def recordInCouch(self, jobs, newstate, oldstate, updatesummary = False):
"""
_recordInCouch_
Record relevant job information in couch. If the job does not yet exist
in couch it will be saved as a seperate document. If the job has a FWJR
attached that will be saved as a seperate document.
"""
if not self._connectDatabases():
logging.error('Databases not connected properly')
return
timestamp = int(time.time())
couchRecordsToUpdate = []
for job in jobs:
couchDocID = job.get("couch_record", None)
if newstate == "new":
oldstate = "none"
if job.get("site_cms_name", None):
if newstate == "executing":
jobLocation = job["site_cms_name"]
else:
jobLocation = "Agent"
else:
jobLocation = "Agent"
if couchDocID == None:
jobDocument = {}
jobDocument["_id"] = str(job["id"])
job["couch_record"] = jobDocument["_id"]
jobDocument["jobid"] = job["id"]
jobDocument["workflow"] = job["workflow"]
jobDocument["task"] = job["task"]
jobDocument["owner"] = job["owner"]
jobDocument["inputfiles"] = []
for inputFile in job["input_files"]:
docInputFile = inputFile.json()
docInputFile["parents"] = []
for parent in inputFile["parents"]:
docInputFile["parents"].append({"lfn": parent["lfn"]})
jobDocument["inputfiles"].append(docInputFile)
jobDocument["states"] = {"0": {"oldstate": oldstate,
"newstate": newstate,
"location": jobLocation,
"timestamp": timestamp}}
jobDocument["jobgroup"] = job["jobgroup"]
jobDocument["mask"] = {"FirstEvent": job["mask"]["FirstEvent"],
"LastEvent": job["mask"]["LastEvent"],
"FirstLumi": job["mask"]["FirstLumi"],
"LastLumi": job["mask"]["LastLumi"],
"FirstRun": job["mask"]["FirstRun"],
"LastRun": job["mask"]["LastRun"]}
if job['mask']['runAndLumis'] != {}:
# Then we have to save the mask runAndLumis
jobDocument['mask']['runAndLumis'] = {}
for key in job['mask']['runAndLumis'].keys():
jobDocument['mask']['runAndLumis'][str(key)] = job['mask']['runAndLumis'][key]
jobDocument["name"] = job["name"]
jobDocument["type"] = "job"
jobDocument["user"] = job.get("user", None)
jobDocument["group"] = job.get("group", None)
jobDocument["taskType"] = job.get("taskType", "Unknown")
jobDocument["jobType"] = job.get("jobType", "Unknown")
couchRecordsToUpdate.append({"jobid": job["id"],
"couchid": jobDocument["_id"]})
self.jobsdatabase.queue(jobDocument, callback = discardConflictingDocument)
else:
# We send a PUT request to the stateTransition update handler.
# Couch expects the parameters to be passed as arguments to in
# the URI while the Requests class will only encode arguments
# this way for GET requests. Changing the Requests class to
# encode PUT arguments as couch expects broke a bunch of code so
# we'll just do our own encoding here.
updateUri = "/" + self.jobsdatabase.name + "/_design/JobDump/_update/stateTransition/" + couchDocID
updateUri += "?oldstate=%s&newstate=%s&location=%s×tamp=%s" % (oldstate,
newstate,
jobLocation,
timestamp)
self.jobsdatabase.makeRequest(uri = updateUri, type = "PUT", decode = False)
# updating the status of the summary doc only when it is explicitely requested
# doc is already in couch
if updatesummary:
jobSummaryId = job["name"]
updateUri = "/" + self.jsumdatabase.name + "/_design/WMStats/_update/jobSummaryState/" + jobSummaryId
# map retrydone state to jobfailed state for monitoring
if newstate == "retrydone":
monitorState = "jobfailed"
else:
monitorState = newstate
updateUri += "?newstate=%s×tamp=%s" % (monitorState, timestamp)
self.jsumdatabase.makeRequest(uri = updateUri, type = "PUT", decode = False)
logging.debug("Updated job summary status for job %s" % jobSummaryId)
updateUri = "/" + self.jsumdatabase.name + "/_design/WMStats/_update/jobStateTransition/" + jobSummaryId
updateUri += "?oldstate=%s&newstate=%s&location=%s×tamp=%s" % (oldstate,
monitorState,
job["location"],
timestamp)
self.jsumdatabase.makeRequest(uri = updateUri, type = "PUT", decode = False)
logging.debug("Updated job summary state history for job %s" % jobSummaryId)
if job.get("fwjr", None):
# If there are too many input files, strip them out
# of the FWJR, as they should already
# be in the database
# This is not critical
try:
if len(job['fwjr'].getAllInputFiles()) > self.maxUploadedInputFiles:
job['fwjr'].stripInputFiles()
except:
logging.error("Error while trying to strip input files from FWJR. Ignoring.")
pass
# complete fwjr document
job["fwjr"].setTaskName(job["task"])
fwjrDocument = {"_id": "%s-%s" % (job["id"], job["retry_count"]),
"jobid": job["id"],
"retrycount": job["retry_count"],
"fwjr": job["fwjr"].__to_json__(None),
"type": "fwjr"}
self.fwjrdatabase.queue(fwjrDocument, timestamp = True, callback = discardConflictingDocument)
updateSummaryDB(self.statsumdatabase, job)
#TODO: can add config switch to swich on and off
# if self.config.JobSateMachine.propagateSuccessJobs or (job["retry_count"] > 0) or (newstate != 'success'):
if (job["retry_count"] > 0) or (newstate != 'success'):
jobSummaryId = job["name"]
# building a summary of fwjr
logging.debug("Pushing job summary for job %s" % jobSummaryId)
errmsgs = {}
inputs = []
if "steps" in fwjrDocument["fwjr"]:
for step in fwjrDocument["fwjr"]["steps"]:
if "errors" in fwjrDocument["fwjr"]["steps"][step]:
errmsgs[step] = [error for error in fwjrDocument["fwjr"]["steps"][step]["errors"]]
if "input" in fwjrDocument["fwjr"]["steps"][step] and "source" in fwjrDocument["fwjr"]["steps"][step]["input"]:
inputs.extend( [source["runs"] for source in fwjrDocument["fwjr"]['steps'][step]["input"]["source"] if "runs" in source] )
outputs = []
outputDataset = None
for singlestep in job["fwjr"].listSteps():
for singlefile in job["fwjr"].getAllFilesFromStep(step=singlestep):
if singlefile:
outputs.append({'type': 'output' if CMSSTEP.match(singlestep) else singlefile.get('module_label', None),
'lfn': singlefile.get('lfn', None),
'location': list(singlefile.get('locations', set([]))) if len(singlefile.get('locations', set([]))) > 1
else singlefile['locations'].pop(),
'checksums': singlefile.get('checksums', {}),
'size': singlefile.get('size', None) })
#it should have one output dataset for all the files
outputDataset = singlefile.get('dataset', None) if not outputDataset else outputDataset
inputFiles = []
for inputFileStruct in job["fwjr"].getAllInputFiles():
# check if inputFileSummary needs to be extended
inputFileSummary = {}
inputFileSummary["lfn"] = inputFileStruct["lfn"]
inputFileSummary["input_type"] = inputFileStruct["input_type"]
inputFiles.append(inputFileSummary)
# Don't record intermediate jobfailed status in the jobsummary
# change to jobcooloff which will be overwritten by error handler anyway
if (job["retry_count"] > 0) and (newstate == 'jobfailed'):
summarystate = 'jobcooloff'
else:
summarystate = newstate
jobSummary = {"_id": jobSummaryId,
"wmbsid": job["id"],
"type": "jobsummary",
"retrycount": job["retry_count"],
"workflow": job["workflow"],
"task": job["task"],
"jobtype": job["jobType"],
"state": summarystate,
"site": job.get("location", None),
"cms_location": job["fwjr"].getSiteName(),
"exitcode": job["fwjr"].getExitCode(),
"errors": errmsgs,
"lumis": inputs,
"outputdataset": outputDataset,
"inputfiles": inputFiles,
"acdc_url": "%s/%s" % (sanitizeURL(self.config.ACDC.couchurl)['url'], self.config.ACDC.database),
"agent_name": self.config.Agent.hostName,
"output": outputs }
if couchDocID is not None:
try:
currentJobDoc = self.jsumdatabase.document(id = jobSummaryId)
jobSummary['_rev'] = currentJobDoc['_rev']
jobSummary['state_history'] = currentJobDoc.get('state_history', [])
# record final status transition
if newstate == 'success':
finalStateDict = {'oldstate': oldstate,
'newstate': newstate,
'location': job["location"],
'timestamp': timestamp}
jobSummary['state_history'].append(finalStateDict)
noEmptyList = ["inputfiles", "lumis"]
for prop in noEmptyList:
jobSummary[prop] = jobSummary[prop] if jobSummary[prop] else currentJobDoc.get(prop, [])
except CouchNotFoundError:
pass
self.jsumdatabase.queue(jobSummary, timestamp = True)
if len(couchRecordsToUpdate) > 0:
self.setCouchDAO.execute(bulkList = couchRecordsToUpdate,
conn = self.getDBConn(),
transaction = self.existingTransaction())
self.jobsdatabase.commit(callback = discardConflictingDocument)
self.fwjrdatabase.commit(callback = discardConflictingDocument)
self.jsumdatabase.commit()
return
def recordInCouch(self, jobs, newstate, oldstate, updatesummary=False):
"""
_recordInCouch_
Record relevant job information in couch. If the job does not yet exist
in couch it will be saved as a seperate document. If the job has a FWJR
attached that will be saved as a seperate document.
"""
if not self._connectDatabases():
logging.error('Databases not connected properly')
return
timestamp = int(time.time())
couchRecordsToUpdate = []
for job in jobs:
couchDocID = job.get("couch_record", None)
if newstate == "new":
oldstate = "none"
if job.get("site_cms_name", None):
if newstate == "executing":
jobLocation = job["site_cms_name"]
else:
jobLocation = "Agent"
else:
jobLocation = "Agent"
if couchDocID is None:
jobDocument = {}
jobDocument["_id"] = str(job["id"])
job["couch_record"] = jobDocument["_id"]
jobDocument["jobid"] = job["id"]
jobDocument["workflow"] = job["workflow"]
jobDocument["task"] = job["task"]
jobDocument["owner"] = job["owner"]
jobDocument["inputfiles"] = []
for inputFile in job["input_files"]:
docInputFile = inputFile.json()
docInputFile["parents"] = []
for parent in inputFile["parents"]:
docInputFile["parents"].append({"lfn": parent["lfn"]})
jobDocument["inputfiles"].append(docInputFile)
jobDocument["states"] = {"0": {"oldstate": oldstate,
"newstate": newstate,
"location": jobLocation,
"timestamp": timestamp}}
jobDocument["jobgroup"] = job["jobgroup"]
jobDocument["mask"] = {"FirstEvent": job["mask"]["FirstEvent"],
"LastEvent": job["mask"]["LastEvent"],
"FirstLumi": job["mask"]["FirstLumi"],
"LastLumi": job["mask"]["LastLumi"],
"FirstRun": job["mask"]["FirstRun"],
"LastRun": job["mask"]["LastRun"]}
if job['mask']['runAndLumis'] != {}:
# Then we have to save the mask runAndLumis
jobDocument['mask']['runAndLumis'] = {}
for key in job['mask']['runAndLumis'].keys():
jobDocument['mask']['runAndLumis'][str(key)] = job['mask']['runAndLumis'][key]
jobDocument["name"] = job["name"]
jobDocument["type"] = "job"
jobDocument["user"] = job.get("user", None)
jobDocument["group"] = job.get("group", None)
jobDocument["taskType"] = job.get("taskType", "Unknown")
jobDocument["jobType"] = job.get("jobType", "Unknown")
couchRecordsToUpdate.append({"jobid": job["id"],
"couchid": jobDocument["_id"]})
self.jobsdatabase.queue(jobDocument, callback=discardConflictingDocument)
else:
# We send a PUT request to the stateTransition update handler.
# Couch expects the parameters to be passed as arguments to in
# the URI while the Requests class will only encode arguments
# this way for GET requests. Changing the Requests class to
# encode PUT arguments as couch expects broke a bunch of code so
# we'll just do our own encoding here.
updateUri = "/" + self.jobsdatabase.name + "/_design/JobDump/_update/stateTransition/" + couchDocID
updateUri += "?oldstate=%s&newstate=%s&location=%s×tamp=%s" % (oldstate,
newstate,
jobLocation,
timestamp)
self.jobsdatabase.makeRequest(uri=updateUri, type="PUT", decode=False)
# updating the status of the summary doc only when it is explicitely requested
# doc is already in couch
if updatesummary:
jobSummaryId = job["name"]
updateUri = "/" + self.jsumdatabase.name + "/_design/WMStatsAgent/_update/jobSummaryState/" + jobSummaryId
# map retrydone state to jobfailed state for monitoring
if newstate == "retrydone":
monitorState = "jobfailed"
else:
monitorState = newstate
updateUri += "?newstate=%s×tamp=%s" % (monitorState, timestamp)
self.jsumdatabase.makeRequest(uri=updateUri, type="PUT", decode=False)
logging.debug("Updated job summary status for job %s", jobSummaryId)
updateUri = "/" + self.jsumdatabase.name + "/_design/WMStatsAgent/_update/jobStateTransition/" + jobSummaryId
updateUri += "?oldstate=%s&newstate=%s&location=%s×tamp=%s" % (oldstate,
monitorState,
job["location"],
timestamp)
self.jsumdatabase.makeRequest(uri=updateUri, type="PUT", decode=False)
logging.debug("Updated job summary state history for job %s", jobSummaryId)
if job.get("fwjr", None):
cachedByWorkflow = self.workloadCache.setdefault(job['workflow'],
getDataFromSpecFile(
self.getWorkflowSpecDAO.execute(job['task'])[
job['task']]['spec']))
job['fwjr'].setCampaign(cachedByWorkflow.get('Campaign', ''))
job['fwjr'].setPrepID(cachedByWorkflow.get(job['task'], ''))
# If there are too many input files, strip them out
# of the FWJR, as they should already
# be in the database
# This is not critical
try:
if len(job['fwjr'].getAllInputFiles()) > self.maxUploadedInputFiles:
job['fwjr'].stripInputFiles()
except Exception as ex:
logging.error("Error while trying to strip input files from FWJR. Ignoring. : %s", str(ex))
if newstate == "retrydone":
jobState = "jobfailed"
else:
jobState = newstate
# there is race condition updating couch record location and job is completed.
# for the fast fail job, it could miss the location update
job["location"] = job["fwjr"].getSiteName() or job.get("location", "Unknown")
# complete fwjr document
job["fwjr"].setTaskName(job["task"])
jsonFWJR = job["fwjr"].__to_json__(None)
# Don't archive cleanup job report
if job["jobType"] == "Cleanup":
archStatus = "skip"
else:
archStatus = "ready"
fwjrDocument = {"_id": "%s-%s" % (job["id"], job["retry_count"]),
"jobid": job["id"],
"jobtype": job["jobType"],
"jobstate": jobState,
"retrycount": job["retry_count"],
"archivestatus": archStatus,
"fwjr": jsonFWJR,
"type": "fwjr"}
self.fwjrdatabase.queue(fwjrDocument, timestamp=True, callback=discardConflictingDocument)
updateSummaryDB(self.statsumdatabase, job)
# TODO: can add config switch to swich on and off
# if self.config.JobSateMachine.propagateSuccessJobs or (job["retry_count"] > 0) or (newstate != 'success'):
if (job["retry_count"] > 0) or (newstate != 'success'):
jobSummaryId = job["name"]
# building a summary of fwjr
logging.debug("Pushing job summary for job %s", jobSummaryId)
errmsgs = {}
inputs = []
if "steps" in fwjrDocument["fwjr"]:
for step in fwjrDocument["fwjr"]["steps"]:
if "errors" in fwjrDocument["fwjr"]["steps"][step]:
errmsgs[step] = [error for error in fwjrDocument["fwjr"]["steps"][step]["errors"]]
if "input" in fwjrDocument["fwjr"]["steps"][step] and "source" in \
fwjrDocument["fwjr"]["steps"][step]["input"]:
inputs.extend(
[source["runs"] for source in fwjrDocument["fwjr"]['steps'][step]["input"]["source"]
if "runs" in source])
outputs = []
outputDataset = None
for singlestep in job["fwjr"].listSteps():
for singlefile in job["fwjr"].getAllFilesFromStep(step=singlestep):
if singlefile:
if len(singlefile.get('locations', set())) > 1:
locations = list(singlefile.get('locations'))
elif singlefile.get('locations'):
locations = singlefile['locations'].pop()
else:
locations = set()
if CMSSTEP.match(singlestep):
outType = 'output'
else:
outType = singlefile.get('module_label', None)
outputs.append({'type': outType,
'lfn': singlefile.get('lfn', None),
'location': locations,
'checksums': singlefile.get('checksums', {}),
'size': singlefile.get('size', None)})
# it should have one output dataset for all the files
outputDataset = singlefile.get('dataset', None) if not outputDataset else outputDataset
inputFiles = []
for inputFileStruct in job["fwjr"].getAllInputFiles():
# check if inputFileSummary needs to be extended
inputFileSummary = {}
inputFileSummary["lfn"] = inputFileStruct["lfn"]
inputFileSummary["input_type"] = inputFileStruct["input_type"]
inputFiles.append(inputFileSummary)
# Don't record intermediate jobfailed status in the jobsummary
# change to jobcooloff which will be overwritten by error handler anyway
if (job["retry_count"] > 0) and (newstate == 'jobfailed'):
summarystate = 'jobcooloff'
else:
summarystate = newstate
jobSummary = {"_id": jobSummaryId,
"wmbsid": job["id"],
"type": "jobsummary",
"retrycount": job["retry_count"],
"workflow": job["workflow"],
"task": job["task"],
"jobtype": job["jobType"],
"state": summarystate,
"site": job.get("location", None),
"cms_location": job["fwjr"].getSiteName(),
"exitcode": job["fwjr"].getExitCode(),
"eos_log_url": job["fwjr"].getLogURL(),
"worker_node_info": job["fwjr"].getWorkerNodeInfo(),
"errors": errmsgs,
"lumis": inputs,
"outputdataset": outputDataset,
"inputfiles": inputFiles,
"acdc_url": "%s/%s" % (
sanitizeURL(self.config.ACDC.couchurl)['url'], self.config.ACDC.database),
"agent_name": self.config.Agent.hostName,
"output": outputs}
if couchDocID is not None:
try:
currentJobDoc = self.jsumdatabase.document(id=jobSummaryId)
jobSummary['_rev'] = currentJobDoc['_rev']
jobSummary['state_history'] = currentJobDoc.get('state_history', [])
# record final status transition
if newstate == 'success':
finalStateDict = {'oldstate': oldstate,
'newstate': newstate,
'location': job["location"],
'timestamp': timestamp}
jobSummary['state_history'].append(finalStateDict)
noEmptyList = ["inputfiles", "lumis"]
for prop in noEmptyList:
jobSummary[prop] = jobSummary[prop] if jobSummary[prop] else currentJobDoc.get(prop, [])
except CouchNotFoundError:
pass
self.jsumdatabase.queue(jobSummary, timestamp=True)
if len(couchRecordsToUpdate) > 0:
self.setCouchDAO.execute(bulkList=couchRecordsToUpdate,
conn=self.getDBConn(),
transaction=self.existingTransaction())
self.jobsdatabase.commit(callback=discardConflictingDocument)
self.fwjrdatabase.commit(callback=discardConflictingDocument)
self.jsumdatabase.commit()
return
