def on_client_ready(_):
dispatcher = get_full_dispatcher(reactor, authenticator, log,
get_service_configs(config),
kz_client, store, supervisor,
cassandra_cluster)
# Setup scheduler service after starting
scheduler = setup_scheduler(parent, dispatcher, store, kz_client)
health_checker.checks['scheduler'] = scheduler.health_check
otter.scheduler = scheduler
# Give dispatcher to Otter REST object
otter.dispatcher = dispatcher
# Set the client after starting
# NOTE: There is small amount of time when the start is
# not finished and the kz_client is not set in which case
# policy execution and group delete will fail
store.kz_client = kz_client
# Setup kazoo to stop when shutting down
parent.addService(FunctionalService(
stop=partial(call_after_supervisor,
kz_client.stop, supervisor)))
setup_converger(
parent, kz_client, dispatcher,
config_value('converger.interval') or 10,
config_value('converger.build_timeout') or 3600,
config_value('converger.limited_retry_iterations') or 10,
config_value('converger.step_limits') or {})
def test_cloudfeeds_optional(self):
"""
Does not return cloud feeds service if the config is not there
"""
del self.config['cloudfeeds']
self.assertNotIn(ServiceType.CLOUD_FEEDS,
get_service_configs(self.config))
def test_cloudfeeds_setup(self):
"""
Cloud feeds observer is setup if it is there in config
"""
self.addCleanup(set_fanout, None)
self.assertEqual(get_fanout(), None)
conf = deepcopy(test_config)
conf['cloudfeeds'] = {'service': 'cloudFeeds', 'tenant_id': 'tid',
'url': 'url'}
makeService(conf)
serv_confs = get_service_configs(conf)
serv_confs[ServiceType.CLOUD_FEEDS] = {'url': 'url'}
self.assertEqual(len(get_fanout().subobservers), 1)
cf_observer = get_fanout().subobservers[0]
self.assertEqual(
cf_observer,
CloudFeedsObserver(
reactor=self.reactor,
authenticator=matches(IsInstance(CachingAuthenticator)),
tenant_id='tid',
region='ord',
service_configs=serv_confs))
# single tenant authenticator is created
authenticator = cf_observer.authenticator
self.assertIsInstance(
authenticator._authenticator._authenticator._authenticator,
SingleTenantAuthenticator)
def test_metrics_optional(self):
"""
Does not return metrics service if the config is not there
"""
del self.config['metrics']
self.assertNotIn(ServiceType.CLOUD_METRICS_INGEST,
get_service_configs(self.config))
def test_metrics_optional(self):
"""
Does not return metrics service if the config is not there
"""
del self.config['metrics']
self.assertNotIn(ServiceType.CLOUD_METRICS_INGEST,
get_service_configs(self.config))
def test_takes_from_config(self):
"""
Returns mapping based on info from config
"""
self.assertEqual(
get_service_configs(self.config), {
ServiceType.CLOUD_SERVERS: {
'name': 'nova',
'region': 'DFW',
},
ServiceType.CLOUD_LOAD_BALANCERS: {
'name': 'clb',
'region': 'DFW',
},
ServiceType.CLOUD_ORCHESTRATION: {
'name': 'orch',
'region': 'DFW',
},
ServiceType.RACKCONNECT_V3: {
'name': 'rc',
'region': 'DFW',
},
ServiceType.CLOUD_METRICS_INGEST: {
'name': 'm',
'region': 'IAD',
},
ServiceType.CLOUD_FEEDS: {
'name': 'cf',
'region': 'DFW',
'url': 'url'
}
})
def test_takes_from_config(self):
"""
Returns mapping based on info from config
"""
self.assertEqual(
get_service_configs(self.config),
{
ServiceType.CLOUD_SERVERS: {
'name': 'nova',
'region': 'DFW',
},
ServiceType.CLOUD_LOAD_BALANCERS: {
'name': 'clb',
'region': 'DFW',
},
ServiceType.CLOUD_ORCHESTRATION: {
'name': 'orch',
'region': 'DFW',
},
ServiceType.RACKCONNECT_V3: {
'name': 'rc',
'region': 'DFW',
},
ServiceType.CLOUD_METRICS_INGEST: {
'name': 'm',
'region': 'IAD',
},
ServiceType.CLOUD_FEEDS: {'url': 'cf_url'},
ServiceType.CLOUD_FEEDS_CAP: {"url": "cap_url"},
})
def test_cloudfeeds_setup(self):
"""
Cloud feeds observer is setup if it is there in config
"""
self.addCleanup(set_fanout, None)
self.assertEqual(get_fanout(), None)
conf = deepcopy(test_config)
conf['cloudfeeds'] = {
'service': 'cloudFeeds',
'tenant_id': 'tid',
'url': 'url'
}
makeService(conf)
serv_confs = get_service_configs(conf)
serv_confs[ServiceType.CLOUD_FEEDS] = {'url': 'url'}
self.assertEqual(len(get_fanout().subobservers), 1)
cf_observer = get_fanout().subobservers[0]
self.assertEqual(
cf_observer,
CloudFeedsObserver(reactor=self.reactor,
authenticator=matches(
IsInstance(CachingAuthenticator)),
tenant_id='tid',
region='ord',
service_configs=serv_confs))
# single tenant authenticator is created
authenticator = cf_observer.authenticator
self.assertIsInstance(
authenticator._authenticator._authenticator._authenticator,
SingleTenantAuthenticator)
def test_cloudfeeds_optional(self):
"""
Does not return cloud feeds service if the config is not there
"""
del self.config['cloudfeeds']
self.assertNotIn(ServiceType.CLOUD_FEEDS,
get_service_configs(self.config))
def set_desired_to_actual(groups, reactor, store, cass_client, authenticator,
conf):
dispatcher = get_full_dispatcher(
reactor, authenticator, mock_log(), get_service_configs(conf),
"kzclient", store, "supervisor", cass_client)
return gatherResults(
map(partial(set_desired_to_actual_group, dispatcher, cass_client),
groups))
def collect_metrics(reactor, config, log, client=None, authenticator=None,
_print=False):
"""
Start collecting the metrics
:param reactor: Twisted reactor
:param dict config: Configuration got from file containing all info
needed to collect metrics
:param :class:`silverberg.client.CQLClient` client:
Optional cassandra client. A new client will be created
if this is not given and disconnected before returing
:param :class:`otter.auth.IAuthenticator` authenticator:
Optional authenticator. A new authenticator will be created
if this is not given
:param bool _print: Should debug messages be printed to stdout?
:return: :class:`Deferred` fired with ``list`` of `GroupMetrics`
"""
_client = client or connect_cass_servers(reactor, config['cassandra'])
authenticator = authenticator or generate_authenticator(reactor,
config['identity'])
store = CassScalingGroupCollection(_client, reactor, 1000)
dispatcher = get_dispatcher(reactor, authenticator, log,
get_service_configs(config), store)
# calculate metrics on launch_server and non-paused groups
groups = yield perform(dispatcher, Effect(GetAllValidGroups()))
groups = [
g for g in groups
if json.loads(g["launch_config"]).get("type") == "launch_server" and
(not g.get("paused", False))]
tenanted_groups = groupby(lambda g: g["tenantId"], groups)
group_metrics = yield get_all_metrics(
dispatcher, tenanted_groups, log, _print=_print)
# Add to cloud metrics
metr_conf = config.get("metrics", None)
if metr_conf is not None:
eff = add_to_cloud_metrics(
metr_conf['ttl'], config['region'], group_metrics,
len(tenanted_groups), config, log, _print)
eff = Effect(TenantScope(eff, metr_conf['tenant_id']))
yield perform(dispatcher, eff)
log.msg('added to cloud metrics')
if _print:
print('added to cloud metrics')
if _print:
group_metrics.sort(key=lambda g: abs(g.desired - g.actual),
reverse=True)
print('groups sorted as per divergence')
print('\n'.join(map(str, group_metrics)))
# Disconnect only if we created the client
if not client:
yield _client.disconnect()
defer.returnValue(group_metrics)
def test_cloudfeeds_optional(self):
"""
Does not return cloud feeds services if the config is not there
"""
del self.config['cloudfeeds']
del self.config['terminator']
confs = get_service_configs(self.config)
self.assertNotIn(ServiceType.CLOUD_FEEDS, confs)
self.assertNotIn(ServiceType.CLOUD_FEEDS_CAP, confs)
def groups_steps(groups, reactor, store, cass_client, authenticator, conf):
"""
Return [(group, steps)] list
"""
eff = parallel(map(group_steps, groups))
disp = get_full_dispatcher(
reactor, authenticator, mock_log(), get_service_configs(conf),
"kzclient", store, "supervisor", cass_client)
return perform(disp, eff).addCallback(lambda steps: zip(groups, steps))
def groups_steps(groups, reactor, store, cass_client, authenticator, conf):
"""
Return [(group, steps)] list
"""
eff = parallel(map(group_steps, groups))
disp = get_full_dispatcher(reactor, authenticator, mock_log(),
get_service_configs(conf), "kzclient", store,
"supervisor", cass_client)
return perform(disp, eff).addCallback(lambda steps: zip(groups, steps))
def collect_metrics(reactor, config, log, client=None, authenticator=None,
_print=False):
"""
Start collecting the metrics
:param reactor: Twisted reactor
:param dict config: Configuration got from file containing all info
needed to collect metrics
:param :class:`silverberg.client.CQLClient` client:
Optional cassandra client. A new client will be created
if this is not given and disconnected before returing
:param :class:`otter.auth.IAuthenticator` authenticator:
Optional authenticator. A new authenticator will be created
if this is not given
:param bool _print: Should debug messages be printed to stdout?
:return: :class:`Deferred` fired with ``list`` of `GroupMetrics`
"""
convergence_tids = config.get('convergence-tenants', [])
_client = client or connect_cass_servers(reactor, config['cassandra'])
authenticator = authenticator or generate_authenticator(reactor,
config['identity'])
store = CassScalingGroupCollection(_client, reactor, 1000)
dispatcher = get_dispatcher(reactor, authenticator, log,
get_service_configs(config), store)
# calculate metrics
fpath = get_in(["metrics", "last_tenant_fpath"], config,
default="last_tenant.txt")
tenanted_groups = yield perform(
dispatcher,
get_todays_scaling_groups(convergence_tids, fpath))
group_metrics = yield get_all_metrics(
dispatcher, tenanted_groups, log, _print=_print)
# Add to cloud metrics
metr_conf = config.get("metrics", None)
if metr_conf is not None:
eff = add_to_cloud_metrics(
metr_conf['ttl'], config['region'], group_metrics,
len(tenanted_groups), log, _print)
eff = Effect(TenantScope(eff, metr_conf['tenant_id']))
yield perform(dispatcher, eff)
log.msg('added to cloud metrics')
if _print:
print('added to cloud metrics')
if _print:
group_metrics.sort(key=lambda g: abs(g.desired - g.actual),
reverse=True)
print('groups sorted as per divergence', *group_metrics, sep='\n')
# Disconnect only if we created the client
if not client:
yield _client.disconnect()
defer.returnValue(group_metrics)
def makeService(config):
"""
Set up the otter-api service.
"""
config = dict(config)
set_config_data(config)
parent = MultiService()
region = config_value('region')
seed_endpoints = [
clientFromString(reactor, str(host))
for host in config_value('cassandra.seed_hosts')]
cassandra_cluster = LoggingCQLClient(
TimingOutCQLClient(
reactor,
RoundRobinCassandraCluster(
seed_endpoints,
config_value('cassandra.keyspace'),
disconnect_on_cancel=True),
config_value('cassandra.timeout') or 30),
log.bind(system='otter.silverberg'))
store = CassScalingGroupCollection(
cassandra_cluster, reactor, config_value('limits.absolute.maxGroups'))
admin_store = CassAdmin(cassandra_cluster)
bobby_url = config_value('bobby_url')
if bobby_url is not None:
set_bobby(BobbyClient(bobby_url))
service_configs = get_service_configs(config)
authenticator = generate_authenticator(reactor, config['identity'])
supervisor = SupervisorService(authenticator, region, coiterate,
service_configs)
supervisor.setServiceParent(parent)
set_supervisor(supervisor)
health_checker = HealthChecker(reactor, {
'store': getattr(store, 'health_check', None),
'kazoo': store.kazoo_health_check,
'supervisor': supervisor.health_check
})
# Setup cassandra cluster to disconnect when otter shuts down
if 'cassandra_cluster' in locals():
parent.addService(FunctionalService(stop=partial(
call_after_supervisor, cassandra_cluster.disconnect, supervisor)))
otter = Otter(store, region, health_checker.health_check)
site = Site(otter.app.resource())
site.displayTracebacks = False
api_service = service(str(config_value('port')), site)
api_service.setServiceParent(parent)
# Setup admin service
admin_port = config_value('admin')
if admin_port:
admin = OtterAdmin(admin_store)
admin_site = Site(admin.app.resource())
admin_site.displayTracebacks = False
admin_service = service(str(admin_port), admin_site)
admin_service.setServiceParent(parent)
# setup cloud feed
cf_conf = config.get('cloudfeeds', None)
if cf_conf is not None:
id_conf = deepcopy(config['identity'])
id_conf['strategy'] = 'single_tenant'
add_to_fanout(CloudFeedsObserver(
reactor=reactor,
authenticator=generate_authenticator(reactor, id_conf),
tenant_id=cf_conf['tenant_id'],
region=region,
service_configs=service_configs))
# Setup Kazoo client
if config_value('zookeeper'):
threads = config_value('zookeeper.threads') or 10
disable_logs = config_value('zookeeper.no_logs')
threadpool = ThreadPool(maxthreads=threads)
sync_kz_client = KazooClient(
hosts=config_value('zookeeper.hosts'),
# Keep trying to connect until the end of time with
# max interval of 10 minutes
connection_retry=dict(max_tries=-1, max_delay=600),
logger=None if disable_logs else TxLogger(log.bind(system='kazoo'))
)
kz_client = TxKazooClient(reactor, threadpool, sync_kz_client)
# Don't timeout. Keep trying to connect forever
d = kz_client.start(timeout=None)
def on_client_ready(_):
dispatcher = get_full_dispatcher(reactor, authenticator, log,
get_service_configs(config),
kz_client, store, supervisor,
cassandra_cluster)
# Setup scheduler service after starting
scheduler = setup_scheduler(parent, dispatcher, store, kz_client)
health_checker.checks['scheduler'] = scheduler.health_check
otter.scheduler = scheduler
# Give dispatcher to Otter REST object
otter.dispatcher = dispatcher
# Set the client after starting
# NOTE: There is small amount of time when the start is
# not finished and the kz_client is not set in which case
# policy execution and group delete will fail
store.kz_client = kz_client
# Setup kazoo to stop when shutting down
parent.addService(FunctionalService(
stop=partial(call_after_supervisor,
kz_client.stop, supervisor)))
setup_converger(
parent, kz_client, dispatcher,
config_value('converger.interval') or 10,
config_value('converger.build_timeout') or 3600,
config_value('converger.limited_retry_iterations') or 10,
config_value('converger.step_limits') or {})
d.addCallback(on_client_ready)
d.addErrback(log.err, 'Could not start TxKazooClient')
return parent
