def setUp(self):
self.orig_dir = os.getcwd()
os.chdir(TestCase.directory)
try:
# Force use of fake 'ssh' and 'scp'.
ssh = ('python', os.path.join(_TST_ROOT, 'ssh.py'), _DMZ_ROOT)
scp = ('python', os.path.join(_TST_ROOT, 'scp.py'), _DMZ_ROOT)
self.orig_ssh = protocol.configure_ssh(ssh)
self.orig_scp = protocol.configure_scp(scp)
# Avoid lots of polling log entries.
if logging.getLogger().getEffectiveLevel() < logging.DEBUG:
logging.getLogger().setLevel(logging.DEBUG)
# Start RJE server.
hostname = socket.gethostname()
self.proc = start_server(hostname)
# Create NAS_Allocator referring to server.
logging.debug('create allocator')
self.allocator = NAS_Allocator()
parser = ConfigParser.ConfigParser()
section = self.allocator.name
parser.add_section(section)
parser.set(section, 'dmz_host', hostname)
parser.set(section, 'server_host', hostname)
self.allocator.configure(parser)
# Add allocator to RAM.
RAM.add_allocator(self.allocator)
except Exception:
os.chdir(self.orig_dir)
raise
def setUp(self):
self.orig_dir = os.getcwd()
os.chdir(TestCase.directory)
try:
# Force use of fake 'ssh' and 'scp'.
ssh = ("python", os.path.join(_TST_ROOT, "ssh.py"), _DMZ_ROOT)
scp = ("python", os.path.join(_TST_ROOT, "scp.py"), _DMZ_ROOT)
self.orig_ssh = protocol.configure_ssh(ssh)
self.orig_scp = protocol.configure_scp(scp)
# Avoid lots of polling log entries.
if logging.getLogger().getEffectiveLevel() < logging.DEBUG:
logging.getLogger().setLevel(logging.DEBUG)
# Start RJE server.
hostname = socket.gethostname()
self.proc = start_server(hostname)
# Create NAS_Allocator referring to server.
logging.debug("create allocator")
self.allocator = NAS_Allocator()
parser = ConfigParser.ConfigParser()
section = self.allocator.name
parser.add_section(section)
parser.set(section, "dmz_host", hostname)
parser.set(section, "server_host", hostname)
self.allocator.configure(parser)
# Add allocator to RAM.
RAM.add_allocator(self.allocator)
except Exception:
os.chdir(self.orig_dir)
raise
def setUp(self):
nodes = []
for i in range(12):
nodes.append('g-0%02d'%i)
# start the fake MPI_Allocator
self.cluster=MPI_Allocator(name='test',machines=nodes)
# add it to to the RAM
RAM.add_allocator(self.cluster)
print 'Job Complete ...'
# -----------------------------
# --- Execute the component ---
# -----------------------------
super(MeshComp, self).execute()
# --- Copy files to local directories
shutil.copy2(
Path('OpenMDAO') + '/patran.out',
Path('Patran') + '\Config' + self.config)
shutil.copy2(
Path('OpenMDAO') + '/Config' + self.config + '.cub', Path('Cubit'))
if __name__ == "__main__":
enable_console()
logging.getLogger().setLevel(logging.DEBUG)
# --- Create 4 allocators (redundancy for reliability) adding each to the manager ---
allocator1 = NAS_Allocator(name='PFE20_DMZ1',
dmz_host='dmzfs1.nas.nasa.gov',
server_host='pfe20')
RAM.add_allocator(allocator1)
Remote_Mesh_Comp = MeshComp()
Remote_Mesh_Comp.run()
def test_configure(self):
logging.debug('')
logging.debug('test_configure')
# Reconfigure.
with open('resources.cfg', 'w') as out:
out.write("""
[LocalHost]
max_load: 100
""")
local = RAM.get_allocator('LocalHost')
max_load = local.max_load
try:
self.assertTrue(max_load < 100)
RAM.configure('resources.cfg')
self.assertEqual(local.max_load, 100)
local.max_load = max_load
finally:
os.remove('resources.cfg')
# Add another local.
with open('resources.cfg', 'w') as out:
out.write("""
[Local2]
classname: openmdao.main.resource.LocalAllocator
authkey: PublicKey
allow_shell: False
total_cpus: 42
max_load: 200
""")
try:
RAM.configure('resources.cfg')
local2 = RAM.get_allocator('Local2')
self.assertEqual(local2.factory._authkey, 'PublicKey')
self.assertEqual(local2.factory._allow_shell, False)
self.assertEqual(local2.total_cpus, 42)
self.assertEqual(local2.max_load, 200)
self.assertEqual(local2.host, socket.gethostname())
self.assertTrue(local2.pid > 0)
RAM.remove_allocator('Local2')
finally:
os.remove('resources.cfg')
# Bad local total_cpus.
with open('resources.cfg', 'w') as out:
out.write("""
[Local2]
classname: openmdao.main.resource.LocalAllocator
total_cpus: 0
""")
try:
assert_raises(self, "RAM.configure('resources.cfg')",
globals(), locals(), ValueError,
'Local2: total_cpus must be > 0, got 0')
finally:
os.remove('resources.cfg')
# Bad local max_load.
with open('resources.cfg', 'w') as out:
out.write("""
[Local2]
classname: openmdao.main.resource.LocalAllocator
max_load: 0
""")
try:
assert_raises(self, "RAM.configure('resources.cfg')",
globals(), locals(), ValueError,
'Local2: max_load must be > 0, got 0')
finally:
os.remove('resources.cfg')
# Bad module.
with open('resources.cfg', 'w') as out:
out.write("""
[BadModule]
classname: no-such-module.Allocator
max_load: 100
""")
try:
assert_raises(self, "RAM.configure('resources.cfg')",
globals(), locals(), RuntimeError,
"RAM configure BadModule: can't import"
" 'no-such-module'")
finally:
os.remove('resources.cfg')
# Bad class.
with open('resources.cfg', 'w') as out:
out.write("""
[BadClass]
classname: openmdao.main.resource.NoSuchAllocator
max_load: 100
""")
try:
assert_raises(self, "RAM.configure('resources.cfg')",
globals(), locals(), RuntimeError,
"RAM configure BadClass: no class"
" 'NoSuchAllocator' in openmdao.main.resource")
finally:
os.remove('resources.cfg')
# Add, insert, get, remove.
local3 = LocalAllocator('Local3')
local4 = LocalAllocator('Local4', total_cpus=4)
RAM.add_allocator(local3)
try:
allocator_names = \
[allocator.name for allocator in RAM.list_allocators()]
self.assertEqual(allocator_names, ['LocalHost', 'Local3'])
self.assertTrue(RAM.get_allocator('Local3') is local3)
self.assertTrue(RAM.get_allocator(1) is local3)
RAM.insert_allocator(0, local4)
try:
allocator_names = \
[allocator.name for allocator in RAM.list_allocators()]
self.assertEqual(allocator_names,
['Local4', 'LocalHost', 'Local3'])
finally:
RAM.remove_allocator('Local4')
finally:
RAM.remove_allocator(1)
assert_raises(self, "RAM.get_allocator('Local3')",
globals(), locals(), ValueError,
"allocator 'Local3' not found")
assert_raises(self, "RAM.remove_allocator('Local3')",
globals(), locals(), ValueError,
"allocator 'Local3' not found")
assert_raises(self, "LocalAllocator('BadLoad', max_load=-2)",
globals(), locals(), ValueError,
"BadLoad: max_load must be > 0, got -2")
def test_configure(self):
logging.debug('')
logging.debug('test_configure')
# Reconfigure.
with open('resources.cfg', 'w') as out:
out.write("""
[LocalHost]
max_load: 100
""")
local = RAM.get_allocator('LocalHost')
max_load = local.max_load
try:
self.assertTrue(max_load < 100)
RAM.configure('resources.cfg')
self.assertEqual(local.max_load, 100)
local.max_load = max_load
finally:
os.remove('resources.cfg')
# Add another local.
with open('resources.cfg', 'w') as out:
out.write("""
[Local2]
classname: openmdao.main.resource.LocalAllocator
authkey: PublicKey
allow_shell: False
total_cpus: 42
max_load: 200
""")
try:
RAM.configure('resources.cfg')
local2 = RAM.get_allocator('Local2')
self.assertEqual(local2.factory._authkey, 'PublicKey')
self.assertEqual(local2.factory._allow_shell, False)
self.assertEqual(local2.total_cpus, 42)
self.assertEqual(local2.max_load, 200)
self.assertEqual(local2.host, socket.gethostname())
self.assertTrue(local2.pid > 0)
RAM.remove_allocator('Local2')
finally:
os.remove('resources.cfg')
# Bad local total_cpus.
with open('resources.cfg', 'w') as out:
out.write("""
[Local2]
classname: openmdao.main.resource.LocalAllocator
total_cpus: 0
""")
try:
assert_raises(self, "RAM.configure('resources.cfg')", globals(),
locals(), ValueError,
'Local2: total_cpus must be > 0, got 0')
finally:
os.remove('resources.cfg')
# Bad local max_load.
with open('resources.cfg', 'w') as out:
out.write("""
[Local2]
classname: openmdao.main.resource.LocalAllocator
max_load: 0
""")
try:
assert_raises(self, "RAM.configure('resources.cfg')", globals(),
locals(), ValueError,
'Local2: max_load must be > 0, got 0')
finally:
os.remove('resources.cfg')
# Bad module.
with open('resources.cfg', 'w') as out:
out.write("""
[BadModule]
classname: no-such-module.Allocator
max_load: 100
""")
try:
assert_raises(
self, "RAM.configure('resources.cfg')", globals(), locals(),
RuntimeError, "RAM configure BadModule: can't import"
" 'no-such-module'")
finally:
os.remove('resources.cfg')
# Bad class.
with open('resources.cfg', 'w') as out:
out.write("""
[BadClass]
classname: openmdao.main.resource.NoSuchAllocator
max_load: 100
""")
try:
assert_raises(
self, "RAM.configure('resources.cfg')", globals(), locals(),
RuntimeError, "RAM configure BadClass: no class"
" 'NoSuchAllocator' in openmdao.main.resource")
finally:
os.remove('resources.cfg')
# Add, insert, get, remove.
local3 = LocalAllocator('Local3')
local4 = LocalAllocator('Local4', total_cpus=4)
RAM.add_allocator(local3)
try:
allocator_names = \
[allocator.name for allocator in RAM.list_allocators()]
self.assertEqual(allocator_names, ['LocalHost', 'Local3'])
self.assertTrue(RAM.get_allocator('Local3') is local3)
self.assertTrue(RAM.get_allocator(1) is local3)
RAM.insert_allocator(0, local4)
try:
allocator_names = \
[allocator.name for allocator in RAM.list_allocators()]
self.assertEqual(allocator_names,
['Local4', 'LocalHost', 'Local3'])
finally:
RAM.remove_allocator('Local4')
finally:
RAM.remove_allocator(1)
assert_raises(self, "RAM.get_allocator('Local3')", globals(), locals(),
ValueError, "allocator 'Local3' not found")
assert_raises(self, "RAM.remove_allocator('Local3')", globals(),
locals(), ValueError, "allocator 'Local3' not found")
assert_raises(self, "LocalAllocator('BadLoad', max_load=-2)",
globals(), locals(), ValueError,
"BadLoad: max_load must be > 0, got -2")
FileMetadata('patran.out', output=True, binary=True),
FileMetadata('Config' + self.config + '.cub', output=True, binary=True)]
self.resources = {'localhost': False}
print 'Job Complete ...'
# -----------------------------
# --- Execute the component ---
# -----------------------------
super(MeshComp, self).execute()
# --- Copy files to local directories
shutil.copy2(Path('OpenMDAO') + '/patran.out', Path('Patran') + '\Config' + self.config)
shutil.copy2(Path('OpenMDAO') + '/Config' + self.config + '.cub', Path('Cubit'))
if __name__ == "__main__":
enable_console()
logging.getLogger().setLevel(logging.DEBUG)
# --- Create 4 allocators (redundancy for reliability) adding each to the manager ---
allocator1 = NAS_Allocator(name='PFE20_DMZ1',
dmz_host='dmzfs1.nas.nasa.gov',
server_host='pfe20')
RAM.add_allocator(allocator1)
Remote_Mesh_Comp = MeshComp()
Remote_Mesh_Comp.run()
def rundlcs(envcmd = None, options=None, args=None, batch_size=5):
"""
run the whole process, including startup and shutdown
to do:
parse input
create load cases
create app assembly
create dispatcher
send cases and app to dispatcher
run cases
collect and save output
envcmd: a text string cmd (e.g. 'source env.sh') to set up the environment for the cluster allocator
"""
if (batch_size != None and line_count(options.cases)-1 > batch_size):
rundlcs_bybatch(envcmd, options, args, batch_size)
return
if options==None:
options, args = get_options()
print options
ctrl = parse_input(options)
# ctrl will be just the input, but broken up into separate categories, e.g.
# ctrl.cases, ctrl.app, ctrl.dispatch, ...
# work in progress; running efficiently at NREL.
if (options.cluster_allocator):
# cluster=ClusterAllocator()
### never had the guts to try this yet!
# env = os.environ
# fname = "%s/.env.sh" % (env['HOME'])
# fout = file(fname, "w")
# for key in env:
# fout.write("export %s=%s\n" % (key,env[key]))
# fout.close()
###
if envcmd == None:
cluster=ClusterAllocator()
else:
cluster=ClusterAllocator(use_modules=False, beforestart=envcmd)
# cluster=ClusterAllocator(use_modules=False, beforestart=". %s;" % fname)
RAM.remove_allocator('LocalHost')
RAM.add_allocator(cluster)
# RAM.insert_allocator(0,cluster)
### using "factory" functions to create specific subclasses (e.g. distinguish between FAST and HAWC2)
# Then we use these to create the cases...
case_params = ctrl.cases
casetab = GenericRunCaseTable()
casetab.initFromFile(case_params['source_file'], verbose=True, start_at = options.start_at)
# solver...
solver = 'FAST'
# solver = 'HAWC2'
if solver=='FAST':
## TODO, changed when we have a real turbine
# aero code stuff: for constructors
aerocode = openFAST(ctrl.output) ## need better name than output_params
aerocode.setOutput(ctrl.output)
elif solver == 'HAWC2':
aerocoe = openHAWC2(None)
raise NotImplementedError, "HAWC2 aeroecode wrapper not implemented in runBatch.py yet"
else:
raise ValueError, "unknown aerocode: %s" % solver
# case iterator
dispatcher = CaseAnalyzer(ctrl.dispatcher)
### After this point everything should be generic, all appropriate subclass object created
# # # # # # # # # # #
dispatcher.presetup_workflow(aerocode, casetab.cases) # just makes sure parts are there when configure() is called
dispatcher.configure()
# Now tell the dispatcher to (setup and ) run the cases using the aerocode on the turbine.
# calling configure() is done inside run(). but now it is done already (above), too.
# norun does not write directories, but it does set us up to process them if they already exist
if (not options.norun):
dispatcher.run()
# TODO: more complexity will be needed for difference between "run now" and "run later" cases.
dispatcher.collect_output(ctrl.output)
def rundlcs(envcmd = None, options=None, args=None, batch_size=100):
"""
run the whole process, including startup and shutdown
to do:
parse input
create load cases
create app assembly
create dispatcher
send cases and app to dispatcher
run cases
collect and save output
envcmd: a text string cmd (e.g. 'source env.sh') to set up the environment for the cluster allocator
"""
if (batch_size != None and line_count(options.cases)-1 > batch_size):
rundlcs_bybatch(envcmd, options, args, batch_size)
return
if options==None:
options, args = get_options()
print options
ctrl = parse_input(options)
# ctrl will be just the input, but broken up into separate categories, e.g.
# ctrl.cases, ctrl.app, ctrl.dispatch, ...
# work in progress; running efficiently at NREL.
if (options.cluster_allocator):
# cluster=ClusterAllocator()
### never had the guts to try this yet!
# env = os.environ
# fname = "%s/.env.sh" % (env['HOME'])
# fout = file(fname, "w")
# for key in env:
# fout.write("export %s=%s\n" % (key,env[key]))
# fout.close()
###
if envcmd == None:
cluster=ClusterAllocator()
else:
cluster=ClusterAllocator(use_modules=False, beforestart=envcmd)
# cluster=ClusterAllocator(use_modules=False, beforestart=". %s;" % fname)
RAM.remove_allocator('LocalHost')
RAM.add_allocator(cluster)
# RAM.insert_allocator(0,cluster)
### using "factory" functions to create specific subclasses (e.g. distinguish between FAST and HAWC2)
# Then we use these to create the cases...
case_params = ctrl.cases
casetab = GenericRunCaseTable()
casetab.initFromFile(case_params['source_file'], verbose=True, start_at = options.start_at)
# solver...
solver = 'FAST'
# solver = 'HAWC2'
if solver=='FAST':
## TODO, changed when we have a real turbine
# aero code stuff: for constructors
aerocode = openFAST(ctrl.output) ## need better name than output_params
aerocode.setOutput(ctrl.output)
elif solver == 'HAWC2':
aerocoe = openHAWC2(None)
raise NotImplementedError, "HAWC2 aeroecode wrapper not implemented in runBatch.py yet"
else:
raise ValueError, "unknown aerocode: %s" % solver
# case iterator
dispatcher = CaseAnalyzer(ctrl.dispatcher)
### After this point everything should be generic, all appropriate subclass object created
# # # # # # # # # # #
dispatcher.presetup_workflow(aerocode, casetab.cases) # just makes sure parts are there when configure() is called
dispatcher.configure()
# Now tell the dispatcher to (setup and ) run the cases using the aerocode on the turbine.
# calling configure() is done inside run(). but now it is done already (above), too.
# norun does not write directories, but it does set us up to process them if they already exist
if (not options.norun):
dispatcher.run()
# TODO: more complexity will be needed for difference between "run now" and "run later" cases.
dispatcher.collect_output(ctrl.output)
