def testScheddQueryPoll(self):
self.launch_daemons(["SCHEDD", "COLLECTOR", "STARTD", "NEGOTIATOR"])
output_file = os.path.join(testdir, "test.out")
if os.path.exists(output_file):
os.unlink(output_file)
schedd = htcondor.Schedd()
ad = classad.parseOne(open("tests/submit.ad"))
ads = []
cluster = schedd.submit(ad, 10, False, ads)
for i in range(60):
ads_iter = schedd.xquery("ClusterId == %d" % cluster,
["JobStatus"],
name="query1")
ads_iter2 = schedd.xquery("ClusterId == %d" % cluster,
["JobStatus"],
name="query2")
ads = []
for query in htcondor.poll([ads_iter, ads_iter2]):
self.assertTrue(query.tag() in ["query1", "query2"])
ads += query.nextAdsNonBlocking()
#print ads
if len(ads) == 0:
break
if i % 2 == 0:
schedd.reschedule()
def q(self, constraint=True):
# type: Union[bool, str] -> defaultdict
"""Return list of running and idle condor jobs (CLI condor_q)."""
condor_q = defaultdict(list)
queries = [schedd.xquery(requirements="%s && %s" % (self.__q_requirement_string, constraint),
projection=["ClusterId", "ProcId", "RequestCpus", "JobStatus"])
for schedd in self.schedds]
for query in htcondor.poll(queries):
for ads in query.nextAdsNonBlocking():
key = "%s.%s" % (ads.get("ClusterId"), ads.get("ProcId"))
condor_q[key].append(int(ads.get("RequestCpus")))
condor_q[key].append(int(ads.get("JobStatus")))
return condor_q
def q(self, constraint=True):
# type: Union[bool, str] -> defaultdict
"""Return list of running and idle condor jobs (CLI condor_q)."""
condor_q = defaultdict(list)
queries = [
schedd.xquery(
requirements="%s && %s" %
(self.__q_requirement_string, constraint),
projection=["ClusterId", "ProcId", "RequestCpus", "JobStatus"])
for schedd in self.schedds
]
for query in htcondor.poll(queries):
for ads in query.nextAdsNonBlocking():
key = "%s.%s" % (ads.get("ClusterId"), ads.get("ProcId"))
condor_q[key].append(int(ads.get("RequestCpus")))
condor_q[key].append(int(ads.get("JobStatus")))
return condor_q
def testScheddQueryPoll(self):
self.launch_daemons(["SCHEDD", "COLLECTOR", "STARTD", "NEGOTIATOR"])
output_file = os.path.join(testdir, "test.out")
if os.path.exists(output_file):
os.unlink(output_file)
schedd = htcondor.Schedd()
ad = classad.parseOne(open("tests/submit.ad"))
ads = []
cluster = schedd.submit(ad, 10, False, ads)
for i in range(60):
ads_iter = schedd.xquery("ClusterId == %d" % cluster, ["JobStatus"], name="query1")
ads_iter2 = schedd.xquery("ClusterId == %d" % cluster, ["JobStatus"], name="query2")
ads = []
for query in htcondor.poll([ads_iter, ads_iter2]):
self.assertTrue(query.tag() in ["query1", "query2"])
ads += query.nextAdsNonBlocking()
#print ads
if len(ads) == 0:
break
if i % 2 == 0:
schedd.reschedule()
def extractData(self):
# Initialize the data for the main table
data = {
'running': 0,
'idle': 0,
'cores': 0,
'ram': 0,
'requested_memory': 0,
'efficiency': 0,
'qtime': [], # format changed to 'int' in the end
'remote': 0,
'filename_plot': ''
}
# Retrieve user priority information
for ad in self.negotiator.getPriorities():
name = ad.get("Name").replace(ad.get("AccountingGroup") + ".", "")
last_prio = self.priorities.setdefault(name, 500.0)
self.priorities[name] = max(ad.get("Priority"), last_prio)
# Extract job information using htcondor python bindings
for query in htcondor.poll(self.queries):
for ads in query:
job_id = ads.get("GlobalJobId")
self.condor_jobs_information[job_id] = {
quantity: ads.get(quantity)
for quantity in self.quantities_list
}
# Fill the main table and the user statistics information
for jobid, job in self.condor_jobs_information.iteritems():
# Count remotable jobs
if job["RemoteJob"]:
data["remote"] += 1
# Determine user and set up user dependent statistics
user = job["User"]
if user not in self.user_statistics:
self.user_statistics[user] = copy.deepcopy(
self.user_statistics_dict)
# Count used RAM in MiB
if job["ResidentSetSize"] is None:
pass
else:
self.user_statistics[user][
"ram"] += job["ResidentSetSize"] / 1024.
data["ram"] += job["ResidentSetSize"] / 1024.
# Count requested RAM in MiB
self.user_statistics[user]["requested_memory"] += job[
"RequestMemory"]
data["requested_memory"] += job["RequestMemory"]
# Count used cores
self.user_statistics[user]["cores"] += job["RequestCpus"]
data["cores"] += job["RequestCpus"]
# Get information on network traffic.
if type(job["NetworkInputMb"]) == types.NoneType:
pass
else:
self.user_statistics[user]["NetworkInputMb"] += job[
"NetworkInputMb"]
if type(job["NetworkOutputMb"]) == types.NoneType:
pass
else:
self.user_statistics[user]["NetworkOutputMb"] += job[
"NetworkOutputMb"]
# Get information on input files.
self.user_statistics[user]["transfer"] += job[
"TransferInputSizeMB"]
# Summarize the status information
status = self.jobs_status_dict.get(job["JobStatus"])
self.user_statistics[user][status] += 1
if status in data:
data[status] += 1
# Calculate the time in the queue for all jobs in seconds
if status == "running":
try:
data["qtime"].append(
max(0, job["JobStartDate"] - job["QDate"]))
except Exception:
pass
# Determine the sites the user is running his jobs on
job["MachineAttrCloudSite0"] = "Undefined" if job[
"MachineAttrCloudSite0"] is None else job[
"MachineAttrCloudSite0"]
if job["MachineAttrCloudSite0"].lower(
) not in self.user_statistics[user][
"sites"] and status == "running":
self.user_statistics[user]["sites"].append(
job["MachineAttrCloudSite0"].lower())
# Calculate runtimes, cputimes and efficiencies of each job of a user
if status == "running":
try:
cputime = job["RemoteUserCpu"] + job["RemoteSysCpu"]
runtime = job["RequestCpus"] * (job["ServerTime"] -
job["JobStartDate"])
efficiency = float(cputime) / float(runtime)
# Avoiding not up to date values of JobCurrentStartDate, that result in efficiencies bigger than 1
if efficiency <= 1.:
self.user_statistics[user]["efficiencies"].append(
efficiency)
except Exception:
pass
all_efficiencies = []
for user in self.user_statistics:
user_data = {"batchsystem_user": user}
for status in self.jobs_status_dict.itervalues():
user_data[status] = self.user_statistics[user][status]
user_data["cores"], user_data["ram"] = self.user_statistics[user][
"cores"], self.determine_diskspace(
self.user_statistics[user]["ram"], given_unit="MiB")
user_data["requested_memory"] = max(
1,
self.determine_diskspace(
self.user_statistics[user]["requested_memory"],
given_unit="MiB"))
user_data["efficiency"] = round(np.mean(self.user_statistics[user]["efficiencies"]),2) \
if len(self.user_statistics[user]["efficiencies"]) > 0 else 1.0
all_efficiencies += self.user_statistics[user]["efficiencies"]
user_data["sites"] = ",\n".join(
self.user_statistics[user]["sites"])
user_data["priority"] = round(self.priorities[user], 1)
user_data["NetworkInputMb"] = round(
self.user_statistics[user]["NetworkInputMb"], 2)
user_data["NetworkOutputMb"] = round(
self.user_statistics[user]["NetworkOutputMb"], 2)
user_data["transfer"] = self.user_statistics[user][
"NetworkOutputMb"]
self.statistics_db_value_list.append(user_data)
data["efficiency"] = round(np.mean(all_efficiencies),
2) if len(all_efficiencies) > 0 else 1.0
data["ram"] = self.determine_diskspace(data["ram"], given_unit="MiB")
data["requested_memory"] = max(
1,
self.determine_diskspace(data["requested_memory"],
given_unit="MiB"))
if len(data["qtime"]) != 0:
data["qtime"] = str(timedelta(seconds=int(np.mean(data["qtime"]))))
else:
data["qtime"] = str(0)
# Plot creation for user statistics
data["filename_plot"] = self.plot()
# Overall status calculation
efficiency_status = 0.0
if data["efficiency"] >= float(self.config["efficiency_warning"]):
efficiency_status = 1.0
elif data["efficiency"] >= float(self.config["efficiency_critical"]):
efficiency_status = 0.5
ram_status = 0.5 if float(data["ram"]) / float(data["cores"]) > float(
self.config["ram_requested_memory_ratio"]) else 1.0
queue_time_status = 0.0 if "day" in data["qtime"] else 1.0
data["status"] = min(efficiency_status, ram_status, queue_time_status)
return data
def extractData(self):
# Initialize the data for the main table
data = {
'running': 0,
'idle': 0,
'cores': 0,
'ram': 0,
'requested_memory' : 0,
'efficiency': 0,
'qtime': [], # format changed to 'int' in the end
'remote': 0,
'filename_plot' : ''
}
# Retrieve user priority information
for ad in self.negotiator.getPriorities():
name = ad.get("Name").replace(ad.get("AccountingGroup")+".","")
last_prio = self.priorities.setdefault(name,500.0)
self.priorities[name]= max(ad.get("Priority"),last_prio)
# Extract job information using htcondor python bindings
for query in htcondor.poll(self.queries):
for ads in query:
job_id = ads.get("GlobalJobId")
self.condor_jobs_information[job_id] = {quantity : ads.get(quantity) for quantity in self.quantities_list}
# Fill the main table and the user statistics information
for jobid, job in self.condor_jobs_information.iteritems():
# Count remotable jobs
if job["RemoteJob"]:
data["remote"] += 1
# Determine user and set up user dependent statistics
user = job["User"]
if user not in self.user_statistics:
self.user_statistics[user] = copy.deepcopy(self.user_statistics_dict)
# Count used RAM in MiB
if job["ResidentSetSize"] is None:
pass
else:
self.user_statistics[user]["ram"] += job["ResidentSetSize"]/1024.
data["ram"] += job["ResidentSetSize"]/1024.
# Count requested RAM in MiB
self.user_statistics[user]["requested_memory"] += job["RequestMemory"]
data["requested_memory"] += job["RequestMemory"]
# Count used cores
self.user_statistics[user]["cores"] += job["RequestCpus"]
data["cores"] += job["RequestCpus"]
# Get information on network traffic.
if type(job["NetworkInputMb"]) == types.NoneType:
pass
else:
self.user_statistics[user]["NetworkInputMb"] += job["NetworkInputMb"]
if type(job["NetworkOutputMb"]) == types.NoneType:
pass
else:
self.user_statistics[user]["NetworkOutputMb"] += job["NetworkOutputMb"]
# Get information on input files.
self.user_statistics[user]["transfer"] += job["TransferInputSizeMB"]
# Summarize the status information
status = self.jobs_status_dict.get(job["JobStatus"])
self.user_statistics[user][status] +=1
if status in data:
data[status] += 1
# Calculate the time in the queue for all jobs in seconds
if status == "running":
try:
data["qtime"].append(max(0, job["JobStartDate"] - job["QDate"]))
except Exception:
pass
# Determine the sites the user is running his jobs on
job["MachineAttrCloudSite0"] = "Undefined" if job["MachineAttrCloudSite0"] is None else job["MachineAttrCloudSite0"]
if job["MachineAttrCloudSite0"].lower() not in self.user_statistics[user]["sites"] and status == "running":
self.user_statistics[user]["sites"].append(job["MachineAttrCloudSite0"].lower())
# Calculate runtimes, cputimes and efficiencies of each job of a user
if status == "running":
try:
cputime = job["RemoteUserCpu"] + job["RemoteSysCpu"]
runtime = job["RequestCpus"] * (job["ServerTime"] - job["JobStartDate"])
efficiency = float(cputime)/float(runtime)
# Avoiding not up to date values of JobCurrentStartDate, that result in efficiencies bigger than 1
if efficiency <= 1.:
self.user_statistics[user]["efficiencies"].append(efficiency)
except Exception:
pass
all_efficiencies = []
for user in self.user_statistics:
user_data = {"batchsystem_user": user}
for status in self.jobs_status_dict.itervalues():
user_data[status] = self.user_statistics[user][status]
user_data["cores"],user_data["ram"] = self.user_statistics[user]["cores"], self.determine_diskspace(self.user_statistics[user]["ram"], given_unit="MiB")
user_data["requested_memory"] = max(1,self.determine_diskspace(self.user_statistics[user]["requested_memory"], given_unit = "MiB"))
user_data["efficiency"] = round(np.mean(self.user_statistics[user]["efficiencies"]),2) \
if len(self.user_statistics[user]["efficiencies"]) > 0 else 1.0
all_efficiencies += self.user_statistics[user]["efficiencies"]
user_data["sites"] = ",\n".join(self.user_statistics[user]["sites"])
user_data["priority"] = round(self.priorities[user],1)
user_data["NetworkInputMb"] = round(self.user_statistics[user]["NetworkInputMb"],2)
user_data["NetworkOutputMb"] = round(self.user_statistics[user]["NetworkOutputMb"],2)
user_data["transfer"] = self.user_statistics[user]["NetworkOutputMb"]
self.statistics_db_value_list.append(user_data)
data["efficiency"] = round(np.mean(all_efficiencies),2) if len(all_efficiencies)> 0 else 1.0
data["ram"] = self.determine_diskspace(data["ram"], given_unit="MiB")
data["requested_memory"] = max(1,self.determine_diskspace(data["requested_memory"], given_unit = "MiB"))
if len(data["qtime"]) != 0:
data["qtime"] = str(timedelta(seconds=int(np.mean(data["qtime"]))))
else:
data["qtime"] = str(0)
# Plot creation for user statistics
data["filename_plot"] = self.plot()
# Overall status calculation
efficiency_status = 0.0
if data["efficiency"] >= float(self.config["efficiency_warning"]):
efficiency_status = 1.0
elif data["efficiency"] >= float(self.config["efficiency_critical"]):
efficiency_status = 0.5
ram_status = 0.5 if float(data["ram"])/float(data["cores"]) > float(self.config["ram_requested_memory_ratio"]) else 1.0
queue_time_status = 0.0 if "day" in data["qtime"] else 1.0
data["status"] = min(efficiency_status, ram_status, queue_time_status)
return data
jobs = []
for schedd_ad in htcondor.Collector().locateAll(htcondor.DaemonTypes.Schedd):
schedd = htcondor.Schedd(schedd_ad)
jobs += schedd.xquery()
print(len(jobs))
# yield the matching jobs;
queries = []
coll_query = htcondor.Collector().locate(htcondor.AdTypes.Schedd)
for schedd_ad in coll_query:
schedd_obj = htcondor.Schedd(schedd_ad)
queries.append(schedd_obj.xquery())
# 返回可用结果
jobs_count = {}
for query in htcondor.poll(queries):
schedd_name = query.tag()
job_counts.setdefault(schedd_name, 0)
count = len(query.nextAdsNonBlocking())
job_counts[schedd_name] += count
print "Got %d results from %s." % (count, schedd_name)
print(job_counts)
# 历史查询
# suspend() 挂起
for ad in schedd.history('true', ['ProcId', 'ClusterId', 'JobStatus', 'WallDuration'], 2):
print(ad)
