
jobStates = [job.split()[4] for job in jobList]
except ValueError:
raise batchelor.BatchelorException(
"parsing of qstat output to get job id failed.")
list_of_states = []
for i, jobID in enumerate(jobIDs):
if select_jobIDs == None or jobID in select_jobIDs:
job_status = JobStatus(jobID)
job_status.setStatus(JobStatus.kUnknown, name=jobStates[i])
if jobStates[i] == 'qw' or jobStates[i] == 'hqw':
job_status.setStatus(JobStatus.kWaiting)
elif jobStates[i] == 't':
job_status.setStatus(JobStatus.kTransmitting)
elif jobStates[i] == 'd' or jobStates[i] == 'dr' or jobStates[
i] == 'dt':
job_status.setStatus(JobStatus.kDeletion)
elif jobStates[i] == 'Eq':
job_status.setStatus(JobStatus.kError)
elif jobStates[i] == 'r' or jobStates[i] == 'hr':
# get detailed job information
command = "qstat -xml -j {0}".format(jobID)
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException(
"qstat failed (stderr: '" + stderr + "')")
elif 'unknown_jobs' in stdout:
continue
# the job has been ended between the qstat command and now
else:
try:
root = ElementTree.fromstring(stdout)
for child in root[0]:
for task in child.findall('JB_ja_tasks'):
for sublist in task.findall('ulong_sublist'):
task_number = sublist.findall(
'JAT_task_number')
if task_number:
task_number = int(task_number[0].text)
job_status.setStatus(
JobStatus.kRunning)
for usage_list in sublist.findall(
'JAT_scaled_usage_list'):
for scaled in usage_list.findall(
'scaled'):
name = scaled.findall(
'UA_name')[0].text
value = scaled.findall(
'UA_value')[0].text
if name == 'cpu':
job_status.setCpuTime(
float(value) / 3600.0,
task_number)
elif name == 'vmem':
job_status.setMemoryUsage(
float(value) /
(1024.0)**3,
task_number)
except xml.etree.ElementTree.ParseError as e:
raise batchelor.BatchelorException(
"xml-parser could not parse output of qstat -xml -j {0}: {1}"
.format(jobID, e))
# end of parsing through the xml tree
list_of_states.append(job_status)
# end of if jobs belongs to the selected jobs
# end of loop over all jobs
return list_of_states
def getListOfJobStates(jobName, username=None, detailed=True):
if detailed:
command = "llq -u `whoami` -m -x"
else:
command = "llq -u `whoami` -m"
(returncode, stdout, stderr) = batchelor.runCommand(command)
if returncode != 0:
raise batchelor.BatchelorException("llq failed (stderr: '" + stderr +
"')")
jobList = []
jobStates = []
currentJobId = -1
currentJobStatus = None
for line in stdout.split('\n'):
line = line.rstrip('\n')
if line.startswith("===== Job Step mgmt."):
try:
currentJobId = int(line[line.find(".") + 1:line.rfind(".")])
currentJobStatus = JobStatus(currentJobId)
except ValueError:
raise batchelor.BatchelorException(
"parsing of llq output to get job id failed.")
line = ' '.join(line.split())
if line.startswith("Job Name: "):
if currentJobId < 0:
raise batchelor.BatchelorException(
"parsing of llq output failed, got job name before job id."
)
name = line[10:]
if name == jobName or jobName == None:
jobList.append(currentJobId)
jobStates.append(currentJobStatus)
elif line.startswith("Step Virtual Memory: "):
if currentJobId < 0:
raise batchelor.BatchelorException(
"parsing of llq output failed, got job name before job id."
)
try:
parsed = line.lstrip().lstrip('Step Virtual Memory:').split()
currentJobStatus.setMemoryUsage(
float(parsed[0]) * _kMemoryUnits[parsed[1]], 0)
except ValueError:
raise batchelor.BatchelorException(
"parsing of llq output to get job id failed.")
elif line.startswith("Status: "):
if currentJobId < 0:
raise batchelor.BatchelorException(
"parsing of llq output failed, got job name before job id."
)
else:
status = line.lstrip().lstrip("Status: ")
currentJobStatus.setStatus(JobStatus.kUnknown, name=status)
if status == 'Running':
currentJobStatus.setStatus(JobStatus.kRunning)
elif status == 'I' or status == 'Idle' or status == 'Pending':
currentJobStatus.setStatus(JobStatus.kWaiting)
elif status == 'Submission Error' or status == 'Terminated' or status == 'Removed' or status == 'Remove Pending':
currentJobStatus.setStatus(JobStatus.kError)
elif line.startswith("Step User Time: "):
if currentJobId < 0:
raise batchelor.BatchelorException(
"parsing of llq output failed, got job name before job id."
)
time_str = line.lstrip().lstrip("Step User Time:").split(':')
try:
hours = float(time_str[0])
minuts = float(time_str[1])
seconds = float(time_str[2])
total_time = hours + minuts / 60.0 + seconds / 3600.0
currentJobStatus.setCpuTime(total_time, 0)
except ValueError:
raise batchelor.BatchelorException(
"parsing of llq output to get job id failed.")
return jobStates
def submitJob(config,
command,
outputFile,
jobName,
wd=None,
arrayStart=None,
arrayEnd=None,
arrayStep=None,
priority=None,
ompNumThreads=None):
# some checks of the job-settings
if wd and os.path.realpath(wd).count(
os.path.realpath(os.path.expanduser('~'))):
raise batchelor.BatchelorException(
"The given working-directory is in your home-folder which is no allowed at E18: '{0}'"
.format(wd))
if os.path.realpath(outputFile).count(
os.path.realpath(os.path.expanduser('~'))):
raise batchelor.BatchelorException(
"The given output-file is in your home-folder which is no allowed at E18: '{0}'"
.format(outputFile))
if priority:
priority = max(int(-1024 + 2048 * (priority + 1.0) / 2.0), -1023)
(fileDescriptor, fileName) = tempfile.mkstemp()
os.close(fileDescriptor)
batchelor.runCommand("cp " + batchelor._getRealPath(
config.get(submoduleIdentifier(), "header_file")) + " " + fileName)
with open(fileName, 'a') as scriptFile:
if ompNumThreads is not None:
scriptFile.write(
"export OMP_NUM_THREADS={0}\n".format(ompNumThreads))
scriptFile.write(command)
cmnd = "qsub "
cmnd += "-j y "
cmnd += "-b no "
cmnd += "-m n "
cmnd += "" if jobName is None else ("-N " + jobName + " ")
if arrayStart is not None:
cmnd += "-t " + str(arrayStart) + "-" + str(arrayEnd) + ":" + str(
arrayStep) + " "
cmnd += "-o '" + outputFile + "' "
cmnd += "-wd '" + ("/tmp/" if not wd else wd) + "' "
if config.has_option(submoduleIdentifier(), "shortqueue") and config.get(
submoduleIdentifier(),
"shortqueue") in [1, "1", "TRUE", "true", "True"]:
cmnd += "-l short=1 "
elif config.has_option(submoduleIdentifier(), "longqueue") and config.get(
submoduleIdentifier(),
"longqueue") in [1, "1", "TRUE", "true", "True"]:
cmnd += "-l long=1 "
else:
cmnd += "-l medium=1 "
cmnd += "-l h_pmem=" + config.get(submoduleIdentifier(), "memory") + " "
cmnd += "-l arch=" + config.get(submoduleIdentifier(), "arch") + " "
cmnd += _getExcludedHostsString(config)
cmnd += "-p {0} ".format(priority) if priority else ""
cmnd += "-pe mt {0} ".format(
ompNumThreads) if ompNumThreads is not None else ""
cmnd += "< " + fileName
(returncode, stdout, stderr) = batchelor.runCommand(cmnd)
if returncode != 0:
raise batchelor.BatchelorException("qsub failed (stderr: '" + stderr +
"')")
# example output: "Your job 1601905 ("J2415c980b8") has been submitted"
if arrayStart is not None:
jobId = stdout.lstrip("Your job-array ")
jobId = jobId[:jobId.find('.')]
else:
jobId = stdout.lstrip("Your job ")
jobId = jobId[:jobId.find(' ')]
try:
jobId = int(jobId)
except ValueError:
raise batchelor.BatchelorException(
'parsing of qsub output to get job id failed.')
batchelor.runCommand("rm -f " + fileName)
return jobId
def getListOfErrorJobs(jobName=None):
raise batchelor.BatchelorException("not implemented")
