def _buildExecutor(self):
"""
Creates and returns an ExecutorInfo instance representing our executor implementation.
"""
# The executor program is installed as a setuptools entry point by setup.py
info = mesos_pb2.ExecutorInfo()
info.name = "toil"
info.command.value = resolveEntryPoint('_toil_mesos_executor')
info.executor_id.value = "toil-%i" % os.getpid()
info.source = pwd.getpwuid(os.getuid()).pw_name
return info
def issueJob(self, jobStoreID, memory, cores, disk, preemptable):
"""
Add a job to the queue of jobs
"""
self.jobsIssued += 1
jobCommand = ' '.join((resolveEntryPoint('_toil_worker'), self.jobStoreString, jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobCommand, memory, cores, disk, preemptable)
self.jobBatchSystemIDToIssuedJob[jobBatchSystemID] = IssuedJob(jobStoreID, memory, cores, disk, preemptable)
logger.debug("Issued job with job store ID: %s and job batch system ID: "
"%s and cores: %.2f, disk: %.2f, and memory: %.2f",
jobStoreID, str(jobBatchSystemID), cores, disk, memory)
def _buildExecutor(self):
"""
Creates and returns an ExecutorInfo instance representing our executor implementation.
"""
# The executor program is installed as a setuptools entry point by setup.py
info = mesos_pb2.ExecutorInfo()
info.name = "toil"
info.command.value = resolveEntryPoint('_toil_mesos_executor')
info.executor_id.value = "toil-%i" % os.getpid()
info.source = pwd.getpwuid(os.getuid()).pw_name
return info
def issueJob(self, jobStoreID, memory, cores, disk, preemptable):
"""
Add a job to the queue of jobs
"""
self.jobsIssued += 1
jobCommand = ' '.join((resolveEntryPoint('_toil_worker'), self.jobStoreLocator, jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobCommand, memory, cores, disk, preemptable)
self.jobBatchSystemIDToIssuedJob[jobBatchSystemID] = IssuedJob(jobStoreID, memory, cores, disk, preemptable)
logger.debug("Issued job with job store ID: %s and job batch system ID: "
"%s and cores: %.2f, disk: %.2f, and memory: %.2f",
jobStoreID, str(jobBatchSystemID), cores, disk, memory)
def issueJob(self, jobStoreID, memory, cores, disk):
"""
Add a job to the queue of jobs
"""
self.jobsIssued += 1
jobCommand = ' '.join((resolveEntryPoint('_toil_worker'), self.jobStoreString, jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobCommand, memory, cores, disk)
self.jobBatchSystemIDToJobStoreIDHash[jobBatchSystemID] = jobStoreID
logger.debug("Issued job with job store ID: %s and job batch system ID: "
"%s and cores: %i, disk: %i, and memory: %i",
jobStoreID, str(jobBatchSystemID), cores, disk, memory)
def issueJob(self, issuedJob):
"""
Add a job to the queue of jobs.
"""
jobCommand = ' '.join((resolveEntryPoint('_toil_worker'),
self.config.jobStore, issuedJob.jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobCommand, issuedJob.memory,
issuedJob.cores, issuedJob.disk, issuedJob.preemptable)
self.jobBatchSystemIDToIssuedJob[jobBatchSystemID] = issuedJob
logger.debug("Issued job with job store ID: %s and job batch system ID: "
"%s and cores: %i, disk: %i, and memory: %i",
issuedJob.jobStoreID, str(jobBatchSystemID), issuedJob.cores,
issuedJob.disk, issuedJob.memory)
def issueJob(self, issuedJob):
"""
Add a job to the queue of jobs.
"""
jobCommand = ' '.join((resolveEntryPoint('_toil_worker'),
self.config.jobStore, issuedJob.jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobCommand, issuedJob.memory,
issuedJob.cores, issuedJob.disk, issuedJob.preemptable)
self.jobBatchSystemIDToIssuedJob[jobBatchSystemID] = issuedJob
logger.debug("Issued job with job store ID: %s and job batch system ID: "
"%s and cores: %i, disk: %i, and memory: %i",
issuedJob.jobStoreID, str(jobBatchSystemID), issuedJob.cores,
issuedJob.disk, issuedJob.memory)
def issueJob(self, jobNode):
"""
Add a job to the queue of jobs
"""
if jobNode.preemptable:
self.preemptableJobsIssued += 1
jobNode.command = ' '.join((resolveEntryPoint('_toil_worker'),
self.jobStoreLocator, jobNode.jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobNode)
self.jobBatchSystemIDToIssuedJob[jobBatchSystemID] = jobNode
logger.debug("Issued job with job store ID: %s and job batch system ID: "
"%s and cores: %.2f, disk: %.2f, and memory: %.2f",
jobNode.jobStoreID, str(jobBatchSystemID), jobNode.cores,
jobNode.disk, jobNode.memory)
def issueJob(self, jobNode):
"""
Add a job to the queue of jobs
"""
jobNode.command = ' '.join((resolveEntryPoint('_toil_worker'),
self.jobStoreLocator, jobNode.jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobNode)
self.jobBatchSystemIDToIssuedJob[jobBatchSystemID] = jobNode
if jobNode.preemptable:
# len(jobBatchSystemIDToIssuedJob) should always be greater than or equal to preemptableJobsIssued,
# so increment this value after the job is added to the issuedJob dict
self.preemptableJobsIssued += 1
cur_logger = (logger.debug if jobNode.jobName.startswith(self.debugJobNames)
else logger.info)
cur_logger("Issued job %s with job batch system ID: "
"%s and cores: %s, disk: %s, and memory: %s",
jobNode, str(jobBatchSystemID), int(jobNode.cores),
bytes2human(jobNode.disk), bytes2human(jobNode.memory))
def issueJob(self, jobNode):
"""
Add a job to the queue of jobs
"""
jobNode.command = ' '.join((resolveEntryPoint('_toil_worker'),
self.jobStoreLocator, jobNode.jobStoreID))
jobBatchSystemID = self.batchSystem.issueBatchJob(jobNode)
self.jobBatchSystemIDToIssuedJob[jobBatchSystemID] = jobNode
if jobNode.preemptable:
# len(jobBatchSystemIDToIssuedJob) should always be greater than or equal to preemptableJobsIssued,
# so increment this value after the job is added to the issuedJob dict
self.preemptableJobsIssued += 1
cur_logger = (logger.debug if jobNode.jobName.startswith(CWL_INTERNAL_JOBS)
else logger.info)
cur_logger("Issued job %s with job batch system ID: "
"%s and cores: %s, disk: %s, and memory: %s",
jobNode, str(jobBatchSystemID), int(jobNode.cores),
bytes2human(jobNode.disk), bytes2human(jobNode.memory))
def toilMain(self):
return resolveEntryPoint("toil")
def _toilSort(self, jobStore, batchSystem,
lines=defaultLines, N=defaultN, testNo=1, lineLen=defaultLineLen):
"""
Generate a file consisting of the given number of random lines, each line of the given
length. Sort the file with Toil by splitting the file recursively until each part is less
than the given number of bytes, sorting each part and merging them back together. Then
verify the result.
:param jobStore: a job store string
:param batchSystem: the name of the batch system
:param lines: the number of random lines to generate
:param N: the size in bytes of each split
:param testNo: the number of repeats of this test
:param lineLen: the length of each random line in the file
"""
for test in xrange(testNo):
try:
# Specify options
options = Job.Runner.getDefaultOptions(jobStore)
options.logLevel = getLogLevelString()
options.retryCount = 2
options.batchSystem = batchSystem
options.clean = "never"
options.badWorker = 0.5
options.badWorkerFailInterval = 0.05
# Make the file to sort
tempSortFile = os.path.join(self.tempDir, "fileToSort.txt")
makeFileToSort(tempSortFile, lines=lines, lineLen=lineLen)
# First make our own sorted version
with open(tempSortFile, 'r') as fileHandle:
l = fileHandle.readlines()
l.sort()
# Make the first job
firstJob = Job.wrapJobFn(setup, tempSortFile, N, memory=sortMemory)
# Check we get an exception if we try to restart a workflow that doesn't exist
options.restart = True
try:
Job.Runner.startToil(firstJob, options)
self.fail()
except JobStoreCreationException:
pass
options.restart = False
# Now actually run the workflow
try:
Job.Runner.startToil(firstJob, options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
# Check we get an exception if we try to run without restart on an existing store
try:
Job.Runner.startToil(firstJob, options)
self.fail()
except JobStoreCreationException:
pass
options.restart = True
# This loop tests the restart behavior
totalTrys = 1
while i != 0:
options.useExistingOptions = random.random() > 0.5
try:
Job.Runner.startToil(firstJob, options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
if totalTrys > 16: #p(fail after this many restarts) = 0.5**32
self.fail() #Exceeded a reasonable number of restarts
totalTrys += 1
# Now check that if you try to restart from here it will raise an exception
# indicating that there are no jobs remaining in the workflow.
try:
Job.Runner.startToil(firstJob, options)
self.fail()
except JobException:
pass
# Now check the file is properly sorted..
with open(tempSortFile, 'r') as fileHandle:
l2 = fileHandle.readlines()
self.assertEquals(l, l2)
finally:
subprocess.check_call([resolveEntryPoint('toil'), 'clean', jobStore])
def _toilSort(self, jobStoreLocator, batchSystem,
lines=defaultLines, N=defaultN, testNo=1, lineLen=defaultLineLen,
retryCount=2, badWorker=0.5, downCheckpoints=False, disableCaching=False):
"""
Generate a file consisting of the given number of random lines, each line of the given
length. Sort the file with Toil by splitting the file recursively until each part is less
than the given number of bytes, sorting each part and merging them back together. Then
verify the result.
:param jobStoreLocator: The location of the job store.
:param batchSystem: the name of the batch system
:param lines: the number of random lines to generate
:param N: the size in bytes of each split
:param testNo: the number of repeats of this test
:param lineLen: the length of each random line in the file
"""
for test in xrange(testNo):
try:
# Specify options
options = Job.Runner.getDefaultOptions(jobStoreLocator)
options.logLevel = getLogLevelString()
options.retryCount = retryCount
options.batchSystem = batchSystem
options.clean = "never"
options.badWorker = badWorker
options.badWorkerFailInterval = 0.05
options.disableCaching = disableCaching # FIXME maybe this line should be deleted
options.downCheckpoints = downCheckpoints
options.N = N
# Make the file to sort
tempSortFile = os.path.join(self.tempDir, "fileToSort.txt")
makeFileToSort(tempSortFile, lines=lines, lineLen=lineLen)
options.fileToSort = tempSortFile
# First make our own sorted version
with open(tempSortFile, 'r') as fileHandle:
l = fileHandle.readlines()
l.sort()
# Check we get an exception if we try to restart a workflow that doesn't exist
options.restart = True
with self.assertRaises(NoSuchJobStoreException):
main(options)
options.restart = False
# Now actually run the workflow
try:
main(options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
# Check we get an exception if we try to run without restart on an existing store
with self.assertRaises(JobStoreExistsException):
main(options)
options.restart = True
# This loop tests the restart behavior
totalTrys = 1
while i != 0:
options.useExistingOptions = random.random() > 0.5
try:
main(options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
if totalTrys > 32: # p(fail after this many restarts) = 0.5**32
self.fail('Exceeded a reasonable number of restarts')
totalTrys += 1
# Now check that if you try to restart from here it will raise an exception
# indicating that there are no jobs remaining in the workflow.
with self.assertRaises(JobException):
main(options)
# Now check the file is properly sorted..
with open(tempSortFile, 'r') as fileHandle:
l2 = fileHandle.readlines()
self.assertEquals(l, l2)
finally:
subprocess.check_call([resolveEntryPoint('toil'), 'clean', jobStoreLocator])
def _toilSort(self,
jobStoreLocator,
batchSystem,
lines=defaultLines,
N=defaultN,
testNo=1,
lineLen=defaultLineLen,
retryCount=2,
badWorker=0.5,
downCheckpoints=False,
disableCaching=False):
"""
Generate a file consisting of the given number of random lines, each line of the given
length. Sort the file with Toil by splitting the file recursively until each part is less
than the given number of bytes, sorting each part and merging them back together. Then
verify the result.
:param jobStoreLocator: The location of the job store.
:param batchSystem: the name of the batch system
:param lines: the number of random lines to generate
:param N: the size in bytes of each split
:param testNo: the number of repeats of this test
:param lineLen: the length of each random line in the file
"""
for test in xrange(testNo):
try:
# Specify options
options = Job.Runner.getDefaultOptions(jobStoreLocator)
options.logLevel = getLogLevelString()
options.retryCount = retryCount
options.batchSystem = batchSystem
options.clean = "never"
options.badWorker = badWorker
options.badWorkerFailInterval = 0.05
options.disableCaching = disableCaching # FIXME maybe this line should be deleted
options.downCheckpoints = downCheckpoints
options.N = N
# Make the file to sort
tempSortFile = os.path.join(self.tempDir, "fileToSort.txt")
makeFileToSort(tempSortFile, lines=lines, lineLen=lineLen)
options.fileToSort = tempSortFile
# First make our own sorted version
with open(tempSortFile, 'r') as fileHandle:
l = fileHandle.readlines()
l.sort()
# Check we get an exception if we try to restart a workflow that doesn't exist
options.restart = True
with self.assertRaises(NoSuchJobStoreException):
main(options)
options.restart = False
# Now actually run the workflow
try:
main(options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
# Check we get an exception if we try to run without restart on an existing store
with self.assertRaises(JobStoreExistsException):
main(options)
options.restart = True
# This loop tests the restart behavior
totalTrys = 1
while i != 0:
options.useExistingOptions = random.random() > 0.5
try:
main(options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
if totalTrys > 32: # p(fail after this many restarts) = 0.5**32
self.fail(
'Exceeded a reasonable number of restarts')
totalTrys += 1
# Now check that if you try to restart from here it will raise an exception
# indicating that there are no jobs remaining in the workflow.
with self.assertRaises(JobException):
main(options)
# Now check the file is properly sorted..
with open(tempSortFile, 'r') as fileHandle:
l2 = fileHandle.readlines()
self.assertEquals(l, l2)
finally:
subprocess.check_call(
[resolveEntryPoint('toil'), 'clean', jobStoreLocator])
def _toilSort(self,
jobStore,
batchSystem,
lines=defaultLines,
N=defaultN,
testNo=1,
lineLen=defaultLineLen):
"""
Generate a file consisting of the given number of random lines, each line of the given
length. Sort the file with Toil by splitting the file recursively until each part is less
than the given number of bytes, sorting each part and merging them back together. Then
verify the result.
:param jobStore: a job store string
:param batchSystem: the name of the batch system
:param lines: the number of random lines to generate
:param N: the size in bytes of each split
:param testNo: the number of repeats of this test
:param lineLen: the length of each random line in the file
"""
for test in xrange(testNo):
try:
# Specify options
options = Job.Runner.getDefaultOptions(jobStore)
options.logLevel = getLogLevelString()
options.retryCount = 2
options.batchSystem = batchSystem
options.clean = "never"
options.badWorker = 0.5
options.badWorkerFailInterval = 0.05
# Make the file to sort
tempSortFile = os.path.join(self.tempDir, "fileToSort.txt")
makeFileToSort(tempSortFile, lines=lines, lineLen=lineLen)
# First make our own sorted version
with open(tempSortFile, 'r') as fileHandle:
l = fileHandle.readlines()
l.sort()
# Make the first job
firstJob = Job.wrapJobFn(setup,
tempSortFile,
N,
memory=sortMemory)
# Check we get an exception if we try to restart a workflow that doesn't exist
options.restart = True
try:
Job.Runner.startToil(firstJob, options)
self.fail()
except JobStoreCreationException:
pass
options.restart = False
# Now actually run the workflow
try:
Job.Runner.startToil(firstJob, options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
# Check we get an exception if we try to run without restart on an existing store
try:
Job.Runner.startToil(firstJob, options)
self.fail()
except JobStoreCreationException:
pass
options.restart = True
# This loop tests the restart behavior
totalTrys = 1
while i != 0:
options.useExistingOptions = random.random() > 0.5
try:
Job.Runner.startToil(firstJob, options)
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
if totalTrys > 16: #p(fail after this many restarts) = 0.5**32
self.fail(
) #Exceeded a reasonable number of restarts
totalTrys += 1
# Now check that if you try to restart from here it will raise an exception
# indicating that there are no jobs remaining in the workflow.
try:
Job.Runner.startToil(firstJob, options)
self.fail()
except JobException:
pass
# Now check the file is properly sorted..
with open(tempSortFile, 'r') as fileHandle:
l2 = fileHandle.readlines()
self.assertEquals(l, l2)
finally:
subprocess.check_call(
[resolveEntryPoint('toil'), 'clean', jobStore])
def toilMain(self):
return resolveEntryPoint('toil')
def _toilSort(self,
jobStoreLocator,
batchSystem,
lines=defaultLines,
N=defaultN,
testNo=1,
lineLen=defaultLineLen,
retryCount=2,
badWorker=0.5,
downCheckpoints=False,
disableCaching=False):
"""
Generate a file consisting of the given number of random lines, each line of the given
length. Sort the file with Toil by splitting the file recursively until each part is less
than the given number of bytes, sorting each part and merging them back together. Then
verify the result.
:param jobStoreLocator: The location of the job store.
:param batchSystem: the name of the batch system
:param lines: the number of random lines to generate
:param N: the size in bytes of each split
:param testNo: the number of repeats of this test
:param lineLen: the length of each random line in the file
"""
for test in range(testNo):
try:
# Specify options
options = Job.Runner.getDefaultOptions(jobStoreLocator)
options.logLevel = getLogLevelString()
options.retryCount = retryCount
options.batchSystem = batchSystem
options.clean = "never"
options.badWorker = badWorker
options.badWorkerFailInterval = 0.05
options.disableCaching = disableCaching
# This is required because mesosMasterAddress now defaults to the IP of the machine
# that is starting the workflow while the mesos *tests* run locally.
if batchSystem == 'mesos':
options.mesosMasterAddress = 'localhost:5050'
options.downCheckpoints = downCheckpoints
options.N = N
options.outputFile = self.outputFile
options.fileToSort = self.inputFile
options.overwriteOutput = True
options.realTimeLogging = True
# Make the file to sort
makeFileToSort(options.fileToSort,
lines=lines,
lineLen=lineLen)
# First make our own sorted version
with open(options.fileToSort, 'r') as fileHandle:
l = fileHandle.readlines()
l.sort()
# Check we get an exception if we try to restart a workflow that doesn't exist
options.restart = True
with self.assertRaises(NoSuchJobStoreException):
with runMain(options):
# Now check the file is properly sorted..
with open(options.outputFile, 'r') as fileHandle:
l2 = fileHandle.readlines()
self.assertEquals(l, l2)
options.restart = False
# Now actually run the workflow
try:
with runMain(options):
pass
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
# Check we get an exception if we try to run without restart on an existing store
with self.assertRaises(JobStoreExistsException):
with runMain(options):
pass
options.restart = True
# This loop tests the restart behavior
totalTrys = 1
while i != 0:
options.useExistingOptions = random.random() > 0.5
try:
with runMain(options):
pass
i = 0
except FailedJobsException as e:
i = e.numberOfFailedJobs
if totalTrys > 32: # p(fail after this many restarts) = 0.5**32
self.fail(
'Exceeded a reasonable number of restarts')
totalTrys += 1
finally:
subprocess.check_call(
[resolveEntryPoint('toil'), 'clean', jobStoreLocator])
# final test to make sure the jobStore was actually deleted
self.assertRaises(NoSuchJobStoreException, Toil.resumeJobStore,
jobStoreLocator)
