def run(self):
setLogLevel("DEBUG")
logger.info("Adding experiments to jobTree\n")
if self.options.inputInfo:
self.addChildTarget(PreProcess(self.options))
else:
self.addChildTarget(Start(self.options))
def run(self):
setLogLevel("DEBUG")
infiles = os.listdir(self.options.indir)
for file in infiles:
if re.search('\.fastq$', file) or re.search('\.fq$', file):
sample = os.path.basename(file).split('.')[0]
self.addChildTarget( Sample(sample, self.options) )
def run(self):
setLogLevel("DEBUG")
options = self.options
system("mkdir -p %s" %(options.outdir))
experiments, samples = getExperiments(options.cactusdir)
for i, exp in enumerate(experiments):
sample = samples[i]
logger.info("Experiment %s, sample %s\n" %(exp, sample) )
self.addChildTarget( RunExperiment(options, exp, sample) )
#Map to other refs, the structure of the directories is going to be:
#outdir/
# otherRefs/
# sampleNA*/
# hg19/
# apd/
# ...
refdir = os.path.join(options.outdir, "otherRefs")
system("mkdir -p %s" %refdir)
for sample in samples:
sampleDir = os.path.join(refdir, sample)
readdir = os.path.join(self.options.readdir, sample)
system("mkdir -p %s" %sampleDir)
for ref in self.options.refs:
rdir = os.path.join(sampleDir, ref)
system("mkdir -p %s" %rdir)
self.addChildTarget( RunMapping(self.options, os.path.join(self.options.refdir, ref), rdir, readdir) )
#Done mapping, now drawPlots
self.setFollowOnTarget( Plots(options.outdir, os.path.join(options.outdir, "plots"), options.cleanup) )
def run(self):
setLogLevel("DEBUG")
logger.info("Adding experiments to jobTree\n")
if self.options.inputInfo:
self.addChildTarget( PreProcess(self.options) )
else:
self.addChildTarget( Start(self.options) )
def main():
parser = getBasicOptionParser("usage: %prog [options]", "%prog 0.1")
parser.add_option("--job", dest="jobFile",
help="Job file containing command to run",
default="None")
parser.add_option("--treePointer", dest="treePointerFile",
help="File containing pointer to the tree data",
default="None")
options, args = parseBasicOptions(parser)
logger.info("Parsed the input arguments")
job = ET.parse(options.jobFile).getroot()
setLogLevel(job.attrib["log_level"])
logger.info("Parsed the job XML")
treePointer = ET.parse(options.treePointerFile).getroot()
logger.info("Parsed the tree pointer XML")
tree = ET.parse(treePointer.attrib["file"]).getroot()
logger.info("Parsed the tree XML")
for child in tree.find("children").findall("child"):
#Make the chuld tree pointer
childTreePointerFile = makeTreePointer(child.attrib["file"], getTempFile(rootDir=job.attrib["global_temp_dir"]))
#Make the child command
unbornChild = ET.SubElement(job.find("children"), "child")
command = "jobTreeTest_CommandFirst.py --treePointer %s --job JOB_FILE" % \
(childTreePointerFile,)
unbornChild.attrib["command"] = command
if random.random() > 0.2:
unbornChild.attrib["time"] = str(random.random() * 10)
#Make the child tree pointer
ET.SubElement(treePointer.find("children"), "child", { "file":childTreePointerFile })
job.attrib["command"] = "jobTreeTest_CommandSecond.py --treePointer %s --job JOB_FILE" % \
(options.treePointerFile,)
logger.info("Made new command")
fileHandle = open(options.jobFile, 'w')
ET.ElementTree(job).write(fileHandle)
fileHandle.close()
logger.info("Updated the job file")
print >>sys.stderr, "Checking that we can report to std err" #These lines should end up in the logs
print "Checking that we can report to std out"
if random.random() > 0.9:
logger.info("Going to fail the job")
sys.exit(1)
logger.info("Going to pass the job done okay")
sys.exit(0)
def run(self):
setLogLevel("DEBUG")
globalTempDir = self.getGlobalTempDir()
for i in xrange( self.options.samnum ) :
samplingdir = os.path.join(globalTempDir, "%d" %i)
system("mkdir -p %s" %samplingdir)
self.addChildTarget( RunSampling(samplingdir, self.options) )
self.setFollowOnTarget( RunDrawNt2aaDist(globalTempDir, self.options) )
def run(self):
setLogLevel("DEBUG")
system("mkdir -p %s" %self.options.outdir)
system("chmod ug+xrw %s" %self.options.outdir)
regions = getList(self.options.regions)
logger.info("Regions: %s\n" %(" ".join(regions)))
for r in regions:
self.addChildTarget(RunRegion(r, self.options))
self.setFollowOnTarget(AggregateResults(self.options))
def run(self):
setLogLevel("DEBUG")
system("mkdir -p %s" % self.options.outdir)
system("chmod ug+xrw %s" % self.options.outdir)
regions = getList(self.options.regions)
logger.info("Regions: %s\n" % (" ".join(regions)))
for r in regions:
self.addChildTarget(RunRegion(r, self.options))
self.setFollowOnTarget(AggregateResults(self.options))
def run(self):
#--------------------------------------------------------------------------------------
#Get parameter sets. For each set, issue job to run cactus on different simulation data
#--------------------------------------------------------------------------------------
setLogLevel("DEBUG")
system("rm -rf %s*" % self.options.outputDir)
logger.info("Remove output directory if exists\n")
#Convert true.mfa of each simulation to maf format
#simTrueMafDir = os.path.join(self.options.outputDir, "sim")
simTrueMafDir = self.options.simTrueMafDir
check_dir(simTrueMafDir)
for sim in self.options.sim:
#convert mfa file of current simulation into MAF format:
sim = modify_dirname(sim)
simName = getRootDir(sim)
trueMAF = os.path.join(simTrueMafDir, "%s_true.maf" % (simName))
if not os.path.exists(trueMAF):
trueMFA = os.path.join(sim, "true.mfa")
runEvalMFAToMAF(trueMFA, trueMAF)
logger.info("Converted true.mfa of simulation %s to %s\n" %
(sim, trueMAF))
else:
logger.info("TrueMAF already exists: %s\n" % (trueMAF))
for parameterFile, parameterName in getParameters(self.options.config):
outDir = os.path.join(self.options.outputDir, parameterName)
#system("rm -rf %s" % outDir)
os.mkdir(outDir)
system("mv %s %s/" % (parameterFile, outDir))
logger.info(
"Created output directory %s for parameter set %s and moved config file to that directory\n"
% (outDir, parameterName))
paraFile = os.path.join(outDir, 'param.xml')
statsDir = os.path.join(outDir, "stats")
os.mkdir(statsDir)
logger.info("Created directory for stats files: %s\n" % (statsDir))
self.addChildTarget(
CactusTuningSimulationsWrapper(self.options, paraFile, outDir))
logger.info(
"Added CactusTuningSimulationsWrapper as child for parameter %s\n"
% (parameterName))
#Summarize results
#self.setFollowOnTarget(CactusTuningSummary(self.options))
logger.info("Added CactusTuningSummary\n")
def run(self):
setLogLevel("DEBUG")
numSim = self.options.numSim
maxSimPerRun = self.options.maxSimPerRun
sims = min( [maxSimPerRun, numSim - self.doneSims] )
outdir = os.path.join( self.options.outdir, "sims", str(self.doneSims) ) #outdir/sims/batchId
system("mkdir -p %s" %outdir)
self.addChildTarget( SimulationBatch(self.options, outdir, sims, self.doneSims) )
doneSims = self.doneSims + sims
if doneSims < numSim:
self.setFollowOnTarget( Setup(self.options, doneSims) ) #recusion call for next batch of simulations
else:
summarydir = os.path.join(self.options.outdir, "summary") #outdir/summary
statsdir = os.path.join(self.options.outdir, "stats") #outdir/stats
system("mkdir -p %s" % statsdir)
readPickle = True
writePickle = False
writeSummary = True
self.setFollowOnTarget( Summary(summarydir, self.options.samSize, self.options.numSamples, statsdir, writePickle, writeSummary, readPickle) )
def run(self):
#--------------------------------------------------------------------------------------
#Get parameter sets. For each set, issue job to run cactus on different simulation data
#--------------------------------------------------------------------------------------
setLogLevel("DEBUG")
system("rm -rf %s*" % self.options.outputDir)
logger.info("Remove output directory if exists\n")
#Convert true.mfa of each simulation to maf format
#simTrueMafDir = os.path.join(self.options.outputDir, "sim")
simTrueMafDir = self.options.simTrueMafDir
check_dir(simTrueMafDir)
for sim in self.options.sim:
#convert mfa file of current simulation into MAF format:
sim = modify_dirname(sim)
simName = getRootDir(sim)
trueMAF = os.path.join(simTrueMafDir, "%s_true.maf" %(simName))
if not os.path.exists(trueMAF):
trueMFA = os.path.join(sim, "true.mfa")
runEvalMFAToMAF(trueMFA, trueMAF)
logger.info("Converted true.mfa of simulation %s to %s\n" % (sim, trueMAF))
else:
logger.info("TrueMAF already exists: %s\n" %(trueMAF))
for parameterFile, parameterName in getParameters(self.options.config):
outDir = os.path.join(self.options.outputDir, parameterName)
#system("rm -rf %s" % outDir)
os.mkdir(outDir)
system("mv %s %s/" % (parameterFile, outDir))
logger.info("Created output directory %s for parameter set %s and moved config file to that directory\n" % (outDir, parameterName))
paraFile = os.path.join(outDir, 'param.xml')
statsDir = os.path.join(outDir, "stats")
os.mkdir(statsDir)
logger.info("Created directory for stats files: %s\n" % (statsDir))
self.addChildTarget(CactusTuningSimulationsWrapper(self.options, paraFile, outDir))
logger.info("Added CactusTuningSimulationsWrapper as child for parameter %s\n" %(parameterName))
#Summarize results
#self.setFollowOnTarget(CactusTuningSummary(self.options))
logger.info("Added CactusTuningSummary\n")
def run(self):
setLogLevel("DEBUG")
logger.info("Adding experiments to jobTree\n")
experiments = getList(self.options.experimentList)
for exp in experiments:
self.addChildTarget(RunExperiment(exp, self.options))
header.start += len( subSequence ) + lenNs
sequence = sequence[m.start() + lenNs: ]
m = re.search( pattern, sequence )
i = fn2(header, searchedSeq + sequence)
if i != None:
yield i
#=========== MAIN ====================
fH = open(sys.argv[1], 'r')
fH2 = open(sys.argv[2], 'w')
lengthOfNs = int(sys.argv[3])
lengthOfFragment = int(sys.argv[4])
if len(sys.argv) == 6:
setLogLevel(sys.argv[5])
headers = set()
for name, sequence in fastaRead(fH):
header = Header( name.split()[0], len(sequence) )
logger.info("Got a sequence of length %i with header %s for processing" % (len(sequence), name.split()[0]))
for newheader, subsequence in fn( header, sequence, lengthOfNs ):
if len( subsequence ) > 0:
logger.info("Writing out a sequence of length %i with header %s" % (len(subsequence), newheader))
assert newheader not in headers
headers.add(newheader)
fastaWrite(fH2, newheader, subsequence)
fH.close()
fH2.close()
def main():
sys.path.append(sys.argv[1])
sys.argv.remove(sys.argv[1])
#Now we can import all the stuff..
from sonLib.bioio import getBasicOptionParser
from sonLib.bioio import parseBasicOptions
from sonLib.bioio import logger
from sonLib.bioio import addLoggingFileHandler, redirectLoggerStreamHandlers
from sonLib.bioio import setLogLevel
from sonLib.bioio import getTotalCpuTime, getTotalCpuTimeAndMemoryUsage
from sonLib.bioio import getTempDirectory
from sonLib.bioio import makeSubDir
from jobTree.src.job import Job
from jobTree.src.master import getEnvironmentFileName, getConfigFileName, listChildDirs, getTempStatsFile, setupJobAfterFailure
from sonLib.bioio import system
##########################################
#Input args
##########################################
jobTreePath = sys.argv[1]
jobFile = sys.argv[2]
##########################################
#Load the environment for the job
##########################################
#First load the environment for the job.
fileHandle = open(getEnvironmentFileName(jobTreePath), 'r')
environment = cPickle.load(fileHandle)
fileHandle.close()
for i in environment:
if i not in ("TMPDIR", "TMP", "HOSTNAME", "HOSTTYPE"):
os.environ[i] = environment[i]
# sys.path is used by __import__ to find modules
if "PYTHONPATH" in environment:
for e in environment["PYTHONPATH"].split(':'):
if e != '':
sys.path.append(e)
#os.environ = environment
#os.putenv(key, value)
##########################################
#Setup the temporary directories.
##########################################
#Dir to put all the temp files in.
localSlaveTempDir = getTempDirectory()
localTempDir = makeSubDir(os.path.join(localSlaveTempDir, "localTempDir"))
##########################################
#Setup the logging
##########################################
#Setup the logging. This is mildly tricky because we don't just want to
#redirect stdout and stderr for this Python process; we want to redirect it
#for this process and all children. Consequently, we can't just replace
#sys.stdout and sys.stderr; we need to mess with the underlying OS-level
#file descriptors. See <http://stackoverflow.com/a/11632982/402891>
#When we start, standard input is file descriptor 0, standard output is
#file descriptor 1, and standard error is file descriptor 2.
#What file do we want to point FDs 1 and 2 to?
tempSlaveLogFile = os.path.join(localSlaveTempDir, "slave_log.txt")
#Save the original stdout and stderr (by opening new file descriptors to the
#same files)
origStdOut = os.dup(1)
origStdErr = os.dup(2)
#Open the file to send stdout/stderr to.
logDescriptor = os.open(tempSlaveLogFile, os.O_WRONLY | os.O_CREAT | os.O_APPEND)
#Replace standard output with a descriptor for the log file
os.dup2(logDescriptor, 1)
#Replace standard error with a descriptor for the log file
os.dup2(logDescriptor, 2)
#Since we only opened the file once, all the descriptors duped from the
#original will share offset information, and won't clobber each others'
#writes. See <http://stackoverflow.com/a/5284108/402891>. This shouldn't
#matter, since O_APPEND seeks to the end of the file before every write, but
#maybe there's something odd going on...
#Close the descriptor we used to open the file
os.close(logDescriptor)
for handler in list(logger.handlers): #Remove old handlers
logger.removeHandler(handler)
#Add the new handler. The sys.stderr stream has been redirected by swapping
#the file descriptor out from under it.
logger.addHandler(logging.StreamHandler(sys.stderr))
#Put a message at the top of the log, just to make sure it's working.
print "---JOBTREE SLAVE OUTPUT LOG---"
sys.stdout.flush()
#Log the number of open file descriptors so we can tell if we're leaking
#them.
logger.debug("Next available file descriptor: {}".format(
nextOpenDescriptor()))
##########################################
#Parse input files
##########################################
config = ET.parse(getConfigFileName(jobTreePath)).getroot()
setLogLevel(config.attrib["log_level"])
job = Job.read(jobFile)
job.messages = [] #This is the only way to stop messages logging twice, as are read only in the master
job.children = [] #Similarly, this is where old children are flushed out.
job.write() #Update status, to avoid reissuing children after running a follow on below.
if os.path.exists(job.getLogFileName()): #This cleans the old log file
os.remove(job.getLogFileName())
logger.info("Parsed arguments and set up logging")
#Try loop for slave logging
##########################################
#Setup the stats, if requested
##########################################
if config.attrib.has_key("stats"):
startTime = time.time()
startClock = getTotalCpuTime()
stats = ET.Element("slave")
else:
stats = None
##########################################
#The max time
##########################################
maxTime = float(config.attrib["job_time"])
assert maxTime > 0.0
assert maxTime < sys.maxint
##########################################
#Slave log file trapped from here on in
##########################################
slaveFailed = False
try:
##########################################
#The next job
##########################################
def globalTempDirName(job, depth):
return job.getGlobalTempDirName() + str(depth)
command, memoryAvailable, cpuAvailable, depth = job.followOnCommands[-1]
defaultMemory = int(config.attrib["default_memory"])
defaultCpu = int(config.attrib["default_cpu"])
assert len(job.children) == 0
startTime = time.time()
while True:
job.followOnCommands.pop()
##########################################
#Global temp dir
##########################################
globalTempDir = makeSubDir(globalTempDirName(job, depth))
i = 1
while os.path.isdir(globalTempDirName(job, depth+i)):
system("rm -rf %s" % globalTempDirName(job, depth+i))
i += 1
##########################################
#Old children, not yet deleted
#
#These may exist because of the lazy cleanup
#we do
##########################################
for childDir in listChildDirs(job.jobDir):
logger.debug("Cleaning up old child %s" % childDir)
system("rm -rf %s" % childDir)
##########################################
#Run the job
##########################################
if command != "": #Not a stub
if command[:11] == "scriptTree ":
##########################################
#Run the target
##########################################
loadStack(command).execute(job=job, stats=stats,
localTempDir=localTempDir, globalTempDir=globalTempDir,
memoryAvailable=memoryAvailable, cpuAvailable=cpuAvailable,
defaultMemory=defaultMemory, defaultCpu=defaultCpu, depth=depth)
else: #Is another command
system(command)
##########################################
#Cleanup/reset a successful job/checkpoint
##########################################
job.remainingRetryCount = int(config.attrib["try_count"])
system("rm -rf %s/*" % (localTempDir))
job.update(depth=depth, tryCount=job.remainingRetryCount)
##########################################
#Establish if we can run another job
##########################################
if time.time() - startTime > maxTime:
logger.info("We are breaking because the maximum time the job should run for has been exceeded")
break
#Deal with children
if len(job.children) >= 1: #We are going to have to return to the parent
logger.info("No more jobs can run in series by this slave, its got %i children" % len(job.children))
break
if len(job.followOnCommands) == 0:
logger.info("No more jobs can run by this slave as we have exhausted the follow ons")
break
#Get the next job and see if we have enough cpu and memory to run it..
command, memory, cpu, depth = job.followOnCommands[-1]
if memory > memoryAvailable:
logger.info("We need more memory for the next job, so finishing")
break
if cpu > cpuAvailable:
logger.info("We need more cpus for the next job, so finishing")
break
logger.info("Starting the next job")
##########################################
#Finish up the stats
##########################################
if stats != None:
totalCpuTime, totalMemoryUsage = getTotalCpuTimeAndMemoryUsage()
stats.attrib["time"] = str(time.time() - startTime)
stats.attrib["clock"] = str(totalCpuTime - startClock)
stats.attrib["memory"] = str(totalMemoryUsage)
tempStatsFile = getTempStatsFile(jobTreePath)
fileHandle = open(tempStatsFile + ".new", "w")
ET.ElementTree(stats).write(fileHandle)
fileHandle.close()
os.rename(tempStatsFile + ".new", tempStatsFile) #This operation is atomic
logger.info("Finished running the chain of jobs on this node, we ran for a total of %f seconds" % (time.time() - startTime))
##########################################
#Where slave goes wrong
##########################################
except: #Case that something goes wrong in slave
traceback.print_exc()
logger.critical("Exiting the slave because of a failed job on host %s", socket.gethostname())
job = Job.read(jobFile)
setupJobAfterFailure(job, config)
job.write()
slaveFailed = True
##########################################
#Cleanup
##########################################
#Close the slave logging
#Flush at the Python level
sys.stdout.flush()
sys.stderr.flush()
#Flush at the OS level
os.fsync(1)
os.fsync(2)
#Close redirected stdout and replace with the original standard output.
os.dup2(origStdOut, 1)
#Close redirected stderr and replace with the original standard error.
os.dup2(origStdOut, 2)
#sys.stdout and sys.stderr don't need to be modified at all. We don't need
#to call redirectLoggerStreamHandlers since they still log to sys.stderr
#Close our extra handles to the original standard output and standard error
#streams, so we don't leak file handles.
os.close(origStdOut)
os.close(origStdErr)
#Now our file handles are in exactly the state they were in before.
#Copy back the log file to the global dir, if needed
if slaveFailed:
truncateFile(tempSlaveLogFile)
system("mv %s %s" % (tempSlaveLogFile, job.getLogFileName()))
#Remove the temp dir
system("rm -rf %s" % localSlaveTempDir)
#This must happen after the log file is done with, else there is no place to put the log
if (not slaveFailed) and len(job.followOnCommands) == 0 and len(job.children) == 0 and len(job.messages) == 0:
##########################################
#Cleanup global files at the end of the chain
##########################################
job.delete()
def main():
parser = getBasicOptionParser("usage: %prog [options]", "%prog 0.1")
parser.add_option("--job", dest="jobFile",
help="Job file containing command to run",
default="None")
parser.add_option("--treePointer", dest="treePointer",
help="File containing pointer to the tree data",
default="None")
options, args = parseBasicOptions(parser)
logger.info("Parsed the input arguments")
print >>sys.stderr, "Checking that we can report to std err" #These lines should end up in the logs
print "Checking that we can report to std out"
job = ET.parse(options.jobFile).getroot()
setLogLevel(job.attrib["log_level"])
logger.info("Parsed the job XML")
treePointer = ET.parse(options.treePointer).getroot()
logger.info("Parsed the tree pointer XML")
tree = ET.parse(treePointer.attrib["file"]).getroot()
logger.info("Parsed the tree XML")
i = 0
children = tree.find("children").findall("child")
if len(children) > 0:
for child in children:
#Parse the child XML tree
childTree = ET.parse(child.attrib["file"]).getroot()
i += int(childTree.attrib["count"])
else:
i = 1
tree.attrib["count"] = str(i)
logger.info("Calculated the leaf count: %i" % i)
fileHandle = open(treePointer.attrib["file"], 'w')
ET.ElementTree(tree).write(fileHandle)
fileHandle.close()
logger.info("Updated the tree file: %s" % treePointer.attrib["file"])
for childPointer in treePointer.find("children").findall("child"):
if os.path.isfile(childPointer.attrib["file"]):
os.remove(childPointer.attrib["file"])
logger.info("Removed the child pointer files")
logger.info("No need to update the job file, as we didn't make anything new!")
if random.random() > 0.9:
logger.info("Going to fail the job")
sys.exit(1)
logger.info("Going to pass the job done okay")
sys.exit(0)
def main():
sys.path.append(sys.argv[1])
sys.argv.remove(sys.argv[1])
#Now we can import all the stuff..
from sonLib.bioio import getBasicOptionParser
from sonLib.bioio import parseBasicOptions
from sonLib.bioio import logger
from sonLib.bioio import addLoggingFileHandler
from sonLib.bioio import setLogLevel
from sonLib.bioio import getTotalCpuTime
from sonLib.bioio import getTempDirectory
from jobTree.src.master import writeJobs
from jobTree.src.master import readJob
from sonLib.bioio import system
##########################################
#Construct the arguments.
##########################################
parser = getBasicOptionParser("usage: %prog [options]", "%prog 0.1")
parser.add_option("--job", dest="jobFile",
help="Job file containing command to run",
default="None")
options, args = parseBasicOptions(parser)
assert len(args) == 0
##########################################
#Parse the job.
##########################################
job = readJob(options.jobFile)
##########################################
#Setup the logging
##########################################
#Setup the logging
setLogLevel(job.attrib["log_level"])
addLoggingFileHandler(job.attrib["slave_log_file"], rotatingLogging=False)
logger.info("Parsed arguments and set up logging")
##########################################
#Setup the stats, if requested
##########################################
if job.attrib.has_key("stats"):
startTime = time.time()
startClock = getTotalCpuTime()
stats = ET.Element("slave")
else:
stats = None
##########################################
#Load the environment for the job
##########################################
#First load the environment for the job.
fileHandle = open(job.attrib["environment_file"], 'r')
environment = cPickle.load(fileHandle)
fileHandle.close()
for i in environment:
if i not in ("TMPDIR", "TMP", "HOSTNAME", "HOSTTYPE"):
os.environ[i] = environment[i]
# sys.path is used by __import__ to find modules
if "PYTHONPATH" in environment:
for e in environment["PYTHONPATH"].split(':'):
if e != '':
sys.path.append(e)
#os.environ = environment
#os.putenv(key, value)
logger.info("Loaded the environment for the process")
##########################################
#Setup the temporary directories.
##########################################
#Dir to put all the temp files in.
localSlaveTempDir = getTempDirectory()
localTempDir = os.path.join(localSlaveTempDir, "localTempDir")
os.mkdir(localTempDir)
os.chmod(localTempDir, 0777)
##########################################
#Run the script.
##########################################
maxTime = float(job.attrib["job_time"])
assert maxTime > 0.0
assert maxTime < sys.maxint
jobToRun = job.find("followOns").findall("followOn")[-1]
memoryAvailable = int(jobToRun.attrib["memory"])
cpuAvailable = int(jobToRun.attrib["cpu"])
startTime = time.time()
while True:
tempLogFile = processJob(job, jobToRun, memoryAvailable, cpuAvailable, stats, environment, localSlaveTempDir, localTempDir)
if job.attrib["colour"] != "black":
logger.critical("Exiting the slave because of a failed job")
system("mv %s %s" % (tempLogFile, job.attrib["log_file"])) #Copy back the job log file, because we saw failure
break
elif job.attrib.has_key("reportAllJobLogFiles"):
logger.info("Exiting because we've been asked to report all logs, and this involves returning to the master")
#Copy across the log file
system("mv %s %s" % (tempLogFile, job.attrib["log_file"]))
break
totalRuntime = float(job.attrib["total_time"]) #This is the estimate runtime of the jobs on the followon stack
childrenNode = job.find("children")
childrenList = childrenNode.findall("child")
#childRuntime = sum([ float(child.attrib["time"]) for child in childrenList ])
if len(childrenList) >= 2: # or totalRuntime + childRuntime > maxTime: #We are going to have to return to the parent
logger.info("No more jobs can run in series by this slave, its got %i children" % len(childrenList))
break
if time.time() - startTime > maxTime:
logger.info("We are breaking because the maximum time the job should run for has been exceeded")
break
followOns = job.find("followOns")
while len(childrenList) > 0:
child = childrenList.pop()
childrenNode.remove(child)
totalRuntime += float(child.attrib["time"])
ET.SubElement(followOns, "followOn", child.attrib.copy())
#assert totalRuntime <= maxTime + 1 #The plus one second to avoid unimportant rounding errors
job.attrib["total_time"] = str(totalRuntime)
assert len(childrenNode.findall("child")) == 0
if len(followOns.findall("followOn")) == 0:
logger.info("No more jobs can run by this slave as we have exhausted the follow ons")
break
#Get the next job and see if we have enough cpu and memory to run it..
jobToRun = job.find("followOns").findall("followOn")[-1]
if int(jobToRun.attrib["memory"]) > memoryAvailable:
logger.info("We need more memory for the next job, so finishing")
break
if int(jobToRun.attrib["cpu"]) > cpuAvailable:
logger.info("We need more cpus for the next job, so finishing")
break
##Updated the job so we can start the next loop cycle
job.attrib["colour"] = "grey"
writeJobs([ job ])
logger.info("Updated the status of the job to grey and starting the next job")
#Write back the job file with the updated jobs, using the checkpoint method.
writeJobs([ job ])
logger.info("Written out an updated job file")
logger.info("Finished running the chain of jobs on this node, we ran for a total of %f seconds" % (time.time() - startTime))
##########################################
#Cleanup the temporary directory
##########################################
system("rm -rf %s" % localSlaveTempDir)
##########################################
#Finish up the stats
##########################################
if stats != None:
stats.attrib["time"] = str(time.time() - startTime)
stats.attrib["clock"] = str(getTotalCpuTime() - startClock)
fileHandle = open(job.attrib["stats"], 'w')
ET.ElementTree(stats).write(fileHandle)
fileHandle.close()
def execute(self, job):
setLogLevel(job.attrib["log_level"])
logger.info("Setup logging with level: %s" % job.attrib["log_level"])
self.tempDirAccessed = False
self.localTempDir = job.attrib["local_temp_dir"]
self.globalTempDir = job.attrib["global_temp_dir"]
maxTime = float(job.attrib["job_time"])
memory = int(job.attrib["available_memory"])
cpu = int(job.attrib["available_cpu"])
if job.attrib.has_key("stats"):
stats = ET.SubElement(job, "stack")
startTime = time.time()
startClock = getTotalCpuTime()
else:
stats = None
newChildren = [] #List to add all the children to before we package them
#off into stacks
newChildCommands = [] #Ditto for the child commands
newFollowOns = [] #Ditto for the follow-ons
baseDir = os.getcwd()
while self.hasRemaining():
if stats is not None: #Getting the runtime of the stats module
targetStartTime = time.time()
targetStartClock = getTotalCpuTime()
target = self.popTarget()
target.setStack(self)
#Debug check that we have the right amount of CPU and memory for the job in hand
targetMemory = target.getMemory()
if targetMemory != sys.maxint:
assert targetMemory <= memory
targetCpu = target.getCpu()
if targetCpu != sys.maxint:
assert targetCpu <= cpu
#Run the target, first cleanup then run.
target.run()
#Change dir back to cwd dir, if changed by target (this is a safety issue)
if os.getcwd() != baseDir:
os.chdir(baseDir)
#Cleanup after the target
if self.tempDirAccessed:
system("rm -rf %s/*" % self.localTempDir)
self.tempDirAccessed = False
#Handle the follow on
followOn = target.getFollowOn()
#if target.__class__ != CleanupGlobalTempDirTarget and followOn == None:
# followOn = CleanupGlobalTempDirTarget()
if followOn is not None: #Target to get rid of follow on when done.
if target.isGlobalTempDirSet():
followOn.setGlobalTempDir(target.getGlobalTempDir())
newFollowOns.append(followOn)
#Now add the children to the newChildren stack
newChildren += target.getChildren()
#Now add the child commands to the newChildCommands stack
newChildCommands += target.getChildCommands()
if stats is not None:
ET.SubElement(stats, "target", { "time":str(time.time() - targetStartTime),
"clock":str(getTotalCpuTime() - targetStartClock),
"class":".".join((target.__class__.__name__,)),
"e_time":str(target.getRunTime())})
for message in target.getMasterLoggingMessages():
if job.find("messages") is None:
ET.SubElement(job, "messages")
ET.SubElement(job.find("messages"), "message", { "message": message} )
#######
#Now build the new stacks and corresponding jobs
#######
#First add all the follow ons to the existing stack and make it a follow on job for job-tree
assert not self.hasRemaining()
#First sort out the follow on job
if len(newFollowOns) > 0: #If we have follow ons
followOnRuntime = sum([ followOn.getRunTime() for followOn in newFollowOns ])
if followOnRuntime > maxTime: #We create a parallel list of follow ons
followOnStack = Stack(ParallelFollowOnTarget(newFollowOns))
else:
followOnStack = Stack(newFollowOns.pop())
while len(newFollowOns) > 0:
followOnStack.addTarget(newFollowOns.pop())
job.attrib["command"] = followOnStack.makeRunnable(self.globalTempDir)
job.attrib["time"] = str(followOnStack.getRunTime())
followOnMemory = followOnStack.getMemory()
assert not job.attrib.has_key("memory")
if followOnMemory != sys.maxint:
job.attrib["memory"] = str(followOnMemory)
assert not job.attrib.has_key("cpu")
followOnCpu = followOnStack.getCpu()
if followOnCpu != sys.maxint:
job.attrib["cpu"] = str(followOnCpu)
#Now build stacks of children..
childrenTag = job.find("children")
while len(newChildren) > 0:
childStack = Stack(newChildren.pop())
while len(newChildren) > 0 and childStack.getRunTime() <= maxTime:
childStack.addTarget(newChildren.pop())
childJob = ET.SubElement(childrenTag, "child", { "command":childStack.makeRunnable(self.globalTempDir),
"time":str(childStack.getRunTime()) })
childMemory = childStack.getMemory()
assert not childJob.attrib.has_key("memory")
if childMemory != sys.maxint:
childJob.attrib["memory"] = str(childMemory)
assert not childJob.attrib.has_key("cpu")
childCpu = childStack.getCpu()
if childCpu != sys.maxint:
childJob.attrib["cpu"] = str(childCpu)
#Now build jobs for each child command
for childCommand, runTime in newChildCommands:
ET.SubElement(childrenTag, "child", { "command":str(childCommand),
"time":str(runTime) })
#Finish up the stats
if stats is not None:
stats.attrib["time"] = str(time.time() - startTime)
stats.attrib["clock"] = str(getTotalCpuTime() - startClock)
def main():
sys.path.append(sys.argv[1])
sys.argv.remove(sys.argv[1])
#Now we can import all the stuff..
from sonLib.bioio import getBasicOptionParser
from sonLib.bioio import parseBasicOptions
from sonLib.bioio import logger
from sonLib.bioio import addLoggingFileHandler, redirectLoggerStreamHandlers
from sonLib.bioio import setLogLevel
from sonLib.bioio import getTotalCpuTime, getTotalCpuTimeAndMemoryUsage
from sonLib.bioio import getTempDirectory
from sonLib.bioio import makeSubDir
from jobTree.src.job import Job
from jobTree.src.master import getEnvironmentFileName, getConfigFileName, listChildDirs, getTempStatsFile, setupJobAfterFailure
from sonLib.bioio import system
##########################################
#Input args
##########################################
jobTreePath = sys.argv[1]
jobFile = sys.argv[2]
##########################################
#Load the environment for the job
##########################################
#First load the environment for the job.
fileHandle = open(getEnvironmentFileName(jobTreePath), 'r')
environment = cPickle.load(fileHandle)
fileHandle.close()
for i in environment:
if i not in ("TMPDIR", "TMP", "HOSTNAME", "HOSTTYPE"):
os.environ[i] = environment[i]
# sys.path is used by __import__ to find modules
if "PYTHONPATH" in environment:
for e in environment["PYTHONPATH"].split(':'):
if e != '':
sys.path.append(e)
#os.environ = environment
#os.putenv(key, value)
##########################################
#Setup the temporary directories.
##########################################
#Dir to put all the temp files in.
localSlaveTempDir = getTempDirectory()
localTempDir = makeSubDir(os.path.join(localSlaveTempDir, "localTempDir"))
##########################################
#Setup the logging
##########################################
#Setup the logging
tempSlaveLogFile = os.path.join(localSlaveTempDir, "slave_log.txt")
slaveHandle = open(tempSlaveLogFile, 'w')
for handler in list(logger.handlers): #Remove old handlers
logger.removeHandler(handler)
logger.addHandler(logging.StreamHandler(slaveHandle))
origStdErr = sys.stderr
origStdOut = sys.stdout
sys.stderr = slaveHandle
sys.stdout = slaveHandle
##########################################
#Parse input files
##########################################
config = ET.parse(getConfigFileName(jobTreePath)).getroot()
setLogLevel(config.attrib["log_level"])
job = Job.read(jobFile)
job.messages = [] #This is the only way to stop messages logging twice, as are read only in the master
job.children = []
if os.path.exists(job.getLogFileName()): #This cleans the old log file
os.remove(job.getLogFileName())
logger.info("Parsed arguments and set up logging")
#Try loop for slave logging
##########################################
#Setup the stats, if requested
##########################################
if config.attrib.has_key("stats"):
startTime = time.time()
startClock = getTotalCpuTime()
stats = ET.Element("slave")
else:
stats = None
##########################################
#The max time
##########################################
maxTime = float(config.attrib["job_time"])
assert maxTime > 0.0
assert maxTime < sys.maxint
##########################################
#Slave log file trapped from here on in
##########################################
slaveFailed = False
try:
##########################################
#The next job
##########################################
def globalTempDirName(job, depth):
return job.getGlobalTempDirName() + str(depth)
command, memoryAvailable, cpuAvailable, depth = job.followOnCommands[-1]
defaultMemory = int(config.attrib["default_memory"])
defaultCpu = int(config.attrib["default_cpu"])
assert len(job.children) == 0
startTime = time.time()
while True:
job.followOnCommands.pop()
##########################################
#Global temp dir
##########################################
globalTempDir = makeSubDir(globalTempDirName(job, depth))
i = 1
while os.path.isdir(globalTempDirName(job, depth+i)):
system("rm -rf %s" % globalTempDirName(job, depth+i))
i += 1
##########################################
#Old children, not yet deleted
#
#These may exist because of the lazy cleanup
#we do
##########################################
for childDir in listChildDirs(job.jobDir):
logger.debug("Cleaning up old child %s" % childDir)
system("rm -rf %s" % childDir)
##########################################
#Run the job
##########################################
if command != "": #Not a stub
if command[:11] == "scriptTree ":
##########################################
#Run the target
##########################################
loadStack(command).execute(job=job, stats=stats,
localTempDir=localTempDir, globalTempDir=globalTempDir,
memoryAvailable=memoryAvailable, cpuAvailable=cpuAvailable,
defaultMemory=defaultMemory, defaultCpu=defaultCpu, depth=depth)
else: #Is another command
system(command)
##########################################
#Cleanup/reset a successful job/checkpoint
##########################################
job.remainingRetryCount = int(config.attrib["try_count"])
system("rm -rf %s/*" % (localTempDir))
job.update(depth=depth, tryCount=job.remainingRetryCount)
##########################################
#Establish if we can run another job
##########################################
if time.time() - startTime > maxTime:
logger.info("We are breaking because the maximum time the job should run for has been exceeded")
break
#Deal with children
if len(job.children) >= 1: #We are going to have to return to the parent
logger.info("No more jobs can run in series by this slave, its got %i children" % len(job.children))
break
if len(job.followOnCommands) == 0:
logger.info("No more jobs can run by this slave as we have exhausted the follow ons")
break
#Get the next job and see if we have enough cpu and memory to run it..
command, memory, cpu, depth = job.followOnCommands[-1]
if memory > memoryAvailable:
logger.info("We need more memory for the next job, so finishing")
break
if cpu > cpuAvailable:
logger.info("We need more cpus for the next job, so finishing")
break
logger.info("Starting the next job")
##########################################
#Finish up the stats
##########################################
if stats != None:
totalCpuTime, totalMemoryUsage = getTotalCpuTimeAndMemoryUsage()
stats.attrib["time"] = str(time.time() - startTime)
stats.attrib["clock"] = str(totalCpuTime - startClock)
stats.attrib["memory"] = str(totalMemoryUsage)
tempStatsFile = getTempStatsFile(jobTreePath)
fileHandle = open(tempStatsFile + ".new", "w")
ET.ElementTree(stats).write(fileHandle)
fileHandle.close()
os.rename(tempStatsFile + ".new", tempStatsFile) #This operation is atomic
logger.info("Finished running the chain of jobs on this node, we ran for a total of %f seconds" % (time.time() - startTime))
##########################################
#Where slave goes wrong
##########################################
except: #Case that something goes wrong in slave
traceback.print_exc(file = slaveHandle)
logger.critical("Exiting the slave because of a failed job on host %s", socket.gethostname())
job = Job.read(jobFile)
setupJobAfterFailure(job, config)
job.write()
slaveFailed = True
##########################################
#Cleanup
##########################################
#Close the slave logging
slaveHandle.flush()
sys.stderr = origStdErr
sys.stdout = origStdOut
redirectLoggerStreamHandlers(slaveHandle, sys.stderr)
slaveHandle.close()
#Copy back the log file to the global dir, if needed
if slaveFailed:
truncateFile(tempSlaveLogFile)
system("mv %s %s" % (tempSlaveLogFile, job.getLogFileName()))
#Remove the temp dir
system("rm -rf %s" % localSlaveTempDir)
#This must happen after the log file is done with, else there is no place to put the log
if (not slaveFailed) and len(job.followOnCommands) == 0 and len(job.children) == 0 and len(job.messages) == 0:
##########################################
#Cleanup global files at the end of the chain
##########################################
job.delete()
def run(self):
setLogLevel("INFO")
file2info = readSeqIndex(self.options.seqIndexFile, self.options.samples)
#logger.info("Done reading sequence.index file\n")
for sample in file2info:
self.addChildTarget( RunSample(self.options, file2info[sample]) )
