class ProcessDispatcherSimpleAPIClient(object):
# State to use when state returned from PD is None
unknown_state = "400-PENDING"
state_map = {
ProcessStateEnum.SPAWN: '500-RUNNING',
ProcessStateEnum.TERMINATE: '700-TERMINATED',
ProcessStateEnum.ERROR: '850-FAILED'
}
def __init__(self, name, **kwargs):
self.real_client = ProcessDispatcherServiceClient(to_name=name, **kwargs)
self.event_pub = EventPublisher()
def dispatch_process(self, upid, spec, subscribers, constraints=None,
immediate=False):
name = spec.get('name')
self.event_pub.publish_event(event_type="ProcessLifecycleEvent",
origin=name, origin_type="DispatchedHAProcess",
state=ProcessStateEnum.SPAWN)
process_def = ProcessDefinition(name=name)
process_def.executable = {'module': spec.get('module'),
'class': spec.get('class')}
process_def_id = self.real_client.create_process_definition(process_def)
pid = self.real_client.create_process(process_def_id)
process_schedule = ProcessSchedule()
sched_pid = self.real_client.schedule_process(process_def_id,
process_schedule, configuration={}, process_id=pid)
proc = self.real_client.read_process(sched_pid)
dict_proc = {'upid': proc.process_id,
'state': self.state_map.get(proc.process_state, self.unknown_state),
}
return dict_proc
def terminate_process(self, pid):
return self.real_client.cancel_process(pid)
def describe_processes(self):
procs = self.real_client.list_processes()
dict_procs = []
for proc in procs:
dict_proc = {'upid': proc.process_id,
'state': self.state_map.get(proc.process_state, self.unknown_state),
}
dict_procs.append(dict_proc)
return dict_procs
class BaseHighAvailabilityAgentTest(IonIntegrationTestCase):
@needs_epu
def setUp(self):
self._start_container()
self.container.start_rel_from_url('res/deploy/r2cei.yml')
#self.pd_cli = ProcessDispatcherServiceClient(node=self.container.node)
self.pd_cli = ProcessDispatcherServiceClient(to_name="process_dispatcher")
self.process_definition_id = uuid4().hex
self.process_definition_name = 'test_haagent_%s' % self.process_definition_id
self.process_definition = ProcessDefinition(name=self.process_definition_name, executable={
'module': 'ion.agents.cei.test.test_haagent',
'class': 'TestProcess'
})
self.pd_cli.create_process_definition(self.process_definition, self.process_definition_id)
service_definition = SERVICE_DEFINITION_TMPL % self.process_definition_name
sd = IonObject(RT.ServiceDefinition, {"name": self.process_definition_name,
"definition": service_definition})
self.service_def_id, _ = self.container.resource_registry.create(sd)
self.resource_id = "haagent_1234"
self._haa_name = "high_availability_agent"
self._haa_dashi_name = "dashi_haa_" + uuid4().hex
self._haa_dashi_uri = get_dashi_uri_from_cfg()
self._haa_dashi_exchange = "hatests"
self._haa_config = self._get_haagent_config()
self._base_services, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
self._base_procs = self.pd_cli.list_processes()
self.waiter = ProcessStateWaiter()
self.waiter.start()
self.container_client = ContainerAgentClient(node=self.container.node,
name=self.container.name)
self._spawn_haagent()
self._setup_haa_client()
def _get_haagent_config(self):
return {
'highavailability': {
'policy': {
'interval': 1,
'name': 'npreserving',
'parameters': {
'preserve_n': 0
}
},
'process_definition_id': self.process_definition_id,
'dashi_messaging': True,
'dashi_exchange': self._haa_dashi_exchange,
'dashi_name': self._haa_dashi_name
},
'agent': {'resource_id': self.resource_id},
}
def _setup_haa_client(self):
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = SimpleResourceAgentClient(self.resource_id, process=FakeProcess())
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
def _spawn_haagent(self, policy_parameters=None):
config = deepcopy(self._haa_config)
if policy_parameters is not None:
config['highavailability']['policy']['parameters'] = policy_parameters
self._haa_pid = self.container_client.spawn_process(name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent", config=config)
def _kill_haagent(self):
self.container.terminate_process(self._haa_pid)
def tearDown(self):
new_policy = {'preserve_n': 0}
self.haa_client.reconfigure_policy(new_policy)
self.assertEqual(len(self.get_running_procs()), 0)
self.await_ha_state('STEADY')
self.waiter.stop()
try:
self._kill_haagent()
except BadRequest:
log.warning("Couldn't terminate HA Agent in teardown (May have been terminated by a test)")
self.container.resource_registry.delete(self.service_def_id, del_associations=True)
self._stop_container()
def get_running_procs(self):
"""returns a normalized set of running procs (removes the ones that
were there at setup time)
"""
base = self._base_procs
base_pids = [proc.process_id for proc in base]
current = self.pd_cli.list_processes()
current_pids = [proc.process_id for proc in current]
print "filtering base procs %s from %s" % (base_pids, current_pids)
normal = [cproc for cproc in current if cproc.process_id not in base_pids and cproc.process_state == ProcessStateEnum.RUNNING]
return normal
def get_new_services(self):
base = self._base_services
base_names = [i.name for i in base]
services_registered, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
normal = [cserv for cserv in services_registered if cserv.name not in base_names]
return normal
def await_ha_state(self, want_state, timeout=20):
for i in range(0, timeout):
try:
status = self.haa_client.status().result
if status == want_state:
return
else:
procs = self.get_running_procs()
num_procs = len(procs)
log.debug("assert wants state %s, got state %s, with %s procs" % (want_state,status, num_procs))
except Exception:
log.exception("Problem getting HA status, trying again...")
gevent.sleep(1)
raise Exception("Took more than %s to get to ha state %s" % (timeout, want_state))
def await_pyon_ha_state(self, want_state, timeout=20):
for i in range(0, timeout):
try:
result = self.haa_client.dump().result
service_id = result.get('service_id')
service = self.container.resource_registry.read(service_id)
if service.state == want_state:
return
else:
log.debug("want state %s, got state %s") % (want_state, service.state)
except Exception:
log.exception("Problem getting HA status, trying again...")
gevent.sleep(1)
raise Exception("Took more than %s to get to pyon ha state %s" % (timeout, want_state))
class HighAvailabilityAgentSensorPolicyTest(IonIntegrationTestCase):
def _start_webserver(self, port=None):
""" Start a webserver for testing code download
Note: tries really hard to get a port, and if it can't use
the suggested port, randomly picks another, and returns it
"""
def log_message(self, format, *args):
#swallow log massages
pass
class TestRequestHandler(BaseHTTPServer.BaseHTTPRequestHandler):
server_version = 'test_server'
extensions_map = ''
def do_GET(self):
self.send_response(200)
self.send_header("Content-type", "text/plain")
self.send_header("Content-Length", len(self.server.response))
self.end_headers()
self.wfile.write(self.server.response)
class Server(HTTPServer):
response = ''
def serve_forever(self):
self._serving = 1
while self._serving:
self.handle_request()
def stop(self):
self._serving = 0
if port is None:
port = 8008
Handler = TestRequestHandler
Handler.log_message = log_message
for i in range(0, 100):
try:
self._webserver = Server(("localhost", port), Handler)
except socket.error:
print "port %s is in use, picking another" % port
port = randint(8000, 10000)
continue
else:
break
self._web_glet = gevent.spawn(self._webserver.serve_forever)
return port
def _stop_webserver(self):
if self._webserver is not None:
self._webserver.stop()
gevent.sleep(2)
self._web_glet.kill()
def await_ha_state(self, want_state, timeout=20):
for i in range(0, timeout):
try:
status = self.haa_client.status().result
if status == want_state:
return
else:
procs = self.get_running_procs()
num_procs = len(procs)
log.debug("assert wants state %s, got state %s, with %s procs" % (want_state,status, num_procs))
except Exception:
log.exception("Problem getting HA status, trying again...")
gevent.sleep(1)
raise Exception("Took more than %s to get to ha state %s" % (timeout, want_state))
@needs_epu
def setUp(self):
self._start_container()
self.container.start_rel_from_url('res/deploy/r2cei.yml')
self.pd_cli = ProcessDispatcherServiceClient(to_name="process_dispatcher")
self.process_definition_id = uuid4().hex
self.process_definition = ProcessDefinition(name='test', executable={
'module': 'ion.agents.cei.test.test_haagent',
'class': 'TestProcess'
})
self.pd_cli.create_process_definition(self.process_definition,
self.process_definition_id)
http_port = 8919
http_port = self._start_webserver(port=http_port)
self.resource_id = "haagent_4567"
self._haa_name = "high_availability_agent"
self._haa_config = {
'server': {
'trafficsentinel': {
'host': 'localhost',
'port': http_port,
'protocol': 'http',
'username': '******',
'password': '******'
}
},
'highavailability': {
'policy': {
'interval': 1,
'name': 'sensor',
'parameters': {
'metric': 'app_attributes:ml',
'sample_period': 600,
'sample_function': 'Average',
'cooldown_period': 5,
'scale_up_threshold': 2.0,
'scale_up_n_processes': 1,
'scale_down_threshold': 1.0,
'scale_down_n_processes': 1,
'maximum_processes': 5,
'minimum_processes': 1,
}
},
'process_definition_id': self.process_definition_id,
"process_dispatchers": [
'process_dispatcher'
]
},
'agent': {'resource_id': self.resource_id},
}
self._base_procs = self.pd_cli.list_processes()
self.waiter = ProcessStateWaiter()
self.waiter.start()
self.container_client = ContainerAgentClient(node=self.container.node,
name=self.container.name)
self._haa_pid = self.container_client.spawn_process(name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent", config=self._haa_config)
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = SimpleResourceAgentClient(self.resource_id, process=FakeProcess())
log.info('Got haa client %s.', str(self._haa_pyon_client))
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
def tearDown(self):
new_policy = { 'metric': 'app_attributes:ml',
'sample_period': 600,
'sample_function': 'Average',
'cooldown_period': 0,
'scale_up_threshold': 2.0,
'scale_up_n_processes': 1,
'scale_down_threshold': 1.0,
'scale_down_n_processes': 1,
'maximum_processes': 0,
'minimum_processes': 0,
}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 0)
self.waiter.stop()
self.container.terminate_process(self._haa_pid)
self._stop_webserver()
self._stop_container()
def get_running_procs(self):
"""returns a normalized set of running procs (removes the ones that
were there at setup time)
"""
base = self._base_procs
base_pids = [proc.process_id for proc in base]
current = self.pd_cli.list_processes()
current_pids = [proc.process_id for proc in current]
print "filtering base procs %s from %s" % (base_pids, current_pids)
normal = [cproc for cproc in current if cproc.process_id not in base_pids and cproc.process_state == ProcessStateEnum.RUNNING]
return normal
def _get_managed_upids(self):
result = self.haa_client.dump().result
upids = result['managed_upids']
return upids
def _set_response(self, response):
self._webserver.response = response
def test_sensor_policy(self):
status = self.haa_client.status().result
# Ensure HA hasn't already failed
assert status in ('PENDING', 'READY', 'STEADY')
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 1)
self.await_ha_state('STEADY')
# Set ml for each proc such that we scale up
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "pid=%s&ml=5\n" % upid
self._set_response(response)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 2)
# Set ml so we stay steady
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "pid=%s&ml=1.5\n" % upid
self._set_response(response)
self.assertEqual(len(self.get_running_procs()), 2)
self.await_ha_state('STEADY')
# Set ml so we scale down
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "pid=%s&ml=0.5\n" % upid
self._set_response(response)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 1)
self.await_ha_state('STEADY')
class ProcessLeak(Plugin):
"""
Plugin to detect runaway processes in tests on the server side. It is intended to work against a live process
dispatcher service during the entire nose run. Therefore, it works in conjunction with --with-pycc plugin
or a light CEI or full launch. Use greenleak plugin instead for leak detection in non-pycc mode.
"""
name = "processleak"
def __init__(self):
Plugin.__init__(self)
def options(self, parser, env):
super(ProcessLeak, self).options(parser, env=env)
parser.add_option("--list-procs", action="store_true", dest="list_procs", \
help="List all new procs created during the test")
def configure(self, options, conf):
super(ProcessLeak, self).configure(options, conf)
self._list_procs = options.list_procs
def begin(self):
from interface.services.cei.iprocess_dispatcher_service import ProcessDispatcherServiceClient
from pyon.net.messaging import make_node
from pyon.core import bootstrap
from pyon.public import CFG
self.rpc_timeout = 2
self.base_pids = []
self._procs_by_test = {}
if not bootstrap.pyon_initialized:
bootstrap.bootstrap_pyon()
self.node, self.ioloop = make_node()
self.node.setup_interceptors(CFG.interceptor)
self.pd_cli = ProcessDispatcherServiceClient(node=self.node)
# Set base_pids once
from pyon.core.exception import Timeout
try:
self.base_pids = [ proc.process_id for proc in self.pd_cli.list_processes(timeout=20) ]
except Timeout:
pass
def beforeTest(self, test):
from pyon.core.exception import Timeout
#Only populate self.base_pids if it' still empty
if not self.base_pids:
try:
self.base_pids = [ proc.process_id for proc in self.pd_cli.list_processes(timeout=self.rpc_timeout) ]
except Timeout:
pass
def afterTest(self, test):
from pyon.core.exception import Timeout
from interface.objects import ProcessStateEnum
try:
current = self.pd_cli.list_processes(timeout=self.rpc_timeout)
from interface.objects import ProcessStateEnum
procs_leaked = [ (proc.process_id, ProcessStateEnum._str_map.get(proc.process_state)) for proc in current \
if proc.process_id not in self.base_pids and proc.process_state \
not in [ProcessStateEnum.TERMINATED, ProcessStateEnum.EXITED] ]
if self._list_procs:
all_procs = [ (proc.process_id, ProcessStateEnum._str_map.get(proc.process_state)) for proc in current \
if proc.process_id not in self.base_pids ]
if len(procs_leaked) > 0:
[ self.pd_cli.cancel_process(proc[0], timeout=self.rpc_timeout) for proc in procs_leaked ]
if self._list_procs:
self._procs_by_test[test.id()] = all_procs
else:
self._procs_by_test[test.id()] = procs_leaked
except Timeout:
pass
def report(self, stream):
table = []
for tid, procs in self._procs_by_test.iteritems():
# printable tid: don't need the full path
ptid = ".".join(tid.split(".")[-2:])
table.append([ptid, ""])
for proc in procs:
table.append(["", str(proc)])
table.append(["", ""])
# header
table.insert(0, ['Test', 'Leaked Processes'])
# get widths
widths = [max([len(row[x]) for row in table]) for x in xrange(len(table[0]))]
fmt_out = [" ".join([x.ljust(widths[i]) for i, x in enumerate(row)]) for row in table]
# insert col separation row
fmt_out.insert(1, " ".join([''.ljust(widths[i], '=') for i in xrange(len(widths))]))
# write this all to sstream
stream.write("Process leak report\n")
stream.write("\n".join(fmt_out))
stream.write("\n")
def finalize(self, result):
self.node.stop_node()
self.ioloop.join()
class ProcessLeak(Plugin):
"""
Plugin to detect runaway processes in tests on the server side. It is intended to work against a live process
dispatcher service during the entire nose run. Therefore, it works in conjunction with --with-pycc plugin
or a light CEI or full launch. Use greenleak plugin instead for leak detection in non-pycc mode.
"""
name = "processleak"
def __init__(self):
Plugin.__init__(self)
def options(self, parser, env):
super(ProcessLeak, self).options(parser, env=env)
parser.add_option("--list-procs", action="store_true", dest="list_procs", \
help="List all new procs created during the test")
def configure(self, options, conf):
super(ProcessLeak, self).configure(options, conf)
self._list_procs = options.list_procs
def begin(self):
from interface.services.cei.iprocess_dispatcher_service import ProcessDispatcherServiceClient
from pyon.net.messaging import make_node
from pyon.core import bootstrap
from pyon.public import CFG
self.base_pids = []
self.rpc_timeout = 2
self._procs_by_test = {}
if not bootstrap.pyon_initialized:
bootstrap.bootstrap_pyon()
self.node, self.ioloop = make_node()
self.node.setup_interceptors(CFG.interceptor)
self.pd_cli = ProcessDispatcherServiceClient(node=self.node)
def beforeTest(self, test):
from pyon.core.exception import Timeout
try:
self.base_pids = [
proc.process_id for proc in self.pd_cli.list_processes(
timeout=self.rpc_timeout)
]
except Timeout:
pass
def afterTest(self, test):
from pyon.core.exception import Timeout
from interface.objects import ProcessStateEnum
try:
current = self.pd_cli.list_processes(timeout=self.rpc_timeout)
from interface.objects import ProcessStateEnum
procs_leaked = [ (proc.process_id, ProcessStateEnum._str_map.get(proc.process_state)) for proc in current \
if proc.process_id not in self.base_pids and proc.process_state \
not in [ProcessStateEnum.TERMINATED, ProcessStateEnum.EXITED] ]
if self._list_procs:
all_procs = [ (proc.process_id, ProcessStateEnum._str_map.get(proc.process_state)) for proc in current \
if proc.process_id not in self.base_pids ]
if len(procs_leaked) > 0:
[
self.pd_cli.cancel_process(proc[0],
timeout=self.rpc_timeout)
for proc in procs_leaked
]
if self._list_procs:
self._procs_by_test[test.id()] = all_procs
else:
self._procs_by_test[test.id()] = procs_leaked
except Timeout:
pass
def report(self, stream):
table = []
for tid, procs in self._procs_by_test.iteritems():
# printable tid: don't need the full path
ptid = ".".join(tid.split(".")[-2:])
table.append([ptid, ""])
for proc in procs:
table.append(["", str(proc)])
table.append(["", ""])
# header
table.insert(0, ['Test', 'Leaked Processes'])
# get widths
widths = [
max([len(row[x]) for row in table]) for x in xrange(len(table[0]))
]
fmt_out = [
" ".join([x.ljust(widths[i]) for i, x in enumerate(row)])
for row in table
]
# insert col separation row
fmt_out.insert(
1,
" ".join([''.ljust(widths[i], '=') for i in xrange(len(widths))]))
# write this all to sstream
stream.write("Process leak report\n")
stream.write("\n".join(fmt_out))
stream.write("\n")
def finalize(self, result):
self.node.stop_node()
self.ioloop.join()
class HighAvailabilityAgentTest(IonIntegrationTestCase):
@needs_epu
def setUp(self):
self._start_container()
self.container.start_rel_from_url('res/deploy/r2cei.yml')
#self.pd_cli = ProcessDispatcherServiceClient(node=self.container.node)
self.pd_cli = ProcessDispatcherServiceClient(to_name="process_dispatcher")
self.process_definition_id = uuid4().hex
self.process_definition_name = 'test'
self.process_definition = ProcessDefinition(name=self.process_definition_name, executable={
'module': 'ion.agents.cei.test.test_haagent',
'class': 'TestProcess'
})
self.pd_cli.create_process_definition(self.process_definition, self.process_definition_id)
self.resource_id = "haagent_1234"
self._haa_name = "high_availability_agent"
self._haa_dashi_name = "dashi_haa_" + uuid4().hex
self._haa_dashi_uri = get_dashi_uri_from_cfg()
self._haa_dashi_exchange = "%s.hatests" % bootstrap.get_sys_name()
self._haa_config = {
'highavailability': {
'policy': {
'interval': 1,
'name': 'npreserving',
'parameters': {
'preserve_n': 0
}
},
'process_definition_id': self.process_definition_id,
'dashi_messaging' : True,
'dashi_exchange' : self._haa_dashi_exchange,
'dashi_name': self._haa_dashi_name
},
'agent': {'resource_id': self.resource_id},
}
self._base_services, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
self._base_procs = self.pd_cli.list_processes()
self.waiter = ProcessStateWaiter()
self.waiter.start()
self.container_client = ContainerAgentClient(node=self.container.node,
name=self.container.name)
self._haa_pid = self.container_client.spawn_process(name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent", config=self._haa_config)
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = SimpleResourceAgentClient(self.resource_id, process=FakeProcess())
log.info('Got haa client %s.', str(self._haa_pyon_client))
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
def tearDown(self):
self.waiter.stop()
try:
self.container.terminate_process(self._haa_pid)
except BadRequest:
log.warning("Couldn't terminate HA Agent in teardown (May have been terminated by a test)")
self._stop_container()
def get_running_procs(self):
"""returns a normalized set of running procs (removes the ones that
were there at setup time)
"""
base = self._base_procs
base_pids = [proc.process_id for proc in base]
current = self.pd_cli.list_processes()
current_pids = [proc.process_id for proc in current]
print "filtering base procs %s from %s" % (base_pids, current_pids)
normal = [cproc for cproc in current if cproc.process_id not in base_pids and cproc.process_state == ProcessStateEnum.RUNNING]
return normal
def get_new_services(self):
base = self._base_services
base_names = [i.name for i in base]
services_registered, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
current_names = [i.name for i in services_registered]
normal = [cserv for cserv in services_registered if cserv.name not in base_names]
return normal
def await_ha_state(self, want_state, timeout=10):
for i in range(0, timeout):
status = self.haa_client.status().result
if status == want_state:
return
gevent.sleep(1)
raise Exception("Took more than %s to get to ha state %s" % (timeout, want_state))
def test_features(self):
status = self.haa_client.status().result
# Ensure HA hasn't already failed
assert status in ('PENDING', 'READY', 'STEADY')
# verifies L4-CI-CEI-RQ44
# Note: the HA agent is started in the setUp() method, with config
# pointing to the test "service". The initial config is set to preserve
# 0 service processes. With this reconfigure step below, we change that
# to launch 1.
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
result = self.haa_client.dump().result
self.assertEqual(result['policy'], new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 1)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, 'STEADY')
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
# verifies L4-CI-CEI-RQ122 and L4-CI-CEI-RQ124
new_policy = {'preserve_n': 2}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 2)
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 1)
new_policy = {'preserve_n': 0}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 0)
def test_associations(self):
# Ensure that once the HA Agent starts, there is a Service object in
# the registry
result = self.haa_client.dump().result
service_id = result.get('service_id')
self.assertIsNotNone(service_id)
service = self.container.resource_registry.read(service_id)
self.assertIsNotNone(service)
# Ensure that once a process is started, there is an association between
# it and the service
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 1)
self.await_ha_state('STEADY')
proc = self.get_running_procs()[0]
processes_associated, _ = self.container.resource_registry.find_resources(
restype="Process", name=proc.process_id)
self.assertEqual(len(processes_associated), 1)
has_processes = self.container.resource_registry.find_associations(
service, "hasProcess")
self.assertEqual(len(has_processes), 1)
self.await_ha_state('STEADY')
# Ensure that once we terminate that process, there are no associations
new_policy = {'preserve_n': 0}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 0)
processes_associated, _ = self.container.resource_registry.find_resources(
restype="Process", name=proc.process_id)
self.assertEqual(len(processes_associated), 0)
has_processes = self.container.resource_registry.find_associations(
service, "hasProcess")
self.assertEqual(len(has_processes), 0)
# Ensure that once we terminate that HA Agent, the Service object is
# cleaned up
self.container.terminate_process(self._haa_pid)
with self.assertRaises(NotFound):
service = self.container.resource_registry.read(service_id)
def test_dashi(self):
import dashi
dashi_conn = dashi.DashiConnection("something", self._haa_dashi_uri,
self._haa_dashi_exchange)
status = dashi_conn.call(self._haa_dashi_name, "status")
assert status in ('PENDING', 'READY', 'STEADY')
new_policy = {'preserve_n': 0}
dashi_conn.call(self._haa_dashi_name, "reconfigure_policy",
new_policy=new_policy)
class HighAvailabilityAgentSensorPolicyTest(IonIntegrationTestCase):
def _start_webserver(self, port=None):
""" Start a webserver for testing code download
Note: tries really hard to get a port, and if it can't use
the suggested port, randomly picks another, and returns it
"""
def log_message(self, format, *args):
#swallow log massages
pass
class TestRequestHandler(BaseHTTPServer.BaseHTTPRequestHandler):
server_version = 'test_server'
extensions_map = ''
def do_GET(self):
self.send_response(200)
self.send_header("Content-type", "text/plain")
self.send_header("Content-Length", len(self.server.response))
self.end_headers()
self.wfile.write(self.server.response)
class Server(HTTPServer):
response = ''
def serve_forever(self):
self._serving = 1
while self._serving:
self.handle_request()
def stop(self):
self._serving = 0
if port is None:
port = 8008
Handler = TestRequestHandler
Handler.log_message = log_message
for i in range(0, 100):
try:
self._webserver = Server(("localhost", port), Handler)
except socket.error:
print "port %s is in use, picking another" % port
port = randint(8000, 10000)
continue
else:
break
self._web_glet = gevent.spawn(self._webserver.serve_forever)
return port
def _stop_webserver(self):
if self._webserver is not None:
self._webserver.stop()
gevent.sleep(2)
self._web_glet.kill()
@needs_epu
def setUp(self):
self._start_container()
self.container.start_rel_from_url('res/deploy/r2cei.yml')
self.pd_cli = ProcessDispatcherServiceClient(to_name="process_dispatcher")
self.process_definition_id = uuid4().hex
self.process_definition = ProcessDefinition(name='test', executable={
'module': 'ion.agents.cei.test.test_haagent',
'class': 'TestProcess'
})
self.pd_cli.create_process_definition(self.process_definition,
self.process_definition_id)
http_port = 8919
http_port = self._start_webserver(port=http_port)
self.resource_id = "haagent_4567"
self._haa_name = "high_availability_agent"
self._haa_config = {
'highavailability': {
'policy': {
'interval': 1,
'name': 'sensor',
'parameters': {
'metric': 'app_attributes:ml',
'sample_period': 600,
'sample_function': 'Average',
'cooldown_period': 20,
'scale_up_threshold': 2.0,
'scale_up_n_processes': 1,
'scale_down_threshold': 1.0,
'scale_down_n_processes': 1,
'maximum_processes': 5,
'minimum_processes': 1,
}
},
'aggregator': {
'type': 'trafficsentinel',
'host': 'localhost',
'port': http_port,
'protocol': 'http',
'username': '******',
'password': '******'
},
'process_definition_id': self.process_definition_id,
"process_dispatchers": [
'process_dispatcher'
]
},
'agent': {'resource_id': self.resource_id},
}
self._base_procs = self.pd_cli.list_processes()
self.waiter = ProcessStateWaiter()
self.waiter.start()
self.container_client = ContainerAgentClient(node=self.container.node,
name=self.container.name)
self._haa_pid = self.container_client.spawn_process(name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent", config=self._haa_config)
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = SimpleResourceAgentClient(self.resource_id, process=FakeProcess())
log.info('Got haa client %s.', str(self._haa_pyon_client))
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
def tearDown(self):
self.waiter.stop()
self.container.terminate_process(self._haa_pid)
self._stop_webserver()
self._stop_container()
def get_running_procs(self):
"""returns a normalized set of running procs (removes the ones that
were there at setup time)
"""
base = self._base_procs
base_pids = [proc.process_id for proc in base]
current = self.pd_cli.list_processes()
current_pids = [proc.process_id for proc in current]
print "filtering base procs %s from %s" % (base_pids, current_pids)
normal = [cproc for cproc in current if cproc.process_id not in base_pids and cproc.process_state == ProcessStateEnum.RUNNING]
return normal
def _get_managed_upids(self):
result = self.haa_client.dump().result
upids = result['managed_upids']
return upids
def _set_response(self, response):
self._webserver.response = response
def test_sensor_policy(self):
status = self.haa_client.status().result
# Ensure HA hasn't already failed
assert status in ('PENDING', 'READY', 'STEADY')
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 1)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, 'STEADY')
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
# Set ml for each proc such that we scale up
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "%s,ml=5\n"
self._set_response(response)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 2)
# Set ml so we stay steady
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "%s,ml=1.5\n"
self._set_response(response)
self.assertEqual(len(self.get_running_procs()), 2)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, 'STEADY')
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
# Set ml so we scale down
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "%s,ml=0.5\n"
self._set_response(response)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 1)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, 'STEADY')
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
class HighAvailabilityAgentTest(IonIntegrationTestCase):
def setUp(self):
self._start_container()
self.container.start_rel_from_url('res/deploy/r2cei.yml')
#self.pd_cli = ProcessDispatcherServiceClient(node=self.container.node)
self.pd_cli = ProcessDispatcherServiceClient(to_name="process_dispatcher")
self.resource_id = "haagent_1234"
self._haa_name = "high_availability_agent"
self._haa_config = {
'highavailability': {
'policy': {
'interval': 1,
'name': 'npreserving',
'parameters': {
'preserve_n': 0
}
},
'process_spec': {
'name': 'test',
'module': 'ion.agents.cei.test.test_haagent',
'class': 'TestProcess'
},
"process_dispatchers": [
'process_dispatcher'
]
},
'agent': {'resource_id': self.resource_id},
}
self._base_procs = self.pd_cli.list_processes()
self.container_client = ContainerAgentClient(node=self.container.node,
name=self.container.name)
self._haa_pid = self.container_client.spawn_process(name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent", config=self._haa_config)
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = ResourceAgentClient(self.resource_id, process=FakeProcess())
log.info('Got haa client %s.', str(self._haa_pyon_client))
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
self.event_queue = queue.Queue()
self.event_sub = None
def tearDown(self):
self.container.terminate_process(self._haa_pid)
self._stop_container()
def _event_callback(self, event, *args, **kwargs):
self.event_queue.put(event)
def subscribe_events(self, origin):
self.event_sub = EventSubscriber(event_type="ProcessLifecycleEvent",
callback=self._event_callback, origin_type="DispatchedProcess")
self.event_sub.start()
def await_state_event(self, pid, state):
event = self.event_queue.get(timeout=10)
log.debug("Got event: %s", event)
self.assertTrue(event.origin.startswith(pid))
self.assertEqual(event.state, state)
return event
def get_procs(self):
"""returns a normalized set of procs (removes the ones that were there
at setup time)
"""
base = self._base_procs
current = self.pd_cli.list_processes()
normal = [proc for proc in current if proc not in base]
return normal
@needs_epu
def test_features(self):
status = self.haa_client.status().result
self.assertEqual(status, 'PENDING')
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
result = self.haa_client.dump().result
self.assertEqual(result['policy'], new_policy)
self.subscribe_events(None)
self.await_state_event("test", ProcessStateEnum.SPAWN)
self.assertEqual(len(self.get_procs()), 1)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, 'STEADY')
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
new_policy = {'preserve_n': 2}
self.haa_client.reconfigure_policy(new_policy)
self.await_state_event("test", ProcessStateEnum.SPAWN)
self.assertEqual(len(self.get_procs()), 2)
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
self.await_state_event("test", ProcessStateEnum.TERMINATE)
self.assertEqual(len(self.get_procs()), 1)
new_policy = {'preserve_n': 0}
self.haa_client.reconfigure_policy(new_policy)
self.await_state_event("test", ProcessStateEnum.TERMINATE)
self.assertEqual(len(self.get_procs()), 0)
class HighAvailabilityAgentTest(IonIntegrationTestCase):
@needs_epu
def setUp(self):
self._start_container()
self.container.start_rel_from_url('res/deploy/r2cei.yml')
#self.pd_cli = ProcessDispatcherServiceClient(node=self.container.node)
self.pd_cli = ProcessDispatcherServiceClient(to_name="process_dispatcher")
self.process_definition_id = uuid4().hex
self.process_definition_name = 'test'
self.process_definition = ProcessDefinition(name=self.process_definition_name, executable={
'module': 'ion.agents.cei.test.test_haagent',
'class': 'TestProcess'
})
self.pd_cli.create_process_definition(self.process_definition, self.process_definition_id)
self.resource_id = "haagent_1234"
self._haa_name = "high_availability_agent"
self._haa_dashi_name = "dashi_haa_" + uuid4().hex
self._haa_dashi_uri = get_dashi_uri_from_cfg()
self._haa_dashi_exchange = "%s.hatests" % bootstrap.get_sys_name()
self._haa_config = {
'highavailability': {
'policy': {
'interval': 1,
'name': 'npreserving',
'parameters': {
'preserve_n': 0
}
},
'process_definition_id': self.process_definition_id,
'dashi_messaging' : True,
'dashi_exchange' : self._haa_dashi_exchange,
'dashi_name': self._haa_dashi_name
},
'agent': {'resource_id': self.resource_id},
}
self._base_services, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
self._base_procs = self.pd_cli.list_processes()
self.waiter = ProcessStateWaiter()
self.waiter.start()
self.container_client = ContainerAgentClient(node=self.container.node,
name=self.container.name)
self._haa_pid = self.container_client.spawn_process(name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent", config=self._haa_config)
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = SimpleResourceAgentClient(self.resource_id, process=FakeProcess())
log.info('Got haa client %s.', str(self._haa_pyon_client))
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
def tearDown(self):
self.waiter.stop()
try:
self.container.terminate_process(self._haa_pid)
except BadRequest:
log.warning("Couldn't terminate HA Agent in teardown (May have been terminated by a test)")
self._stop_container()
def get_running_procs(self):
"""returns a normalized set of running procs (removes the ones that
were there at setup time)
"""
base = self._base_procs
base_pids = [proc.process_id for proc in base]
current = self.pd_cli.list_processes()
current_pids = [proc.process_id for proc in current]
print "filtering base procs %s from %s" % (base_pids, current_pids)
normal = [cproc for cproc in current if cproc.process_id not in base_pids and cproc.process_state == ProcessStateEnum.RUNNING]
return normal
def get_new_services(self):
base = self._base_services
base_names = [i.name for i in base]
services_registered, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
current_names = [i.name for i in services_registered]
normal = [cserv for cserv in services_registered if cserv.name not in base_names]
return normal
def await_ha_state(self, want_state, timeout=10):
for i in range(0, timeout):
status = self.haa_client.status().result
if status == want_state:
return
gevent.sleep(1)
raise Exception("Took more than %s to get to ha state %s" % (timeout, want_state))
def test_features(self):
status = self.haa_client.status().result
# Ensure HA hasn't already failed
assert status in ('PENDING', 'READY', 'STEADY')
# verifies L4-CI-CEI-RQ44
# Note: the HA agent is started in the setUp() method, with config
# pointing to the test "service". The initial config is set to preserve
# 0 service processes. With this reconfigure step below, we change that
# to launch 1.
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
result = self.haa_client.dump().result
self.assertEqual(result['policy'], new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 1)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, 'STEADY')
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
# Ensure Service object has the correct state
result = self.haa_client.dump().result
service_id = result.get('service_id')
service = self.container.resource_registry.read(service_id)
self.assertEqual(service.state, ServiceStateEnum.STEADY)
# verifies L4-CI-CEI-RQ122 and L4-CI-CEI-RQ124
new_policy = {'preserve_n': 2}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 2)
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 1)
new_policy = {'preserve_n': 0}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 0)
def test_associations(self):
# Ensure that once the HA Agent starts, there is a Service object in
# the registry
result = self.haa_client.dump().result
service_id = result.get('service_id')
self.assertIsNotNone(service_id)
service = self.container.resource_registry.read(service_id)
self.assertIsNotNone(service)
# Ensure that once a process is started, there is an association between
# it and the service
new_policy = {'preserve_n': 1}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 1)
self.await_ha_state('STEADY')
proc = self.get_running_procs()[0]
processes_associated, _ = self.container.resource_registry.find_resources(
restype="Process", name=proc.process_id)
self.assertEqual(len(processes_associated), 1)
has_processes = self.container.resource_registry.find_associations(
service, "hasProcess")
self.assertEqual(len(has_processes), 1)
self.await_ha_state('STEADY')
# Ensure that once we terminate that process, there are no associations
new_policy = {'preserve_n': 0}
self.haa_client.reconfigure_policy(new_policy)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 0)
processes_associated, _ = self.container.resource_registry.find_resources(
restype="Process", name=proc.process_id)
self.assertEqual(len(processes_associated), 0)
has_processes = self.container.resource_registry.find_associations(
service, "hasProcess")
self.assertEqual(len(has_processes), 0)
# Ensure that once we terminate that HA Agent, the Service object is
# cleaned up
self.container.terminate_process(self._haa_pid)
with self.assertRaises(NotFound):
service = self.container.resource_registry.read(service_id)
def test_dashi(self):
import dashi
dashi_conn = dashi.DashiConnection("something", self._haa_dashi_uri,
self._haa_dashi_exchange)
status = dashi_conn.call(self._haa_dashi_name, "status")
assert status in ('PENDING', 'READY', 'STEADY')
new_policy = {'preserve_n': 0}
dashi_conn.call(self._haa_dashi_name, "reconfigure_policy",
new_policy=new_policy)
class HighAvailabilityAgentSensorPolicyTest(IonIntegrationTestCase):
def _start_webserver(self, port=None):
""" Start a webserver for testing code download
Note: tries really hard to get a port, and if it can't use
the suggested port, randomly picks another, and returns it
"""
def log_message(self, format, *args):
# swallow log massages
pass
class TestRequestHandler(BaseHTTPServer.BaseHTTPRequestHandler):
server_version = "test_server"
extensions_map = ""
def do_GET(self):
self.send_response(200)
self.send_header("Content-type", "text/plain")
self.send_header("Content-Length", len(self.server.response))
self.end_headers()
self.wfile.write(self.server.response)
class Server(HTTPServer):
response = ""
def serve_forever(self):
self._serving = 1
while self._serving:
self.handle_request()
def stop(self):
self._serving = 0
if port is None:
port = 8008
Handler = TestRequestHandler
Handler.log_message = log_message
for i in range(0, 100):
try:
self._webserver = Server(("localhost", port), Handler)
except socket.error:
print "port %s is in use, picking another" % port
port = randint(8000, 10000)
continue
else:
break
self._web_glet = gevent.spawn(self._webserver.serve_forever)
return port
def _stop_webserver(self):
if self._webserver is not None:
self._webserver.stop()
gevent.sleep(2)
self._web_glet.kill()
@needs_epu
def setUp(self):
self._start_container()
self.container.start_rel_from_url("res/deploy/r2cei.yml")
self.pd_cli = ProcessDispatcherServiceClient(to_name="process_dispatcher")
self.process_definition_id = uuid4().hex
self.process_definition = ProcessDefinition(
name="test", executable={"module": "ion.agents.cei.test.test_haagent", "class": "TestProcess"}
)
self.pd_cli.create_process_definition(self.process_definition, self.process_definition_id)
http_port = 8919
http_port = self._start_webserver(port=http_port)
self.resource_id = "haagent_4567"
self._haa_name = "high_availability_agent"
self._haa_config = {
"highavailability": {
"policy": {
"interval": 1,
"name": "sensor",
"parameters": {
"metric": "app_attributes:ml",
"sample_period": 600,
"sample_function": "Average",
"cooldown_period": 20,
"scale_up_threshold": 2.0,
"scale_up_n_processes": 1,
"scale_down_threshold": 1.0,
"scale_down_n_processes": 1,
"maximum_processes": 5,
"minimum_processes": 1,
},
},
"aggregator": {
"type": "trafficsentinel",
"host": "localhost",
"port": http_port,
"protocol": "http",
"username": "******",
"password": "******",
},
"process_definition_id": self.process_definition_id,
"process_dispatchers": ["process_dispatcher"],
},
"agent": {"resource_id": self.resource_id},
}
self._base_procs = self.pd_cli.list_processes()
self.waiter = ProcessStateWaiter()
self.waiter.start()
self.container_client = ContainerAgentClient(node=self.container.node, name=self.container.name)
self._haa_pid = self.container_client.spawn_process(
name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent",
config=self._haa_config,
)
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = SimpleResourceAgentClient(self.resource_id, process=FakeProcess())
log.info("Got haa client %s.", str(self._haa_pyon_client))
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
def tearDown(self):
self.waiter.stop()
self.container.terminate_process(self._haa_pid)
self._stop_webserver()
self._stop_container()
def get_running_procs(self):
"""returns a normalized set of running procs (removes the ones that
were there at setup time)
"""
base = self._base_procs
base_pids = [proc.process_id for proc in base]
current = self.pd_cli.list_processes()
current_pids = [proc.process_id for proc in current]
print "filtering base procs %s from %s" % (base_pids, current_pids)
normal = [
cproc
for cproc in current
if cproc.process_id not in base_pids and cproc.process_state == ProcessStateEnum.RUNNING
]
return normal
def _get_managed_upids(self):
result = self.haa_client.dump().result
upids = result["managed_upids"]
return upids
def _set_response(self, response):
self._webserver.response = response
def test_sensor_policy(self):
status = self.haa_client.status().result
# Ensure HA hasn't already failed
assert status in ("PENDING", "READY", "STEADY")
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 1)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, "STEADY")
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
# Set ml for each proc such that we scale up
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "%s,ml=5\n"
self._set_response(response)
self.waiter.await_state_event(state=ProcessStateEnum.RUNNING)
self.assertEqual(len(self.get_running_procs()), 2)
# Set ml so we stay steady
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "%s,ml=1.5\n"
self._set_response(response)
self.assertEqual(len(self.get_running_procs()), 2)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, "STEADY")
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
# Set ml so we scale down
upids = self._get_managed_upids()
response = ""
for upid in upids:
response += "%s,ml=0.5\n"
self._set_response(response)
self.waiter.await_state_event(state=ProcessStateEnum.TERMINATED)
self.assertEqual(len(self.get_running_procs()), 1)
for i in range(0, 5):
status = self.haa_client.status().result
try:
self.assertEqual(status, "STEADY")
break
except:
gevent.sleep(1)
else:
assert False, "HA Service took too long to get to state STEADY"
class BaseHighAvailabilityAgentTest(IonIntegrationTestCase):
@needs_epu
def setUp(self):
self._start_container()
self.container.start_rel_from_url('res/deploy/r2cei.yml')
#self.pd_cli = ProcessDispatcherServiceClient(node=self.container.node)
self.pd_cli = ProcessDispatcherServiceClient(
to_name="process_dispatcher")
self.process_definition_id = uuid4().hex
self.process_definition_name = 'test_haagent_%s' % self.process_definition_id
self.process_definition = ProcessDefinition(
name=self.process_definition_name,
executable={
'module': 'ion.agents.cei.test.test_haagent',
'class': 'TestProcess'
})
self.pd_cli.create_process_definition(self.process_definition,
self.process_definition_id)
service_definition = SERVICE_DEFINITION_TMPL % self.process_definition_name
sd = IonObject(RT.ServiceDefinition, {
"name": self.process_definition_name,
"definition": service_definition
})
self.service_def_id, _ = self.container.resource_registry.create(sd)
self.resource_id = "haagent_1234"
self._haa_name = "high_availability_agent"
self._haa_dashi_name = "dashi_haa_" + uuid4().hex
self._haa_dashi_uri = get_dashi_uri_from_cfg()
self._haa_dashi_exchange = "hatests"
self._haa_config = self._get_haagent_config()
self._base_services, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
self._base_procs = self.pd_cli.list_processes()
self.waiter = ProcessStateWaiter()
self.waiter.start()
self.container_client = ContainerAgentClient(node=self.container.node,
name=self.container.name)
self._spawn_haagent()
self._setup_haa_client()
def _get_haagent_config(self):
return {
'highavailability': {
'policy': {
'interval': 1,
'name': 'npreserving',
'parameters': {
'preserve_n': 0
}
},
'process_definition_id': self.process_definition_id,
'dashi_messaging': True,
'dashi_exchange': self._haa_dashi_exchange,
'dashi_name': self._haa_dashi_name
},
'agent': {
'resource_id': self.resource_id
},
}
def _setup_haa_client(self):
# Start a resource agent client to talk with the instrument agent.
self._haa_pyon_client = SimpleResourceAgentClient(
self.resource_id, process=FakeProcess())
self.haa_client = HighAvailabilityAgentClient(self._haa_pyon_client)
def _spawn_haagent(self, policy_parameters=None):
config = deepcopy(self._haa_config)
if policy_parameters is not None:
config['highavailability']['policy'][
'parameters'] = policy_parameters
self._haa_pid = self.container_client.spawn_process(
name=self._haa_name,
module="ion.agents.cei.high_availability_agent",
cls="HighAvailabilityAgent",
config=config)
def _kill_haagent(self):
self.container.terminate_process(self._haa_pid)
def tearDown(self):
new_policy = {'preserve_n': 0}
self.haa_client.reconfigure_policy(new_policy)
self.assertEqual(len(self.get_running_procs()), 0)
self.await_ha_state('STEADY')
self.waiter.stop()
try:
self._kill_haagent()
except BadRequest:
log.warning(
"Couldn't terminate HA Agent in teardown (May have been terminated by a test)"
)
self.container.resource_registry.delete(self.service_def_id,
del_associations=True)
self._stop_container()
def get_running_procs(self):
"""returns a normalized set of running procs (removes the ones that
were there at setup time)
"""
base = self._base_procs
base_pids = [proc.process_id for proc in base]
current = self.pd_cli.list_processes()
current_pids = [proc.process_id for proc in current]
print "filtering base procs %s from %s" % (base_pids, current_pids)
normal = [
cproc for cproc in current if cproc.process_id not in base_pids
and cproc.process_state == ProcessStateEnum.RUNNING
]
return normal
def get_new_services(self):
base = self._base_services
base_names = [i.name for i in base]
services_registered, _ = self.container.resource_registry.find_resources(
restype="Service", name=self.process_definition_name)
normal = [
cserv for cserv in services_registered
if cserv.name not in base_names
]
return normal
def await_ha_state(self, want_state, timeout=20):
for i in range(0, timeout):
try:
status = self.haa_client.status().result
if status == want_state:
return
else:
procs = self.get_running_procs()
num_procs = len(procs)
log.debug(
"assert wants state %s, got state %s, with %s procs" %
(want_state, status, num_procs))
except Exception:
log.exception("Problem getting HA status, trying again...")
gevent.sleep(1)
raise Exception("Took more than %s to get to ha state %s" %
(timeout, want_state))
def await_pyon_ha_state(self, want_state, timeout=20):
for i in range(0, timeout):
try:
result = self.haa_client.dump().result
service_id = result.get('service_id')
service = self.container.resource_registry.read(service_id)
if service.state == want_state:
return
else:
log.debug("want state %s, got state %s") % (want_state,
service.state)
except Exception:
log.exception("Problem getting HA status, trying again...")
gevent.sleep(1)
raise Exception("Took more than %s to get to pyon ha state %s" %
(timeout, want_state))
