def fetch_hardware():
'''Autodetect which system we are trying to contact and perform hardware
information fetch'''
global partition_table
global node_state
global indexes
global system_type
# Generate a list of all possible nodecards in all possible partitions
system = ComponentProxy('system', defer=True)
if system_type is None:
# We are self-discovering. Can save this step if we
# already know the system type.
system_type = system.get_implementation()
if system_type in bg_types:
partition_table = dict((part['name'], part['node_card_names'])
for part in system.get_partitions([{'name': '*',
'node_card_names': '*'}]))
node_state = {}
elif system_type in cluster_types:
partition_table = {}
node_state = dict((node[0], node[1]) for node in system.get_node_status())
elif system_type in cray_types:
partition_table = {}
# Using JSON for speed and avoinding the XML-RPC marshaller.
stst = json.loads(system.get_nodes(True, None, alps_system_query_fields, True))
indexes = {}
for idx in stst:
r = stst[idx]
if r['status'] == 'busy':
indexes[idx] = r['name']
node_state = dict((k, v['status']) for k, v in stst.iteritems())
else:
raise RuntimeError('The %s system implementation is not supported by cweb')
return system_type
opts, args = opt_parser.parse_args()
except optparse.OptParseError, msg:
print msg
print helpmsg
raise SystemExit, 1
try:
system = ComponentProxy("system", defer=False)
except ComponentLookupError:
print "Failed to connect to system component"
raise SystemExit, 1
whoami = pwd.getpwuid(os.getuid())[0]
if opts.recursive:
partdata = system.get_partitions([{'tag':'partition', 'name':name, 'children_list':'*'} for name in args])
parts = args
for part in partdata:
for child in part['children']:
if child not in parts:
parts.append(child)
else:
parts = args
if opts.add:
args = ([{'tag':'partition', 'name':partname, 'size':"*", 'functional':False,
'scheduled':False, 'queue':'default', 'deps':[]} for partname in parts])
parts = system.add_partitions(args, whoami)
elif opts.delete:
args = ([{'tag':'partition', 'name':partname} for partname in parts], whoami)
try:
system = ComponentProxy("system", defer=False)
except ComponentLookupError:
print "Failed to connect to system"
raise SystemExit, 1
try:
scheduler = ComponentProxy("scheduler", defer=False)
except ComponentLookupError:
print "Failed to connect to scheduler"
raise SystemExit, 1
spec = [{'tag':'partition', 'name':'*', 'queue':'*', 'state':'*', 'size':'*',
'functional':'*', 'scheduled':'*', 'children':'*', 'backfill_time':"*", 'draining':"*"}]
try:
parts = system.get_partitions(spec)
except xmlrpclib.Fault, flt:
if flt.faultCode == NotSupportedError.fault_code:
print "incompatible with cluster support: try nodelist"
raise SystemExit, 1
else:
raise
reservations = scheduler.get_reservations([{'queue':"*", 'partitions':"*", 'active':True}])
expanded_parts = {}
for res in reservations:
for res_part in res['partitions'].split(":"):
for p in parts:
if p['name'] == res_part:
if expanded_parts.has_key(res['queue']):
except getopt.GetoptError, msg:
print msg
print helpmsg
raise SystemExit, 1
try:
system = ComponentProxy("system", defer=False)
except ComponentLookupError:
print "Failed to connect to system component"
raise SystemExit, 1
whoami = getpass.getuser()
if '-r' in sys.argv:
partdata = system.get_partitions([{
'tag': 'partition',
'name': name,
'children': '*'
} for name in args])
parts = args
for part in partdata:
for child in part['children']:
if child not in parts:
parts.append(child)
else:
parts = args
if '-a' in sys.argv:
func = system.add_partitions
args = ([{
'tag': 'partition',
'name': partname,
['dump', 'free', 'load=', 'enable', 'disable', 'activate', 'deactivate',
'queue=', 'deps=', 'xml', 'diag=', 'fail', 'unfail', 'savestate'])
except getopt.GetoptError, msg:
print msg
print helpmsg
raise SystemExit, 1
try:
system = ComponentProxy("system", defer=False)
except ComponentLookupError:
print "Failed to connect to system component"
raise SystemExit, 1
whoami = getpass.getuser()
if '-r' in sys.argv:
partdata = system.get_partitions([{'tag':'partition', 'name':name, 'children':'*'} for name in args])
parts = args
for part in partdata:
for child in part['children']:
if child not in parts:
parts.append(child)
else:
parts = args
if '-a' in sys.argv:
func = system.add_partitions
args = ([{'tag':'partition', 'name':partname, 'size':"*", 'functional':False,
'scheduled':False, 'queue':'default', 'deps':[]} for partname in parts], whoami)
elif '-d' in sys.argv:
func = system.del_partitions
args = ([{'tag':'partition', 'name':partname} for partname in parts], whoami)
def test_something(self):
logging.basicConfig()
try:
cqm = ComponentProxy("queue-manager")
except ComponentLookupError:
assert not "failed to connect to queue manager"
# add a queue
queues = cqm.add_queues([{"tag": "queue", "name": "default"}])
assert len(queues) == 1
# try adding a job to a queue that doesn't exist
try:
jobs = cqm.add_jobs([{"tag": "job", "queue": "jonx"}])
except xmlrpclib.Fault:
# trying to add a job to a queue that doesn't exist results in an xmlrpc Fault
pass
else:
assert not "Adding job to non-existent queue should raise xmlrpclib.Fault"
# get the list of available partitions and add them to the pool of managed partitions
try:
simulator = ComponentProxy("system")
except ComponentLookupError:
assert not "failed to connect to simulator"
for part_name in self.system._partitions:
partitions = simulator.add_partitions([{"tag": "partition", "name": part_name, "queue": "default"}])
assert len(partitions) == 1
partitions = simulator.set_partitions(
[{"tag": "partition", "name": part_name}], {"functional": True, "scheduled": True}
)
assert len(partitions) == 1
partitions = simulator.get_partitions([{"name": "*", "size": "*", "queue": "*"}])
assert len(partitions) > 0
# now run a real job
#
# 1. add the job to the default queue
# 2. obtain a partition for it to run on
# 3. start running it on that paritition
# 4. check that it started running
# 5. sleep for a bit, and then check that it's still running
# 6. sleep some more and then check to see if it actually finished running
nodes = partitions[0]["size"]
jobs = cqm.add_jobs(
[
{
"queue": "default",
"mode": "co",
"command": "/bin/ls",
"outputdir": os.getcwd(),
"walltime": 4,
"nodes": nodes,
"procs": nodes,
"args": [],
"user": "******",
"jobid": "*",
}
]
)
assert len(jobs) == 1
job = jobs[0]
jobid = job["jobid"]
job_location_args = [
{
"jobid": jobid,
"nodes": job["nodes"],
"queue": job["queue"],
"utility_score": 1,
"threshold": 1,
"walltime": job["walltime"],
"attrs": {},
}
]
locations = simulator.find_job_location(job_location_args, [])
assert locations.has_key(jobid)
location = locations[jobid]
cqm.run_jobs([{"jobid": jobid}], location)
r = cqm.get_jobs([{"jobid": jobid, "state": "*", "is_active": True}])
if not r:
assert not "the job didn't start"
time.sleep(20)
r = cqm.get_jobs([{"jobid": jobid, "state": "*", "is_active": True}])
if len(r) != 1:
assert not "the job has stopped running prematurely"
start_time = time.time()
while True:
r = cqm.get_jobs([{"jobid": jobid, "state": "*", "is_active": True}])
if r:
if time.time() - start_time > 240:
assert not "the job seems to have run overtime"
else:
time.sleep(5)
else:
break
# this time, we'll add a job to the queue, start the job, sleep for a bit
# and then try to kill the job before it has finished
nodes = partitions[0]["size"]
jobs = cqm.add_jobs(
[
{
"queue": "default",
"mode": "co",
"command": "/bin/ls",
"outputdir": os.getcwd(),
"walltime": 4,
"nodes": nodes,
"procs": nodes,
"args": [],
"user": "******",
"jobid": "*",
}
]
)
assert len(jobs) == 1
job = jobs[0]
jobid = job["jobid"]
job_location_args = [
{
"jobid": jobid,
"nodes": job["nodes"],
"queue": job["queue"],
"utility_score": 1,
"threshold": 1,
"walltime": job["walltime"],
"attrs": {},
}
]
locations = simulator.find_job_location(job_location_args, [])
assert locations.has_key(jobid)
location = locations[jobid]
cqm.run_jobs([{"jobid": jobid}], location)
r = cqm.get_jobs([{"jobid": jobid, "state": "*", "is_active": True}])
if not r:
assert not "the job didn't start"
time.sleep(20)
r = cqm.get_jobs([{"jobid": jobid, "is_active": True}])
if len(r) != 1:
assert not "the job has stopped running prematurely"
cqm.del_jobs([{"jobid": jobid}])
start_time = time.time()
while True:
r = cqm.get_jobs([{"jobid": jobid, "is_active": True, "state": "*"}])
if r:
if time.time() - start_time > 30:
assert not "the job didn't die when asked to"
else:
time.sleep(1)
else:
break
def test_something(self):
logging.basicConfig()
try:
cqm = ComponentProxy("queue-manager")
except ComponentLookupError:
assert not "failed to connect to queue manager"
# add a queue
queues = cqm.add_queues([{'tag': "queue", 'name': "default"}])
assert len(queues) == 1
# try adding a job to a queue that doesn't exist
try:
jobs = cqm.add_jobs([{'tag': "job", 'queue': "jonx"}])
except xmlrpclib.Fault:
# trying to add a job to a queue that doesn't exist results in an xmlrpc Fault
pass
else:
assert not "Adding job to non-existent queue should raise xmlrpclib.Fault"
# get the list of available partitions and add them to the pool of managed partitions
try:
simulator = ComponentProxy("system")
except ComponentLookupError:
assert not "failed to connect to simulator"
for part_name in self.system._partitions:
partitions = simulator.add_partitions([{
'tag': "partition",
'name': part_name,
'queue': "default"
}])
assert len(partitions) == 1
partitions = simulator.set_partitions([{
'tag': "partition",
'name': part_name
}], {
'functional': True,
'scheduled': True
})
assert len(partitions) == 1
partitions = simulator.get_partitions([{
'name': "*",
'size': "*",
'queue': "*"
}])
assert len(partitions) > 0
# now run a real job
#
# 1. add the job to the default queue
# 2. obtain a partition for it to run on
# 3. start running it on that paritition
# 4. check that it started running
# 5. sleep for a bit, and then check that it's still running
# 6. sleep some more and then check to see if it actually finished running
nodes = partitions[0]['size']
jobs = cqm.add_jobs([{
'queue': "default",
'mode': "co",
'command': "/bin/ls",
'outputdir': os.getcwd(),
'walltime': 4,
'nodes': nodes,
'procs': nodes,
'args': [],
'user': "******",
'jobid': "*"
}])
assert len(jobs) == 1
job = jobs[0]
jobid = job['jobid']
job_location_args = [{
'jobid': jobid,
'nodes': job['nodes'],
'queue': job['queue'],
'utility_score': 1,
'threshold': 1,
'walltime': job['walltime'],
'attrs': {}
}]
locations = simulator.find_job_location(job_location_args, [])
assert locations.has_key(jobid)
location = locations[jobid]
cqm.run_jobs([{'jobid': jobid}], location)
r = cqm.get_jobs([{'jobid': jobid, 'state': "*", 'is_active': True}])
if not r:
assert not "the job didn't start"
time.sleep(20)
r = cqm.get_jobs([{'jobid': jobid, 'state': "*", 'is_active': True}])
if len(r) != 1:
assert not "the job has stopped running prematurely"
start_time = time.time()
while True:
r = cqm.get_jobs([{
'jobid': jobid,
'state': "*",
'is_active': True
}])
if r:
if time.time() - start_time > 240:
assert not "the job seems to have run overtime"
else:
time.sleep(5)
else:
break
# this time, we'll add a job to the queue, start the job, sleep for a bit
# and then try to kill the job before it has finished
nodes = partitions[0]['size']
jobs = cqm.add_jobs([{
'queue': "default",
'mode': "co",
'command': "/bin/ls",
'outputdir': os.getcwd(),
'walltime': 4,
'nodes': nodes,
'procs': nodes,
'args': [],
'user': "******",
'jobid': "*"
}])
assert len(jobs) == 1
job = jobs[0]
jobid = job['jobid']
job_location_args = [{
'jobid': jobid,
'nodes': job['nodes'],
'queue': job['queue'],
'utility_score': 1,
'threshold': 1,
'walltime': job['walltime'],
'attrs': {}
}]
locations = simulator.find_job_location(job_location_args, [])
assert locations.has_key(jobid)
location = locations[jobid]
cqm.run_jobs([{'jobid': jobid}], location)
r = cqm.get_jobs([{'jobid': jobid, 'state': "*", 'is_active': True}])
if not r:
assert not "the job didn't start"
time.sleep(20)
r = cqm.get_jobs([{'jobid': jobid, 'is_active': True}])
if len(r) != 1:
assert not "the job has stopped running prematurely"
cqm.del_jobs([{'jobid': jobid}])
start_time = time.time()
while True:
r = cqm.get_jobs([{
'jobid': jobid,
'is_active': True,
'state': "*"
}])
if r:
if time.time() - start_time > 30:
assert not "the job didn't die when asked to"
else:
time.sleep(1)
else:
break
print msg
print helpmsg
raise SystemExit, 1
try:
system = ComponentProxy("system", defer=False)
except ComponentLookupError:
print "Failed to connect to system component"
raise SystemExit, 1
whoami = pwd.getpwuid(os.getuid())[0]
if opts.recursive:
partdata = system.get_partitions([{
'tag': 'partition',
'name': name,
'children_list': '*'
} for name in args])
parts = args
for part in partdata:
for child in part['children']:
if child not in parts:
parts.append(child)
else:
parts = args
if opts.add:
args = ([{
'tag': 'partition',
'name': partname,
