class KafkaCluster(object):
# conf dict structure with defaults:
# commented-out fields are not defaults but show what is available.
default_conf = {
'version': '2.8.0', # Apache Kafka version
'cp_version': '6.1.0', # Confluent Platform version (for SR)
'broker_cnt': 3,
'sasl_mechanism': '', # GSSAPI, PLAIN, SCRAM-.., ...
'realm_cnt': 1,
'krb_renew_lifetime': 30,
'krb_ticket_lifetime': 120,
# KRaft (No zookeeper). Requires AK >=2.8
'kraft': False,
# SASL PLAIN/SCRAM
'sasl_users': 'testuser=testpass',
# With SSL
'with_ssl': False,
# With SchemaRegistry
'with_sr': False,
# Debug trivup
'debug': False,
# Cleanup
'cleanup': True,
'oidc': False,
# Additional broker server.properties configuration
# 'broker_conf': ['connections.max.idle.ms=1234', ..]
}
def __init__(self, **kwargs):
""" Create and start a KafkaCluster.
See default_conf above for parameters. """
super(KafkaCluster, self).__init__()
conf = kwargs
self.conf = deepcopy(self.default_conf)
if conf is not None:
self.conf.update(conf)
self.version = self.conf.get('version')
self.version_num = [int(x) for x in self.version.split('.')][:3]
self.kraft = self.conf.get('kraft')
# Create trivup Cluster
self.cluster = Cluster(self.__class__.__name__,
os.environ.get(
'TRIVUP_ROOT',
'tmp-%s' % self.__class__.__name__),
debug=bool(self.conf.get('debug', False)),
cleanup=bool(self.conf.get('cleanup', True)))
self._client_conf = dict()
self.env = dict()
self.sasl_mechanism = self.conf.get('sasl_mechanism')
# Add OIDC server app
if bool(self.conf.get('oidc', False)):
self.oidc = OauthbearerOIDCApp(self.cluster)
if not self.sasl_mechanism:
self.sasl_mechanism = 'OAUTHBEARER'
elif self.sasl_mechanism.upper() != 'OAUTHBEARER':
raise RuntimeError(
f"OIDC requires sasl.mechanism OAUTHBEARER, not '{self.sasl_mechanism}'"
)
# Generate SSL certs if enabled
if bool(self.conf.get('with_ssl')):
self.ssl = SslApp(self.cluster, self.conf)
else:
self.ssl = None
# Map mechanism and SSL to security protocol
self.security_protocol = {
(True, True): 'SASL_SSL',
(True, False): 'SASL_PLAINTEXT',
(False, True): 'SSL',
(False, False): 'PLAINTEXT'
}[(bool(self.sasl_mechanism), bool(self.ssl is not None))]
if not self.kraft:
# Create single ZK for the cluster (don't start yet)
self.zk = ZookeeperApp(self.cluster)
else:
self.zk = None
# Allocate (but don't use) a dummy appid so that the brokers get
# the same appid/nodeid for both KRaft and ZK modes.
Allocator(self.cluster).next()
# Broker configuration
broker_cnt = int(self.conf.get('broker_cnt'))
self.broker_conf = {
'replication_factor': min(3, broker_cnt),
'num_partitions': 4,
'version': self.version,
'sasl_mechanisms': self.sasl_mechanism,
'sasl_users': self.conf.get('sasl_users'),
'conf': self.conf.get('broker_conf', []),
'kafka_path': self.conf.get('kafka_path', None)
}
# Start Kerberos KDCs if GSSAPI (Kerberos) is configured
if self.sasl_mechanism == 'GSSAPI':
self._setup_kerberos()
self.broker_conf['realm'] = self.broker_realm
self.broker_conf['listener_host'] = 'localhost'
# Create brokers (don't start yet)
self.brokers = dict()
for n in range(0, broker_cnt):
bconf = copy.deepcopy(self.broker_conf)
if self.version_num >= [2, 4, 0]:
# Configure rack & replica selector if broker supports
# fetch-from-follower
bconf.update({
'conf': [
'broker.rack=RACK${appid}',
'replica.selector.class=org.apache.kafka.common.replica.RackAwareReplicaSelector'
]
}) # noqa: E501
broker = KafkaBrokerApp(self.cluster, bconf)
self.brokers[broker.appid] = broker
# Generate bootstrap servers list
all_listeners = (','.join(
self.cluster.get_all('advertised_listeners', '',
KafkaBrokerApp))).split(',')
self.bootstrap_servers = ','.join(
[x for x in all_listeners if x.startswith(self.security_protocol)])
assert len(self.bootstrap_servers) >= broker_cnt, \
"{} < {} expected bootstrap servers".format(
len(self.bootstrap_servers), broker_cnt)
# Create SchemaRegistry if enabled
if bool(self.conf.get('with_sr', False)):
self.sr = SchemaRegistryApp(
self.cluster, {'version': self.conf.get('cp_version')})
self.env['SR_URL'] = self.sr.get('url')
# Create librdkafka client configuration
self._setup_client_conf()
# Deploy cluster
self.cluster.deploy()
# Start cluster
self.start()
def __del__(self):
""" Destructor: forcibly stop the cluster """
self.stop(force=True)
def _setup_env(self):
""" Set up convenience envs """
self.env['KAFKA_PATH'] = self.cluster.find_app(KafkaBrokerApp).get(
'destdir') # noqa: E501
if not self.kraft:
self.env['ZK_ADDRESS'] = self.zk.get('address')
self.env['BROKERS'] = self.bootstrap_servers
self.env['KAFKA_VERSION'] = self.version
self.env['TRIVUP_ROOT'] = self.cluster.instance_path()
# Add each broker pid as an env so they can be killed indivdidually.
for b in self.cluster.find_apps(KafkaBrokerApp, 'started'):
self.env['BROKER_PID_%d' % b.appid] = str(b.proc.pid)
def _setup_client_conf(self):
""" Set up librdkafka client configuration """
self._client_conf['bootstrap.servers'] = self.bootstrap_servers
self._client_conf['broker.address.family'] = 'v4'
if self.security_protocol != 'PLAINTEXT':
self._client_conf['security.protocol'] = self.security_protocol
broker_version = self.conf.get('version')
brver = broker_version.split('.')
if brver[0] == 0 and brver[1] < 10:
self._client_conf['broker.version.fallback'] = broker_version
self._client_conf['api.version.request'] = 'false'
# Client SASL configuration
if self.sasl_mechanism:
self._client_conf['sasl.mechanism'] = self.sasl_mechanism
if self.sasl_mechanism == 'PLAIN' or \
self.sasl_mechanism.find('SCRAM') != -1:
# Use first user as SASL user/pass
for up in self.conf.get('sasl_users', '').split(','):
u, p = up.split('=')
self._client_conf['sasl.username'] = u
self._client_conf['sasl.password'] = p
break
elif self.sasl_mechanism == 'OAUTHBEARER':
if self.oidc is not None:
self._client_conf['sasl.oauthbearer.method'] = 'OIDC'
self._client_conf[
'sasl.oauthbearer.token.endpoint.url'] = self.oidc.get(
'valid_url') # noqa: E501
self._client_conf['sasl.oauthbearer.client.id'] = '123'
self._client_conf['sasl.oauthbearer.client.secret'] = 'abc'
self._client_conf['sasl.oauthbearer.scope'] = 'test'
self._client_conf['sasl.oauthbearer.extensions'] = \
'ExtensionworkloadIdentity=develC348S,Extensioncluster=lkc123'
else:
self._client_conf[
'enable.sasl.oauthbearer.unsecure.jwt'] = True # noqa: E501
self._client_conf['sasl.oauthbearer.config'] = \
'scope=requiredScope principal=admin'
# Client SSL configuration
if self.ssl is not None:
key = self.ssl.create_cert('client')
self._client_conf['ssl.ca.location'] = self.ssl.ca['pem']
self._client_conf['ssl.certificate.location'] = key['pub']['pem']
self._client_conf['ssl.key.location'] = key['priv']['pem']
self._client_conf['ssl.key.password'] = key['password']
# Add envs pointing out locations of the generated certs
for k, v in self.ssl.ca.items():
self.env['SSL_ca_{}'.format(k)] = v
# Set envs for all generated keys so tests can find them.
for k, v in key.items():
if type(v) is dict:
for k2, v2 in v.items():
# E.g. "SSL_priv_der=path/to/librdkafka-priv.der"
self.env['SSL_{}_{}'.format(k, k2)] = v2
else:
self.env['SSL_{}'.format(k)] = v
def _setup_kerberos(self):
""" Set up Kerberos KDCs """
# Create KDCs for each realm.
# First realm will be the default / broker realm.
#
realm_cnt = int(self.conf.get('realm_cnt', 1))
# No point in having more than two realms
assert realm_cnt > 0 and realm_cnt < 3
realms = ['REALM{}.TRIVUP'.format(x + 1) for x in range(0, realm_cnt)]
# Pre-Allocate ports for the KDCs so they can reference eachother
# in the krb5.conf configuration.
kdc_ports = {
x: TcpPortAllocator(self.cluster).next("dummy")
for x in realms
}
# Set up realm=kdc:port cross-realm mappings
cross_realms = ",".join([
"{}={}:{}".format(x,
self.cluster.get_node().name, kdc_ports[x])
for x in realms
])
kdcs = dict()
for realm in realms:
kdc = KerberosKdcApp(self.cluster,
realm,
conf={
'port':
kdc_ports[realm],
'cross_realms':
cross_realms,
'renew_lifetime':
int(self.conf.get('krb_renew_lifetime')),
'ticket_lifetime':
int(self.conf.get('krb_ticket_lifetime'))
})
# Kerberos needs to be started prior to Kafka so that principals
# and keytabs are available at the time of Kafka config generation.
kdc.start()
kdcs[realm] = kdc
self.broker_realm = realms[0]
self.client_realm = realms[-1]
self.broker_kdc = kdcs[self.broker_realm]
self.client_kdc = kdcs[self.client_realm]
# Add cross-realm TGTs
if realm_cnt > 1:
KerberosKdcApp.add_cross_realm_tgts(kdcs)
# Add client envs and configuration
self.env['KRB5CCNAME'] = self.client_kdc.mkpath('krb5cc')
self.env['KRB5_CONFIG'] = self.client_kdc.conf['krb5_conf']
self.env['KRB5_KDC_PROFILE'] = self.client_kdc.conf['kdc_conf']
principal, keytab = self.client_kdc.add_principal('admin')
self._client_conf['sasl.kerberos.keytab'] = keytab
self._client_conf['sasl.kerberos.principal'] = principal.split('@')[0]
# Refresh ticket 60s before renew timeout.
self._client_conf['sasl.kerberos.min.time.before.relogin'] = \
max(1, int(self.conf.get('krb_renew_lifetime')) - 60) * 1000
def start(self, timeout=0):
""" Start cluster """
self.cluster.start()
# Set up additional convenience envs
self._setup_env()
if timeout > 0:
self.wait_operational(timeout)
def stop(self, cleanup=True, keeptypes=['log'], force=False, timeout=0):
""" Stop cluster and clean up """
self.cluster.stop(force=True)
if timeout > 0:
self.cluster.wait_stopped(timeout)
if cleanup:
self.cluster.cleanup(keeptypes)
def stopped(self):
""" Returns True when all components of the cluster are stopped """
return len([x for x in self.cluster.apps
if x.status() == 'stopped']) == len(self.cluster.apps)
def wait_operational(self, timeout=60):
""" Wait for cluster to go operational """
if not self.cluster.wait_operational(timeout):
self.stop(force=True)
raise Exception(
"Cluster {} did not go operational, see logs in {}/{}".format(
self.cluster.name, self.cluster.root_path,
self.cluster.instance))
def stop_broker(self, broker_id):
""" Stop single broker """
broker = self.brokers.get(broker_id, None)
if broker is None:
raise LookupError("Unknown broker id {}".format(broker_id))
broker.stop()
def start_broker(self, broker_id):
""" Start single broker """
broker = self.brokers.get(broker_id, None)
if broker is None:
raise LookupError("Unknown broker id {}".format(broker_id))
broker.start()
def interactive(self, cmd=None):
""" Execute an interactive shell that has all the
environment variables set. """
if cmd is None:
print('# Interactive mode')
print("# - Waiting for cluster to go operational in {}/{}".format(
self.cluster.root_path, self.cluster.instance))
kc.wait_operational()
env = self.env.copy()
# Avoids 'dumb' terminal mode
env['TERM'] = os.environ.get('TERM', 'vt100')
# Write librdkafka client configuration to file.
cf_path = self.cluster.mkpath('rdkafka.conf', in_instance=True)
self.write_client_conf(cf_path)
env['RDKAFKA_TEST_CONF'] = cf_path
print("# - Client configuration in {}".format(cf_path))
print("# - Connect to cluster with bootstrap.servers {}".format(
self.bootstrap_servers))
# Prefix the standard prompt with cluster info.
if cmd is None:
pfx = '[TRIVUP:{}@{}] '.format(self.cluster.name, self.version)
fullcmd = 'bash --rcfile <(cat ~/.bashrc; echo \'PS1="{}$PS1"\') -i'.format(
pfx) # noqa: E501
print("# - You're now in an interactive sub-shell, type 'exit' "
"to stop the cluster and exit back to your shell.\n")
retcode = subprocess.call(fullcmd,
env=env,
shell=True,
executable='/bin/bash')
else:
fullcmd = cmd
print("# - Executing: {}".format(fullcmd))
retcode = subprocess.call(fullcmd, env=env, shell=True)
if retcode != 0:
print("# - Shell exited with returncode {}: {}".format(
retcode, fullcmd))
def client_conf(self):
""" Get a dict copy of the client configuration """
return deepcopy(self._client_conf)
def write_client_conf(self, path, additional_blob=None):
""" Write client configuration (librdkafka) to @param path """
with open(path, "w") as f:
for k, v in self._client_conf.items():
f.write(str('%s=%s\n' % (k, v)))
if additional_blob is not None:
f.write(str('#\n# Additional configuration:'))
f.write(str(additional_blob))
def test_kerberos_cross_realm():
""" Test Kerberos cross-realm trusts """
topic = "test"
cluster = Cluster('KafkaCluster',
root_path=os.environ.get('TRIVUP_ROOT', 'tmp'),
debug=True)
ZookeeperApp(cluster)
#
# Create KDCs for each realm.
# First realm will be the default / broker realm.
#
realm_cnt = 2
realms = ["REALM{}.COM".format(x + 1) for x in range(0, realm_cnt)]
# Pre-Allocate ports for the KDCs so they can reference eachother
# in the krb5.conf configuration.
kdc_ports = {x: TcpPortAllocator(cluster).next("dummy") for x in realms}
# Set up realm=kdc:port cross-realm mappings
cross_realms = ",".join(["{}={}:{}".format(x, cluster.get_node().name, kdc_ports[x]) for x in realms])
kdcs = dict()
for realm in realms:
kdc = KerberosKdcApp(cluster, realm,
conf={'port': kdc_ports[realm],
'cross_realms': cross_realms,
'renew_lifetime': '30',
'ticket_lifetime': '120'})
kdc.start()
kdcs[realm] = kdc
broker_realm = realms[0]
client_realm = realms[1]
broker_kdc = kdcs[broker_realm]
client_kdc = kdcs[client_realm]
# Create broker_cnt brokers
broker_cnt = 4
brokerconf = {'replication_factor': min(3, int(broker_cnt)),
'num_partitions': broker_cnt * 2,
'version': '2.2.0',
'sasl_mechanisms': 'GSSAPI',
'realm': broker_realm,
'conf': ['connections.max.idle.ms=60000']}
brokers = dict()
for n in range(0, broker_cnt):
broker = KafkaBrokerApp(cluster, brokerconf)
brokers[broker.appid] = broker
# Get bootstrap server list
security_protocol = 'SASL_PLAINTEXT'
all_listeners = (','.join(cluster.get_all(
'listeners', '', KafkaBrokerApp))).split(',')
bootstrap_servers = ','.join([x for x in all_listeners
if x.startswith(security_protocol)])
assert len(bootstrap_servers) > 0, "no bootstrap servers"
print("## Deploying cluster")
cluster.deploy()
print("## Starting cluster")
cluster.start(timeout=30)
# Add cross-realm TGTs
for realm in realms:
for crealm in [x for x in realms if x != realm]:
kdcs[realm].execute('kadmin.local -d "{}" -q "addprinc -requires_preauth -pw password krbtgt/{}@{}"'.format(kdcs[realm].conf.get('dbpath'), crealm, realm)).wait()
kdcs[realm].execute('kadmin.local -d "{}" -q "addprinc -requires_preauth -pw password krbtgt/{}@{}"'.format(kdcs[realm].conf.get('dbpath'), realm, crealm)).wait()
# Create client base configuration
client_config = {
'bootstrap.servers': bootstrap_servers,
'enable.sparse.connections': False,
'broker.address.family': 'v4',
'sasl.mechanisms': 'GSSAPI',
'security.protocol': security_protocol,
'debug': 'broker,security'
}
os.environ['KRB5CCNAME'] = client_kdc.mkpath('krb5cc')
os.environ['KRB5_CONFIG'] = client_kdc.conf['krb5_conf']
os.environ['KRB5_KDC_PROFILE'] = client_kdc.conf['kdc_conf']
principal,keytab = client_kdc.add_principal("admin")
client_config['sasl.kerberos.keytab'] = keytab
client_config['sasl.kerberos.principal'] = principal.split('@')[0]
client_config['sasl.kerberos.min.time.before.relogin'] = 120*1000*3
print(client_config)
print("bootstraps: {}".format(client_config['bootstrap.servers']))
p = Producer(client_config)
time.sleep(10)
for n in range(1, 100):
p.produce(topic, "msg #{}".format(n))
p.poll(1.0)
p.flush(1.0)
print("####### {} messages remaining\n\n\n".format(len(p)))
start = time.time()
end = start + (90*60)
until = start + (12*60)
while time.time() < end:
now = time.time()
if until < now:
print("### Producing 2 messages")
for n in range(1, 2):
p.produce(topic, "msg #{}".format(n))
until = now + (12*60)
p.poll(1.0)
del p
cluster.stop()
