def getRunningCondorStatus(status_dict):
"""Return a dictionary of collectors containing running(claimed) slots
Each element is a condorStatus
:param status_dict: output of getCondorStatus
:return: dictionary of collectors containing running(claimed) slots
"""
out = {}
for collector_name in status_dict:
# Consider following slots
# 1. Static - running slots
# 2. Dynamic slots (They are always running)
# 3. p-slot with one or more dynamic slots
# We get them here so we can use them easily in appendRealRunning()
sq = condorMonitor.SubQuery(
status_dict[collector_name],
lambda el: (
((el.get('State') == 'Claimed') and
(el.get('Activity') in ('Busy', 'Retiring'))
) or
((el.get('PartitionableSlot') == True) and
(el.get('TotalSlots', 1) > 1)
)
)
)
sq.load()
out[collector_name] = sq
return out
def getRunningCondorQ(condorq_dict):
out = {}
for schedd_name in condorq_dict.keys():
sq = condorMonitor.SubQuery(condorq_dict[schedd_name], lambda el:('JobStatus' in el and (el['JobStatus'] == 2)))
sq.load()
out[schedd_name] = sq
return out
def getRunningJobsCondorStatus(status_dict):
"""Return a dictionary of collectors containing running(claimed) slots
This includes Fixed slots and Dynamic slots (no partitionable slots)
Each one is matched with a single job (gives number of running jobs)
Each element is a condorStatus
:param status_dict: output of getCondorStatus
:return: dictionary of collectors containing running(claimed) slots
"""
out = {}
for collector_name in status_dict.keys():
# This counts the running slots: fixed (static/not partitionable) or dynamic
# It may give problems when matching with RemoteHost in the jobs
# since dynamic slots report the parent partitionable slot in GLIDEIN_SiteWMS_Slot
sq = condorMonitor.SubQuery(
status_dict[collector_name],
lambda el: (
((el.get('State') == 'Claimed') and
(el.get('Activity') in ('Busy', 'Retiring'))
)
)
)
sq.load()
out[collector_name] = sq
return out
def getIdleCondorStatus(status_dict):
out = {}
for collector_name in status_dict.keys():
# Exclude partitionable slots with no free memory/cpus
# Minimum memory required by CMS is 2500 MB
#
# 1. (el.get('PartitionableSlot') != True)
# Includes static slots irrespective of the free cpu/mem
#
# 2. (el.get('TotalSlots') == 1)
# p-slots not yet partitioned
#
# 3. (el.get('Cpus', 0) > 0 and el.get('Memory', 2501) > 2500)
# p-slots that have enough idle resources.
sq = condorMonitor.SubQuery(
status_dict[collector_name],
lambda el: (
(el.get('State') == 'Unclaimed') and
(el.get('Activity') == 'Idle') and
(
(el.get('PartitionableSlot') != True) or
(el.get('TotalSlots') == 1) or
(el.get('Cpus', 0) > 0 and el.get('Memory', 2501) > 2500)
)
)
)
sq.load()
out[collector_name] = sq
return out
def getIdleProxyCondorQ(condorq_dict):
out={}
for schedd_name in condorq_dict.keys():
sq=condorMonitor.SubQuery(condorq_dict[schedd_name], lambda el:((el.get('JobStatus')==1) and ('x509userproxy' in el)))
sq.load()
out[schedd_name]=sq
return out
def getOldCondorQ(condorq_dict, min_age):
out = {}
for schedd_name in condorq_dict.keys():
sq = condorMonitor.SubQuery(condorq_dict[schedd_name], lambda el:('ServerTime' in el and 'EnteredCurrentStatus' in el and ((el['ServerTime'] - el['EnteredCurrentStatus']) >= min_age)))
sq.load()
out[schedd_name] = sq
return out
def getClientCondorStatus(status_dict, frontend_name, group_name, request_name):
client_name_old = "%s@%s.%s" % (request_name, frontend_name, group_name)
client_name_new = "%s.%s" % (frontend_name, group_name)
out = {}
for collector_name in status_dict.keys():
sq = condorMonitor.SubQuery(
status_dict[collector_name],
lambda el:('GLIDECLIENT_Name' in el and ((el['GLIDECLIENT_Name'] == client_name_old) or ((el['GLIDECLIENT_Name'] == client_name_new) and (("%s@%s@%s" % (el['GLIDEIN_Entry_Name'], el['GLIDEIN_Name'], el['GLIDEIN_Factory'])) == request_name)))))
sq.load()
out[collector_name] = sq
return out
def getClientCondorStatusCredIdOnly(status_dict, cred_id):
out = {}
for collector_name, collector_status in status_dict.iteritems():
sq = condorMonitor.SubQuery(
collector_status,
lambda el: (
'GLIDEIN_CredentialIdentifier' in el and
(el['GLIDEIN_CredentialIdentifier'] == cred_id)
)
)
sq.load()
out[collector_name] = sq
return out
def getFailedCondorStatus(status_dict):
out = {}
for collector_name in status_dict.keys():
sq = condorMonitor.SubQuery(
status_dict[collector_name],
lambda el: (
(el.get('State') == "Drained") and
(el.get('Activity') == "Retiring")
)
)
sq.load()
out[collector_name] = sq
return out
def getCondorStatusNonDynamic(status_dict):
"""
Return a dictionary of collectors containing static+partitionable slots
and exclude any dynamic slots
Each element is a condorStatus
Use the output of getCondorStatus
"""
out = {}
for collector_name in status_dict.keys():
# Exclude partitionable slots with no free memory/cpus
sq = condorMonitor.SubQuery(
status_dict[collector_name],
lambda el: (
(el.get('SlotType') != 'Dynamic')
)
)
sq.load()
out[collector_name] = sq
return out
def getRunningPSlotCondorStatus(status_dict):
"""Return a dictionary of collectors containing running(claimed) partitionable slots
Each element is a condorStatus
:param status_dict: output of getCondorStatus
:return: collectors containing running(claimed) partitionable slots
"""
out = {}
for collector_name in status_dict.keys():
# Get p-slot where there is atleast one dynamic slot
sq = condorMonitor.SubQuery(
status_dict[collector_name],
lambda el:(
(el.get('PartitionableSlot') == True) and
(el.get('TotalSlots', 1) > 1)
)
)
sq.load()
out[collector_name] = sq
return out
