def fixupTask(task):
""" Fixup some values obtained by the query. """
result = task._asdict()
#fixup timestamps
for field in ['tm_start_time', 'tm_start_injection', 'tm_end_injection']:
current = result[field]
result[field] = str(getEpochFromDBTime(current)) if current else ''
#fixup CLOBS values by calling read (only for Oracle)
for field in [
'tm_task_failure', 'tm_split_args', 'tm_outfiles',
'tm_tfile_outfiles', 'tm_edm_outfiles', 'panda_resubmitted_jobs',
'tm_arguments', 'tm_scriptargs', 'tm_user_files', 'tm_arguments'
]:
current = result[field]
fixedCurr = current if (
current is None or isinstance(current, str)) else current.read()
result[field] = fixedCurr
#liter_evaluate values
for field in [
'tm_site_whitelist', 'tm_site_blacklist', 'tm_split_args',
'tm_outfiles', 'tm_tfile_outfiles', 'tm_edm_outfiles',
'panda_resubmitted_jobs', 'tm_user_infiles', 'tm_arguments',
'tm_scriptargs', 'tm_user_files'
]:
current = result[field]
result[field] = literal_eval(current)
#convert tm_arguments to the desired values
extraargs = result['tm_arguments']
result['resubmit_publication'] = extraargs[
'resubmit_publication'] if 'resubmit_publication' in extraargs else None
result['resubmit_jobids'] = extraargs[
'resubmit_jobids'] if 'resubmit_jobids' in extraargs else None
if result[
'resubmit_jobids'] is None and 'resubmitList' in extraargs: ## For backward compatibility only.
result['resubmit_jobids'] = extraargs['resubmitList']
result['resubmit_site_whitelist'] = extraargs[
'site_whitelist'] if 'site_whitelist' in extraargs else None
if result[
'resubmit_site_whitelist'] is None and 'siteWhiteList' in extraargs: ## For backward compatibility only.
result['resubmit_site_whitelist'] = extraargs['siteWhiteList']
result['resubmit_site_blacklist'] = extraargs[
'site_blacklist'] if 'site_blacklist' in extraargs else None
if result[
'resubmit_site_blacklist'] is None and 'siteBlackList' in extraargs: ## For backward compatibility only.
result['resubmit_site_blacklist'] = extraargs['siteBlackList']
result['resubmit_maxjobruntime'] = extraargs[
'maxjobruntime'] if 'maxjobruntime' in extraargs else None
result['resubmit_maxmemory'] = extraargs[
'maxmemory'] if 'maxmemory' in extraargs else None
result['resubmit_numcores'] = extraargs[
'numcores'] if 'numcores' in extraargs else None
result['resubmit_priority'] = extraargs[
'priority'] if 'priority' in extraargs else None
return result
def makeStatusReturnDict(self,
crabDBInfo,
combinedStatus,
dagStatus='',
statusFailureMsg='',
shortResult={},
statusCacheInfo={},
pubStatus={},
proxiedWebDir=''):
""" Create a dictionary which is mostly identical to the dictionary
that was being returned by the old status (plus a few other keys
needed by the other client commands). This is to ensure backward
compatibility after the status2 transition for users relying on
this dictionary in their scripts.
"""
statusDict = {}
statusDict['status'] = combinedStatus
statusDict['dbStatus'] = getColumn(crabDBInfo, 'tm_task_status')
statusDict['dagStatus'] = dagStatus
statusDict['username'] = getColumn(crabDBInfo, 'tm_username')
statusDict['taskFailureMsg'] = getColumn(crabDBInfo, 'tm_task_failure')
statusDict['taskWarningMsg'] = getColumn(crabDBInfo,
'tm_task_warnings')
statusDict['outdatasets'] = getColumn(crabDBInfo, 'tm_output_dataset')
statusDict['schedd'] = getColumn(crabDBInfo, 'tm_schedd')
statusDict['collector'] = getColumn(crabDBInfo, 'tm_collector')
statusDict['ASOURL'] = getColumn(crabDBInfo, 'tm_asourl')
statusDict['command'] = getColumn(crabDBInfo, 'tm_task_command')
statusDict['publicationEnabled'] = True if getColumn(
crabDBInfo, 'tm_publication') == 'T' else False
statusDict['userWebDirURL'] = getColumn(crabDBInfo, 'tm_user_webdir')
statusDict['inputDataset'] = getColumn(crabDBInfo, 'tm_input_dataset')
dbStartTime = getColumn(crabDBInfo, 'tm_start_time')
statusDict['submissionTime'] = getEpochFromDBTime(
datetime.strptime(dbStartTime, '%Y-%m-%d %H:%M:%S.%f'))
statusDict['statusFailureMsg'] = statusFailureMsg
statusDict['proxiedWebDir'] = proxiedWebDir
statusDict['jobsPerStatus'] = shortResult.get('jobsPerStatus', {})
statusDict['jobList'] = shortResult.get('jobList', {})
statusDict['publication'] = pubStatus.get('status', {})
statusDict['publicationFailures'] = pubStatus.get(
'failure_reasons', {})
statusDict['jobs'] = statusCacheInfo
return statusDict
def fixupTask(task):
""" Fixup some values obtained by the query. """
result = task._asdict()
#fixup timestamps
for field in ['tm_start_time', 'tm_start_injection', 'tm_end_injection']:
current = result[field]
result[field] = str(getEpochFromDBTime(current)) if current else ''
#fixup CLOBS values by calling read (only for Oracle)
for field in ['tm_task_failure', 'tm_split_args', 'tm_outfiles', 'tm_tfile_outfiles', 'tm_edm_outfiles',
'panda_resubmitted_jobs', 'tm_arguments', 'tm_scriptargs', 'tm_user_files', 'tm_arguments']:
current = result[field]
fixedCurr = current if (current is None or isinstance(current, str)) else current.read()
result[field] = fixedCurr
#liter_evaluate values
for field in ['tm_site_whitelist', 'tm_site_blacklist', 'tm_split_args', 'tm_outfiles', 'tm_tfile_outfiles',
'tm_edm_outfiles', 'panda_resubmitted_jobs', 'tm_user_infiles', 'tm_arguments', 'tm_scriptargs',
'tm_user_files']:
current = result[field]
result[field] = literal_eval(current)
#convert tm_arguments to the desired values
extraargs = result['tm_arguments']
result['resubmit_publication'] = extraargs['resubmit_publication'] if 'resubmit_publication' in extraargs else None
result['resubmit_jobids'] = extraargs['resubmit_jobids'] if 'resubmit_jobids' in extraargs else None
if result['resubmit_jobids'] is None and 'resubmitList' in extraargs: ## For backward compatibility only.
result['resubmit_jobids'] = extraargs['resubmitList']
result['resubmit_site_whitelist'] = extraargs['site_whitelist'] if 'site_whitelist' in extraargs else None
if result['resubmit_site_whitelist'] is None and 'siteWhiteList' in extraargs: ## For backward compatibility only.
result['resubmit_site_whitelist'] = extraargs['siteWhiteList']
result['resubmit_site_blacklist'] = extraargs['site_blacklist'] if 'site_blacklist' in extraargs else None
if result['resubmit_site_blacklist'] is None and 'siteBlackList' in extraargs: ## For backward compatibility only.
result['resubmit_site_blacklist'] = extraargs['siteBlackList']
result['resubmit_maxjobruntime'] = extraargs['maxjobruntime'] if 'maxjobruntime' in extraargs else None
result['resubmit_maxmemory'] = extraargs['maxmemory'] if 'maxmemory' in extraargs else None
result['resubmit_numcores'] = extraargs['numcores'] if 'numcores' in extraargs else None
result['resubmit_priority'] = extraargs['priority'] if 'priority' in extraargs else None
result['kill_ids'] = extraargs['killList'] if 'killList' in extraargs else []
result['kill_all'] = extraargs['killAll'] if 'killAll' in extraargs else False
return result
def resubmit2(self, workflow, publication, jobids, siteblacklist, sitewhitelist, maxjobruntime, maxmemory,
numcores, priority, userproxy):
"""Request to reprocess what the workflow hasn't finished to reprocess.
This needs to create a new workflow in the same campaign
"""
retmsg = "ok"
self.logger.info("Getting task ID tuple from DB for task %s", workflow)
row = self.api.query(None, None, self.Task.ID_sql, taskname = workflow)
try:
#just one row is picked up by the previous query
row = self.Task.ID_tuple(*next(row))
except StopIteration:
raise ExecutionError("Impossible to find task %s in the database." % workflow)
submissionTime = getEpochFromDBTime(row.start_time)
self.logger.info("Checking if resubmission is possible: we don't allow resubmission %s days before task expiration date", NUM_DAYS_FOR_RESUBMITDRAIN)
retmsg = checkTaskLifetime(submissionTime)
if retmsg != "ok":
return [{'result': retmsg}]
task_status = row.task_status
task_splitting = row.split_algo
resubmitWhat = "publications" if publication else "jobs"
self.logger.info("About to resubmit %s for workflow: %s.", resubmitWhat, workflow)
## Ignore the following options if this is a publication resubmission or if the
## task was never submitted.
if publication or task_status == 'SUBMITFAILED':
jobids = None
siteblacklist, sitewhitelist, maxjobruntime, maxmemory, numcores, priority = None, None, None, None, None, None
# We only allow resubmission of tasks that are in a final state, listed here:
allowedTaskStates = ['SUBMITTED', 'KILLED', 'KILLFAILED', 'RESUBMITFAILED', 'FAILED']
# Do not resubmit publication for tasks that were not submitted since they don't have any output.
if not publication:
allowedTaskStates += ['SUBMITFAILED'] #NB submitfailed goes to NEW, not RESUBMIT
## If the task status is not an allowed one, fail the resubmission.
if task_status not in allowedTaskStates:
msg = "You cannot resubmit %s if the task is in status %s." % (resubmitWhat, task_status)
raise ExecutionError(msg)
if task_status == 'KILLED' and task_splitting == 'Automatic':
msg = "You cannot resubmit {0} if the task is in status {1} and uses automatic splitting.".format(resubmitWhat, task_status)
raise ExecutionError(msg)
if task_status != 'SUBMITFAILED':
if publication:
## Retrieve publication information.
publicationEnabled = row.publication
asourl = row.asourl
asodb = row.asodb
username = row.username
publicationInfo = self.publicationStatusWrapper(workflow, asourl, asodb, username, publicationEnabled)
if 'status' not in publicationInfo:
msg = "Cannot resubmit publication."
msg += " Unable to retrieve the publication status."
raise ExecutionError(msg)
if 'disabled' in publicationInfo:
msg = "Cannot resubmit publication."
msg += " Publication was disabled in the CRAB configuration."
raise ExecutionError(msg)
if 'error' in publicationInfo:
msg = "Cannot resubmit publication."
msg += " Error in publication status: %s" % (publicationInfo['error'])
raise ExecutionError(msg)
if isCouchDBURL(asourl) and publicationInfo['status'].get('publication_failed', 0) == 0:
msg = "There are no failed publications to resubmit."
raise ExecutionError(msg)
## Here we can add a check on the publication status of the documents
## corresponding to the job ids in resubmitjobids and jobids. So far the
## publication resubmission will resubmit all the failed publications.
self.resubmitPublication(asourl, asodb, userproxy, workflow)
return [{'result': retmsg}]
else:
self.logger.info("Jobs to resubmit: %s", jobids)
## If these parameters are not set, give them the same values they had in the
## original task submission.
if (siteblacklist is None) or (sitewhitelist is None) or (maxjobruntime is None) or (maxmemory is None) or (numcores is None) or (priority is None):
## origValues = [orig_siteblacklist, orig_sitewhitelist, orig_maxjobruntime, orig_maxmemory, orig_numcores, orig_priority]
origValues = next(self.api.query(None, None, self.Task.GetResubmitParams_sql, taskname = workflow))
if siteblacklist is None:
siteblacklist = literal_eval(origValues[0])
if sitewhitelist is None:
sitewhitelist = literal_eval(origValues[1])
if maxjobruntime is None:
maxjobruntime = origValues[2]
if maxmemory is None:
maxmemory = origValues[3]
if numcores is None:
numcores = origValues[4]
if priority is None:
priority = origValues[5]
## These are the parameters that we want to writte down in the 'tm_arguments'
## column of the Tasks DB each time a resubmission is done.
## DagmanResubmitter will read these parameters and write them into the task ad.
arguments = {'resubmit_jobids' : jobids,
'site_blacklist' : siteblacklist,
'site_whitelist' : sitewhitelist,
'maxjobruntime' : maxjobruntime,
'maxmemory' : maxmemory,
'numcores' : numcores,
'priority' : priority,
'resubmit_publication' : publication
}
## Change the 'tm_arguments' column of the Tasks DB for this task to contain the
## above parameters.
self.api.modify(self.Task.SetArgumentsTask_sql, taskname = [workflow], arguments = [str(arguments)])
## Change the status of the task in the Tasks DB to RESUBMIT (or NEW).
if task_status == 'SUBMITFAILED':
newstate = ["NEW"]
newcommand = ["SUBMIT"]
else:
newstate = ["NEW"]
newcommand = ["RESUBMIT"]
self.api.modify(self.Task.SetStatusTask_sql, status = newstate, command = newcommand, taskname = [workflow])
return [{'result': retmsg}]
def resubmit2(self, workflow, publication, jobids, siteblacklist, sitewhitelist, maxjobruntime, maxmemory,
numcores, priority, userproxy):
"""Request to reprocess what the workflow hasn't finished to reprocess.
This needs to create a new workflow in the same campaign
"""
retmsg = "ok"
self.logger.info("Getting task ID tuple from DB for task %s" % workflow)
row = self.api.query(None, None, self.Task.ID_sql, taskname = workflow)
try:
#just one row is picked up by the previous query
row = self.Task.ID_tuple(*next(row))
except StopIteration:
raise ExecutionError("Impossible to find task %s in the database." % workflow)
submissionTime = getEpochFromDBTime(row.start_time)
self.logger.info("Checking if resubmission is possible: we don't allow resubmission %s days before task expiration date", NUM_DAYS_FOR_RESUBMITDRAIN)
retmsg = self.checkTaskLifetime(submissionTime)
if retmsg != "ok":
return [{'result': retmsg}]
task_status = row.task_status
resubmitWhat = "publications" if publication else "jobs"
self.logger.info("About to resubmit %s for workflow: %s." % (resubmitWhat, workflow))
## Ignore the following options if this is a publication resubmission or if the
## task was never submitted.
if publication or task_status == 'SUBMITFAILED':
jobids = None
siteblacklist, sitewhitelist, maxjobruntime, maxmemory, numcores, priority = None, None, None, None, None, None
# We only allow resubmission of tasks that are in a final state, listed here:
allowedTaskStates = ['SUBMITTED', 'KILLED', 'KILLFAILED', 'RESUBMITFAILED', 'FAILED']
# Do not resubmit publication for tasks that were not submitted since they don't have any output.
if not publication:
allowedTaskStates += ['SUBMITFAILED'] #NB submitfailed goes to NEW, not RESUBMIT
## If the task status is not an allowed one, fail the resubmission.
if task_status not in allowedTaskStates:
msg = "You cannot resubmit %s if the task is in status %s." % (resubmitWhat, task_status)
raise ExecutionError(msg)
if task_status != 'SUBMITFAILED':
if publication:
## Retrieve publication information.
publicationEnabled = row.publication
asourl = row.asourl
asodb = row.asodb
username = row.username
publicationInfo = self.publicationStatusWrapper(workflow, asourl, asodb, username, publicationEnabled)
if 'status' not in publicationInfo:
msg = "Cannot resubmit publication."
msg += " Unable to retrieve the publication status."
raise ExecutionError(msg)
if 'disabled' in publicationInfo:
msg = "Cannot resubmit publication."
msg += " Publication was disabled in the CRAB configuration."
raise ExecutionError(msg)
if 'error' in publicationInfo:
msg = "Cannot resubmit publication."
msg += " Error in publication status: %s" % (publicationInfo['error'])
raise ExecutionError(msg)
if publicationInfo['status'].get('publication_failed', 0) == 0:
msg = "There are no failed publications to resubmit."
raise ExecutionError(msg)
## Here we can add a check on the publication status of the documents
## corresponding to the job ids in resubmitjobids and jobids. So far the
## publication resubmission will resubmit all the failed publications.
self.resubmitPublication(asourl, asodb, userproxy, workflow)
return [{'result': retmsg}]
else:
self.logger.info("Jobs to resubmit: %s" % (jobids))
## If these parameters are not set, give them the same values they had in the
## original task submission.
if (siteblacklist is None) or (sitewhitelist is None) or (maxjobruntime is None) or (maxmemory is None) or (numcores is None) or (priority is None):
## origValues = [orig_siteblacklist, orig_sitewhitelist, orig_maxjobruntime, orig_maxmemory, orig_numcores, orig_priority]
origValues = next(self.api.query(None, None, self.Task.GetResubmitParams_sql, taskname = workflow))
if siteblacklist is None:
siteblacklist = literal_eval(origValues[0])
if sitewhitelist is None:
sitewhitelist = literal_eval(origValues[1])
if maxjobruntime is None:
maxjobruntime = origValues[2]
if maxmemory is None:
maxmemory = origValues[3]
if numcores is None:
numcores = origValues[4]
if priority is None:
priority = origValues[5]
## These are the parameters that we want to writte down in the 'tm_arguments'
## column of the Tasks DB each time a resubmission is done.
## DagmanResubmitter will read these parameters and write them into the task ad.
arguments = {'resubmit_jobids' : jobids,
'site_blacklist' : siteblacklist,
'site_whitelist' : sitewhitelist,
'maxjobruntime' : maxjobruntime,
'maxmemory' : maxmemory,
'numcores' : numcores,
'priority' : priority,
'resubmit_publication' : publication
}
## Change the 'tm_arguments' column of the Tasks DB for this task to contain the
## above parameters.
self.api.modify(self.Task.SetArgumentsTask_sql, taskname = [workflow], arguments = [str(arguments)])
## Change the status of the task in the Tasks DB to RESUBMIT (or NEW).
if task_status == 'SUBMITFAILED':
newstate = ["NEW"]
newcommand = ["SUBMIT"]
else:
newstate = ["NEW"]
newcommand = ["RESUBMIT"]
self.api.modify(self.Task.SetStatusTask_sql, status = newstate, command = newcommand, taskname = [workflow])
return [{'result': retmsg}]
def fixupTask(task):
""" Fixup some values obtained by the query. """
result = task._asdict()
# fixup timestamps
for field in ["tm_start_time", "tm_start_injection", "tm_end_injection"]:
current = result[field]
result[field] = str(getEpochFromDBTime(current)) if current else ""
# fixup CLOBS values by calling read (only for Oracle)
for field in [
"tm_task_failure",
"tm_split_args",
"tm_outfiles",
"tm_tfile_outfiles",
"tm_edm_outfiles",
"panda_resubmitted_jobs",
"tm_arguments",
"tm_scriptargs",
"tm_user_files",
"tm_arguments",
]:
current = result[field]
fixedCurr = current if (current is None or isinstance(current, str)) else current.read()
result[field] = fixedCurr
# liter_evaluate values
for field in [
"tm_site_whitelist",
"tm_site_blacklist",
"tm_split_args",
"tm_outfiles",
"tm_tfile_outfiles",
"tm_edm_outfiles",
"panda_resubmitted_jobs",
"tm_user_infiles",
"tm_arguments",
"tm_scriptargs",
"tm_user_files",
]:
current = result[field]
result[field] = literal_eval(current)
# convert tm_arguments to the desired values
extraargs = result["tm_arguments"]
result["resubmit_publication"] = extraargs["resubmit_publication"] if "resubmit_publication" in extraargs else None
result["resubmit_jobids"] = extraargs["resubmit_jobids"] if "resubmit_jobids" in extraargs else None
if result["resubmit_jobids"] is None and "resubmitList" in extraargs: ## For backward compatibility only.
result["resubmit_jobids"] = extraargs["resubmitList"]
result["resubmit_site_whitelist"] = extraargs["site_whitelist"] if "site_whitelist" in extraargs else None
if result["resubmit_site_whitelist"] is None and "siteWhiteList" in extraargs: ## For backward compatibility only.
result["resubmit_site_whitelist"] = extraargs["siteWhiteList"]
result["resubmit_site_blacklist"] = extraargs["site_blacklist"] if "site_blacklist" in extraargs else None
if result["resubmit_site_blacklist"] is None and "siteBlackList" in extraargs: ## For backward compatibility only.
result["resubmit_site_blacklist"] = extraargs["siteBlackList"]
result["resubmit_maxjobruntime"] = extraargs["maxjobruntime"] if "maxjobruntime" in extraargs else None
result["resubmit_maxmemory"] = extraargs["maxmemory"] if "maxmemory" in extraargs else None
result["resubmit_numcores"] = extraargs["numcores"] if "numcores" in extraargs else None
result["resubmit_priority"] = extraargs["priority"] if "priority" in extraargs else None
result["kill_ids"] = extraargs["killList"] if "killList" in extraargs else []
return result
def status(self, workflow, userdn, userproxy=None):
"""Retrieve the status of the workflow.
:arg str workflow: a valid workflow name
:return: a workflow status summary document"""
#Empty results
result = {
"status": '', #from the db
"command": '', #from the db
"taskFailureMsg": '', #from the db
"taskWarningMsg": [], #from the db
"submissionTime": 0, #from the db
"statusFailureMsg": '', #errors of the status itself
"jobList": [],
"schedd": '', #from the db
"splitting": '', #from the db
"taskWorker": '', #from the db
"webdirPath": '', #from the db
"username": ''
} #from the db
# First, verify the task has been submitted by the backend.
self.logger.info("Got status request for workflow %s" % workflow)
row = self.api.query(None, None, self.Task.ID_sql, taskname=workflow)
try:
#just one row is picked up by the previous query
row = self.Task.ID_tuple(*next(row))
except StopIteration:
raise ExecutionError(
"Impossible to find task %s in the database." % workflow)
result['submissionTime'] = getEpochFromDBTime(row.start_time)
if row.task_command:
result['command'] = row.task_command
## Add scheduler and collector to the result dictionary.
if row.username:
result['username'] = row.username
if row.user_webdir:
result['webdirPath'] = '/'.join(['/home/grid'] +
row.user_webdir.split('/')[-2:])
if row.schedd:
result['schedd'] = row.schedd
if row.twname:
result['taskWorker'] = row.twname
if row.split_algo:
result['splitting'] = row.split_algo
self.asoDBURL = row.asourl
# 0 - simple crab status
# 1 - crab status -long
# 2 - crab status -idle
self.logger.info("Status result for workflow %s: %s " %
(workflow, row.task_status))
## Apply taskWarning flag to output.
taskWarnings = literal_eval(row.task_warnings if isinstance(
row.task_warnings, str) else row.task_warnings.read())
result["taskWarningMsg"] = taskWarnings
## Helper function to add the task status and the failure message (both as taken
## from the Task DB) to the result dictionary.
def addStatusAndFailureFromDB(result, row):
result['status'] = row.task_status
if row.task_failure is not None:
if isinstance(row.task_failure, str):
result['taskFailureMsg'] = row.task_failure
else:
result['taskFailureMsg'] = row.task_failure.read()
## Helper function to add a failure message in retrieving the task/jobs status
## (and eventually a task status if there was none) to the result dictionary.
def addStatusAndFailure(result, status, failure=None):
if not result['status']:
result['status'] = status
if failure:
#if not result['statusFailureMsg']:
result['statusFailureMsg'] = failure
#else:
# result['statusFailureMsg'] += "\n%s" % (failure)
#get rid of this? If there is a clusterid we go ahead and get jobs info, otherwise we return result
self.logger.debug("Cluster id: %s" % row.clusterid)
if row.task_status in [
'NEW', 'HOLDING', 'UPLOADED', 'SUBMITFAILED', 'KILLFAILED',
'RESUBMITFAILED', 'FAILED'
]:
addStatusAndFailureFromDB(result, row)
if row.task_status in [
'NEW', 'UPLOADED', 'SUBMITFAILED'
] and row.task_command not in ['KILL', 'RESUBMIT']:
self.logger.debug("Detailed result for workflow %s: %s\n" %
(workflow, result))
return [result]
#even if we get rid these two should be filled
# "taskFailureMsg" : '', #from the db
# "taskWarningMsg" : [], #from the db
#here we know we have a clusterid. But what if webdir is not there? return setting a proper statusFailureMsg
#Now what to do
# get node_state/job_log from the schedd. Needs Justas patch (is it ok?)
# get error_report
# get aso_status (it is going to change once we are done whith the oracle implementation)
# combine everything
## Here we start to retrieve the jobs statuses.
jobsPerStatus = {}
taskJobCount = 0
taskStatus = {}
jobList = []
results = []
# task_codes are used if condor_q command is done to retrieve task status
task_codes = {
1: 'SUBMITTED',
2: 'SUBMITTED',
4: 'COMPLETED',
5: 'KILLED'
}
# dagman_codes are used if task status retrieved using node_state file
# 1 = STATUS_READY (Means that task was not yet started)
# 2 = STATUS_PRERUN (Means that task is doing PRE run)
# 3 = STATUS_SUBMITTED (Means that task is submitted)
# 4 = STATUS_POSTRUN (Means that task in PostRun)
# 5 = STATUS_DONE (Means that task is Done)
# 6 = STATUS_ERROR (Means that task is Failed/Killed)
dagman_codes = {
1: 'SUBMITTED',
2: 'SUBMITTED',
3: 'SUBMITTED',
4: 'SUBMITTED',
5: 'COMPLETED',
6: 'FAILED'
}
# User web directory is needed for getting files from scheduler.
if not row.user_webdir:
self.logger.error(
"webdir not found in DB. Impossible to retrieve task status")
addStatusAndFailure(result,
status='UNKNOWN',
failure='missing webdir info')
return [result]
else:
self.logger.info(
"Getting status for workflow %s using node state file.",
workflow)
try:
taskStatus = self.taskWebStatus(
{'CRAB_UserWebDir': row.user_webdir}, result)
#Check timestamp, if older then 2 minutes warn about stale info
nodeStateUpd = int(
taskStatus.get('DagStatus', {}).get("Timestamp", 0))
DAGStatus = int(
taskStatus.get('DagStatus', {}).get('DagStatus', -1))
epochTime = int(time.time())
# If DAGStatus is 5 or 6, it means it is final state and node_state file will not be updated anymore
# and there is no need to query schedd to get information about task.
# If not, we check when the last time file was updated. It should update every 30s, which is set in
# job classad:
# https://github.com/dmwm/CRABServer/blob/5caac0d379f5e4522f026eeaf3621f7eb5ced98e/src/python/TaskWorker/Actions/DagmanCreator.py#L39
if (nodeStateUpd > 0 and
(int(epochTime - nodeStateUpd) < 120)) or DAGStatus in [
5, 6
]:
self.logger.info("Node state is up to date, using it")
taskJobCount = int(
taskStatus.get('DagStatus', {}).get('NodesTotal'))
self.logger.info(taskStatus)
if row.task_status in [
'QUEUED', 'KILLED', 'KILLFAILED', 'RESUBMITFAILED',
'FAILED'
]:
result['status'] = row.task_status
else:
result['status'] = dagman_codes.get(
DAGStatus, row.task_status)
# make sure taskStatusCode is defined
if result['status'] in ['KILLED', 'KILLFAILED']:
taskStatusCode = 5
else:
taskStatusCode = 1
else:
self.logger.info(
"Node state file is too old or does not have an update time. Stale info is shown"
)
except Exception as ee:
addStatusAndFailure(result, status='UNKNOWN', failure=ee.info)
return [result]
if 'DagStatus' in taskStatus:
del taskStatus['DagStatus']
for i in range(1, taskJobCount + 1):
i = str(i)
if i not in taskStatus:
if taskStatusCode == 5:
taskStatus[i] = {'State': 'killed'}
else:
taskStatus[i] = {'State': 'unsubmitted'}
for job, info in taskStatus.items():
status = info['State']
jobsPerStatus.setdefault(status, 0)
jobsPerStatus[status] += 1
jobList.append((status, job))
result['jobList'] = jobList
#result['jobs'] = taskStatus
if len(taskStatus) == 0 and results and results['JobStatus'] == 2:
result['status'] = 'Running (jobs not submitted)'
#Always returning ASOURL also, it is required for kill, resubmit
self.logger.info("ASO: %s" % row.asourl)
result['ASOURL'] = row.asourl
## Retrieve publication information.
publicationInfo = {}
if (row.publication == 'T' and 'finished' in jobsPerStatus):
#let's default asodb to asynctransfer, for old task this is empty!
asodb = row.asodb or 'asynctransfer'
publicationInfo = self.publicationStatus(workflow, row.asourl,
asodb, row.username)
self.logger.info("Publication status for workflow %s done",
workflow)
elif (row.publication == 'F'):
publicationInfo['status'] = {'disabled': []}
else:
self.logger.info(
"No files to publish: Publish flag %s, files transferred: %s" %
(row.publication, jobsPerStatus.get('finished', 0)))
result['publication'] = publicationInfo.get('status', {})
result['publicationFailures'] = publicationInfo.get(
'failure_reasons', {})
## The output datasets are written into the Task DB by the post-job
## when uploading the output files metadata.
outdatasets = literal_eval(
row.output_dataset.read() if row.output_dataset else 'None')
result['outdatasets'] = outdatasets
return [result]
