class Profiler:
DEBUG_PORTS = {
'fts': 8094,
'index': 9102,
'goxdcr': 9998,
'kv': 9998, # will be deprecated in future
'n1ql': 8093,
'eventing': 8096,
'projector': 9999
}
ENDPOINTS = {
'cpu': 'http://127.0.0.1:{}/debug/pprof/profile',
'heap': 'http://127.0.0.1:{}/debug/pprof/heap',
'goroutine': 'http://127.0.0.1:{}/debug/pprof/goroutine?debug=2',
}
def __init__(self, cluster_spec: ClusterSpec, test_config: TestConfig):
self.test_config = test_config
self.rest = RestHelper(cluster_spec)
self.master_node = next(cluster_spec.masters)
self.ssh_username, self.ssh_password = cluster_spec.ssh_credentials
self.cluster_spec = cluster_spec
self.profiling_settings = test_config.profiling_settings
self.linux_perf_path = '/opt/couchbase/var/lib/couchbase/logs/'
def new_tunnel(self, host: str, port: int) -> SSHTunnelForwarder:
return SSHTunnelForwarder(
ssh_address_or_host=host,
ssh_username=self.ssh_username,
ssh_password=self.ssh_password,
remote_bind_address=('127.0.0.1', port),
)
def save(self, host: str, service: str, profile: str, content: bytes):
fname = '{}_{}_{}_{}.pprof'.format(host, service, profile, uhex()[:6])
logger.info('Collected {} '.format(fname))
with open(fname, 'wb') as fh:
fh.write(content)
def linux_perf_profile(self, host: str, fname: str, path: str):
client = paramiko.SSHClient()
client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
try:
client.connect(hostname=host,
username=self.ssh_username,
password=self.ssh_password)
except Exception:
logger.info(
'Cannot connect to the "{}" via SSH Server'.format(host))
exit()
logger.info('Capturing linux profile using linux perf record ')
cmd = 'perf record -a -F {} -g --call-graph {} ' \
'-p $(pgrep memcached) -o {}{} ' \
'-- sleep {}'.format(self.profiling_settings.linux_perf_frequency,
self.profiling_settings.linux_perf_callgraph,
path,
fname,
self.profiling_settings.linux_perf_profile_duration)
stdin, stdout, stderr = client.exec_command(cmd)
exit_status = stdout.channel.recv_exit_status()
if exit_status == 0:
logger.info(
"linux perf record: linux perf profile capture completed")
else:
logger.info("perf record failed , exit_status : ", exit_status)
client.close()
def profile(self, host: str, service: str, profile: str):
logger.info('Collecting {} profile on {}'.format(profile, host))
endpoint = self.ENDPOINTS[profile]
port = self.DEBUG_PORTS[service]
if self.profiling_settings.linux_perf_profile_flag:
logger.info('Collecting {} profile on {} using linux perf '
'reccord'.format(profile, host))
fname = 'linux_{}_{}_{}_perf.data'.format(host, profile,
uhex()[:4])
self.linux_perf_profile(host=host,
fname=fname,
path=self.linux_perf_path)
else:
logger.info('Collecting {} profile on {}'.format(profile, host))
with self.new_tunnel(host, port) as tunnel:
url = endpoint.format(tunnel.local_bind_port)
response = requests.get(url=url, auth=self.rest.auth)
self.save(host, service, profile, response.content)
def timer(self, **kwargs):
timer = Timer(
function=self.profile,
interval=self.test_config.profiling_settings.interval,
num_runs=self.test_config.profiling_settings.num_profiles,
kwargs=kwargs,
)
timer.start()
def schedule(self):
for service in self.test_config.profiling_settings.services:
logger.info(
'Scheduling profiling of "{}" services'.format(service))
if service == 'projector':
active_nodes_by_role = self.rest.get_active_nodes_by_role(
self.master_node, role='kv')
else:
active_nodes_by_role = self.rest.get_active_nodes_by_role(
self.master_node, role=service)
for server in active_nodes_by_role:
for profile in self.test_config.profiling_settings.profiles:
self.timer(host=server, service=service, profile=profile)
class PerfTest:
COLLECTORS = {}
ROOT_CERTIFICATE = 'root.pem'
def __init__(self, cluster_spec: ClusterSpec, test_config: TestConfig,
verbose: bool):
self.cluster_spec = cluster_spec
self.test_config = test_config
self.target_iterator = TargetIterator(cluster_spec, test_config)
self.cluster = ClusterManager(cluster_spec, test_config)
self.memcached = MemcachedHelper(test_config)
self.monitor = Monitor(cluster_spec, test_config, verbose)
self.rest = RestHelper(cluster_spec)
self.remote = RemoteHelper(cluster_spec, verbose)
self.profiler = Profiler(cluster_spec, test_config)
self.master_node = next(cluster_spec.masters)
self.build = self.rest.get_version(self.master_node)
self.metrics = MetricHelper(self)
self.reporter = ShowFastReporter(cluster_spec, test_config, self.build)
self.cbmonitor_snapshots = []
self.cbmonitor_clusters = []
if self.test_config.test_case.use_workers:
self.worker_manager = WorkerManager(cluster_spec, test_config,
verbose)
def __enter__(self):
return self
def __exit__(self, exc_type, exc_val, exc_tb):
failure = self.debug()
self.tear_down()
if exc_type == KeyboardInterrupt:
logger.warn('The test was interrupted')
return True
if failure:
logger.interrupt(failure)
@property
def query_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'n1ql')
@property
def index_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'index')
@property
def fts_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'fts')
@property
def analytics_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'cbas')
@property
def eventing_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'eventing')
def tear_down(self):
if self.test_config.test_case.use_workers:
self.worker_manager.download_celery_logs()
self.worker_manager.terminate()
if self.test_config.cluster.online_cores:
self.remote.enable_cpu()
if self.test_config.cluster.kernel_mem_limit:
self.collect_logs()
self.cluster.reset_memory_settings()
def collect_logs(self):
self.remote.collect_info()
for hostname in self.cluster_spec.servers:
for fname in glob.glob('{}/*.zip'.format(hostname)):
shutil.move(fname, '{}.zip'.format(hostname))
def reset_memory_settings(self):
if self.test_config.cluster.kernel_mem_limit:
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.reset_memory_settings(host_string=server)
self.monitor.wait_for_servers()
def debug(self) -> str:
failure = self.check_core_dumps()
failure = self.check_rebalance() or failure
return self.check_failover() or failure
def download_certificate(self):
cert = self.rest.get_certificate(self.master_node)
with open(self.ROOT_CERTIFICATE, 'w') as fh:
fh.write(cert)
def check_rebalance(self) -> str:
for master in self.cluster_spec.masters:
if self.rest.is_not_balanced(master):
return 'The cluster is not balanced'
def check_failover(self) -> Optional[str]:
if hasattr(self, 'rebalance_settings'):
if self.rebalance_settings.failover or \
self.rebalance_settings.graceful_failover:
return
for master in self.cluster_spec.masters:
num_failovers = self.rest.get_failover_counter(master)
if num_failovers:
return 'Failover happened {} time(s)'.format(num_failovers)
def check_core_dumps(self) -> str:
dumps_per_host = self.remote.detect_core_dumps()
core_dumps = {
host: dumps
for host, dumps in dumps_per_host.items() if dumps
}
if core_dumps:
return pretty_dict(core_dumps)
def restore(self):
logger.info('Restoring data')
self.remote.restore_data(
self.test_config.restore_settings.backup_storage,
self.test_config.restore_settings.backup_repo,
)
def restore_local(self):
logger.info('Restoring data')
local.extract_cb(filename='couchbase.rpm')
local.cbbackupmgr_restore(
master_node=self.master_node,
cluster_spec=self.cluster_spec,
threads=self.test_config.restore_settings.threads,
archive=self.test_config.restore_settings.backup_storage,
repo=self.test_config.restore_settings.backup_repo,
)
def import_data(self):
logger.info('Importing data')
for bucket in self.test_config.buckets:
self.remote.import_data(
self.test_config.restore_settings.import_file,
bucket,
)
def compact_bucket(self, wait: bool = True):
for target in self.target_iterator:
self.rest.trigger_bucket_compaction(target.node, target.bucket)
if wait:
for target in self.target_iterator:
self.monitor.monitor_task(target.node, 'bucket_compaction')
def wait_for_persistence(self):
for target in self.target_iterator:
self.monitor.monitor_disk_queues(target.node, target.bucket)
self.monitor.monitor_dcp_queues(target.node, target.bucket)
self.monitor.monitor_replica_count(target.node, target.bucket)
def wait_for_indexing(self):
if self.test_config.index_settings.statements:
for server in self.index_nodes:
self.monitor.monitor_indexing(server)
def check_num_items(self):
num_items = self.test_config.load_settings.items * (
1 + self.test_config.bucket.replica_number)
for target in self.target_iterator:
self.monitor.monitor_num_items(target.node, target.bucket,
num_items)
def reset_kv_stats(self):
master_node = next(self.cluster_spec.masters)
for bucket in self.test_config.buckets:
for server in self.rest.get_server_list(master_node, bucket):
port = self.rest.get_memcached_port(server)
self.memcached.reset_stats(server, port, bucket)
def create_indexes(self):
logger.info('Creating and building indexes')
if not self.test_config.index_settings.couchbase_fts_index_name:
for statement in self.test_config.index_settings.statements:
self.rest.exec_n1ql_statement(self.query_nodes[0], statement)
else:
self.create_fts_index_n1ql()
def create_fts_index_n1ql(self):
definition = read_json(
self.test_config.index_settings.couchbase_fts_index_configfile)
definition.update(
{'name': self.test_config.index_settings.couchbase_fts_index_name})
logger.info('Index definition: {}'.format(pretty_dict(definition)))
self.rest.create_fts_index(
self.fts_nodes[0],
self.test_config.index_settings.couchbase_fts_index_name,
definition)
self.monitor.monitor_fts_indexing_queue(
self.fts_nodes[0],
self.test_config.index_settings.couchbase_fts_index_name,
int(self.test_config.access_settings.items))
def create_functions(self):
logger.info('Creating n1ql functions')
for statement in self.test_config.n1ql_function_settings.statements:
self.rest.exec_n1ql_statement(self.query_nodes[0], statement)
def sleep(self):
access_settings = self.test_config.access_settings
logger.info('Running phase for {} seconds'.format(
access_settings.time))
time.sleep(access_settings.time)
def run_phase(self,
phase: str,
task: Callable,
settings: PhaseSettings,
target_iterator: Iterable,
timer: int = None,
wait: bool = True):
logger.info('Running {}: {}'.format(phase, pretty_dict(settings)))
self.worker_manager.run_tasks(task, settings, target_iterator, timer)
if wait:
self.worker_manager.wait_for_workers()
def load(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.load_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('load phase', task, settings, target_iterator)
def hot_load(self, task: Callable = spring_task):
settings = self.test_config.hot_load_settings
self.run_phase('hot load phase', task, settings, self.target_iterator)
def xattr_load(self,
task: Callable = spring_task,
target_iterator: Iterable = None):
if target_iterator is None:
target_iterator = self.target_iterator
settings = self.test_config.xattr_load_settings
self.run_phase('xattr phase', task, settings, target_iterator)
def access(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.access_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('access phase',
task,
settings,
target_iterator,
timer=settings.time)
def access_bg(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.access_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('background access phase',
task,
settings,
target_iterator,
timer=settings.time,
wait=False)
def report_kpi(self, *args, **kwargs):
if self.test_config.stats_settings.enabled:
self._report_kpi(*args, **kwargs)
def _report_kpi(self, *args, **kwargs):
pass
def _measure_curr_ops(self) -> int:
ops = 0
for bucket in self.test_config.buckets:
for server in self.rest.get_active_nodes_by_role(
self.master_node, "kv"):
port = self.rest.get_memcached_port(server)
stats = self.memcached.get_stats(server, port, bucket)
for stat in b'cmd_get', b'cmd_set':
ops += int(stats[stat])
return ops
class ClusterManager:
def __init__(self,
cluster_spec: ClusterSpec,
test_config: TestConfig,
verbose: bool = False):
self.cluster_spec = cluster_spec
self.test_config = test_config
self.rest = RestHelper(cluster_spec)
self.remote = RemoteHelper(cluster_spec, verbose)
self.monitor = Monitor(cluster_spec, test_config, verbose)
self.memcached = MemcachedHelper(test_config)
self.master_node = next(self.cluster_spec.masters)
self.initial_nodes = test_config.cluster.initial_nodes
def is_compatible(self, min_release: str) -> bool:
for master in self.cluster_spec.masters:
version = self.rest.get_version(master)
return version >= min_release
def set_data_path(self):
for server in self.cluster_spec.servers:
self.remote.change_owner(server, self.cluster_spec.data_path)
self.rest.set_data_path(server, self.cluster_spec.data_path)
def set_index_path(self):
for server in self.cluster_spec.servers:
self.remote.change_owner(server, self.cluster_spec.index_path)
self.rest.set_index_path(server, self.cluster_spec.index_path)
def set_analytics_path(self):
paths = []
for path in self.cluster_spec.analytics_paths:
for i in range(self.test_config.analytics_settings.num_io_devices):
io_device = '{}/dev{}'.format(path, i)
paths.append(io_device)
for server in self.cluster_spec.servers_by_role('cbas'):
for path in self.cluster_spec.analytics_paths:
self.remote.change_owner(server, path)
self.rest.set_analytics_paths(server, paths)
def rename(self):
for server in self.cluster_spec.servers:
self.rest.rename(server)
def set_auth(self):
for server in self.cluster_spec.servers:
self.rest.set_auth(server)
def set_mem_quotas(self):
for master in self.cluster_spec.masters:
self.rest.set_mem_quota(master, self.test_config.cluster.mem_quota)
self.rest.set_index_mem_quota(
master, self.test_config.cluster.index_mem_quota)
if self.test_config.cluster.fts_index_mem_quota:
self.rest.set_fts_index_mem_quota(
master, self.test_config.cluster.fts_index_mem_quota)
if self.test_config.cluster.analytics_mem_quota:
self.rest.set_analytics_mem_quota(
master, self.test_config.cluster.analytics_mem_quota)
if self.test_config.cluster.eventing_mem_quota:
self.rest.set_eventing_mem_quota(
master, self.test_config.cluster.eventing_mem_quota)
def set_query_settings(self):
logger.info('Setting query settings')
query_nodes = self.cluster_spec.servers_by_role('n1ql')
if query_nodes:
settings = self.test_config.n1ql_settings.cbq_settings
if settings:
self.rest.set_query_settings(query_nodes[0], settings)
settings = self.rest.get_query_settings(query_nodes[0])
settings = pretty_dict(settings)
logger.info('Query settings: {}'.format(settings))
def set_index_settings(self):
logger.info('Setting index settings')
index_nodes = self.cluster_spec.servers_by_role('index')
if index_nodes:
settings = self.test_config.gsi_settings.settings
if settings:
self.rest.set_index_settings(index_nodes[0], settings)
settings = self.rest.get_index_settings(index_nodes[0])
settings = pretty_dict(settings)
logger.info('Index settings: {}'.format(settings))
def set_services(self):
if not self.is_compatible(min_release='4.0.0'):
return
for master in self.cluster_spec.masters:
roles = self.cluster_spec.roles[master]
self.rest.set_services(master, roles)
def add_nodes(self):
for (_, servers), initial_nodes in zip(self.cluster_spec.clusters,
self.initial_nodes):
if initial_nodes < 2: # Single-node cluster
continue
master = servers[0]
for node in servers[1:initial_nodes]:
roles = self.cluster_spec.roles[node]
self.rest.add_node(master, node, roles)
def rebalance(self):
for (_, servers), initial_nodes in zip(self.cluster_spec.clusters,
self.initial_nodes):
master = servers[0]
known_nodes = servers[:initial_nodes]
ejected_nodes = []
self.rest.rebalance(master, known_nodes, ejected_nodes)
self.monitor.monitor_rebalance(master)
self.wait_until_healthy()
def increase_bucket_limit(self, num_buckets: int):
for master in self.cluster_spec.masters:
self.rest.increase_bucket_limit(master, num_buckets)
def flush_buckets(self):
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.flush_bucket(host=master, bucket=bucket_name)
def delete_buckets(self):
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.delete_bucket(host=master, name=bucket_name)
def create_buckets(self):
mem_quota = self.test_config.cluster.mem_quota
if self.test_config.cluster.num_buckets > 7:
self.increase_bucket_limit(self.test_config.cluster.num_buckets +
3)
if self.test_config.cluster.eventing_metadata_bucket_mem_quota:
mem_quota -= (
self.test_config.cluster.eventing_metadata_bucket_mem_quota +
self.test_config.cluster.eventing_bucket_mem_quota)
per_bucket_quota = mem_quota // self.test_config.cluster.num_buckets
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.create_bucket(
host=master,
name=bucket_name,
ram_quota=per_bucket_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.eviction_policy,
bucket_type=self.test_config.bucket.bucket_type,
conflict_resolution_type=self.test_config.bucket.
conflict_resolution_type,
compression_mode=self.test_config.bucket.compression_mode,
)
def create_eventing_buckets(self):
if not self.test_config.cluster.eventing_bucket_mem_quota:
return
per_bucket_quota = \
self.test_config.cluster.eventing_bucket_mem_quota \
// self.test_config.cluster.eventing_buckets
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.eventing_buckets:
self.rest.create_bucket(
host=master,
name=bucket_name,
ram_quota=per_bucket_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.eviction_policy,
bucket_type=self.test_config.bucket.bucket_type,
conflict_resolution_type=self.test_config.bucket.
conflict_resolution_type,
)
def create_eventing_metadata_bucket(self):
if not self.test_config.cluster.eventing_metadata_bucket_mem_quota:
return
for master in self.cluster_spec.masters:
self.rest.create_bucket(
host=master,
name=self.test_config.cluster.EVENTING_METADATA_BUCKET_NAME,
ram_quota=self.test_config.cluster.
eventing_metadata_bucket_mem_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.EVICTION_POLICY,
bucket_type=self.test_config.bucket.BUCKET_TYPE,
)
def configure_auto_compaction(self):
compaction_settings = self.test_config.compaction
for master in self.cluster_spec.masters:
self.rest.configure_auto_compaction(master, compaction_settings)
settings = self.rest.get_auto_compaction_settings(master)
logger.info('Auto-compaction settings: {}'.format(
pretty_dict(settings)))
def configure_internal_settings(self):
internal_settings = self.test_config.internal_settings
for master in self.cluster_spec.masters:
for parameter, value in internal_settings.items():
self.rest.set_internal_settings(master,
{parameter: maybe_atoi(value)})
def configure_xdcr_settings(self):
xdcr_cluster_settings = self.test_config.xdcr_cluster_settings
for master in self.cluster_spec.masters:
for parameter, value in xdcr_cluster_settings.items():
self.rest.set_xdcr_cluster_settings(
master, {parameter: maybe_atoi(value)})
def tweak_memory(self):
self.remote.reset_swap()
self.remote.drop_caches()
self.remote.set_swappiness()
self.remote.disable_thp()
def enable_n2n_encryption(self):
if self.test_config.cluster.enable_n2n_encryption:
for master in self.cluster_spec.masters:
self.remote.enable_n2n_encryption(
master, self.test_config.cluster.enable_n2n_encryption)
def restart_with_alternative_num_vbuckets(self):
num_vbuckets = self.test_config.cluster.num_vbuckets
if num_vbuckets is not None:
self.remote.restart_with_alternative_num_vbuckets(num_vbuckets)
def restart_with_alternative_bucket_options(self):
"""Apply custom buckets settings.
Tune bucket settings (e.g., max_num_shards or max_num_auxio) using
"/diag/eval" and restart the entire cluster.
"""
if self.test_config.bucket_extras:
self.remote.enable_nonlocal_diag_eval()
cmd = 'ns_bucket:update_bucket_props("{}", ' \
'[{{extra_config_string, "{}={}"}}]).'
for option, value in self.test_config.bucket_extras.items():
if option == 'num_writer_threads':
self.rest.set_num_writer_threads(self.master_node, int(value))
elif option == 'num_reader_threads':
self.rest.set_num_reader_threads(self.master_node, int(value))
else:
logger.info('Changing {} to {}'.format(option, value))
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
diag_eval = cmd.format(bucket, option, value)
self.rest.run_diag_eval(master, diag_eval)
if self.test_config.bucket_extras:
self.remote.restart()
self.wait_until_healthy()
def tune_logging(self):
self.remote.tune_log_rotation()
self.remote.restart()
def enable_auto_failover(self):
enabled = self.test_config.bucket.autofailover_enabled
failover_min = self.test_config.bucket.failover_min
failover_max = self.test_config.bucket.failover_max
for master in self.cluster_spec.masters:
self.rest.set_auto_failover(master, enabled, failover_min,
failover_max)
def wait_until_warmed_up(self):
if self.test_config.bucket.bucket_type in ('ephemeral', 'memcached'):
return
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
self.monitor.monitor_warmup(self.memcached, master, bucket)
def wait_until_healthy(self):
for master in self.cluster_spec.masters:
self.monitor.monitor_node_health(master)
for analytics_node in self.rest.get_active_nodes_by_role(
master, 'cbas'):
self.monitor.monitor_analytics_node_active(analytics_node)
def gen_disabled_audit_events(self, master: str) -> List[str]:
curr_settings = self.rest.get_audit_settings(master)
curr_disabled = {str(event) for event in curr_settings['disabled']}
disabled = curr_disabled - self.test_config.audit_settings.extra_events
return list(disabled)
def enable_audit(self):
if not self.is_compatible(min_release='4.0.0') or \
self.rest.is_community(self.master_node):
return
if not self.test_config.audit_settings.enabled:
return
for master in self.cluster_spec.masters:
disabled = []
if self.test_config.audit_settings.extra_events:
disabled = self.gen_disabled_audit_events(master)
self.rest.enable_audit(master, disabled)
def generate_ce_roles(self) -> List[str]:
return ['admin']
def generate_ee_roles(self) -> List[str]:
existing_roles = {
r['role']
for r in self.rest.get_rbac_roles(self.master_node)
}
roles = []
for role in (
'bucket_admin',
'data_dcp_reader',
'data_monitoring',
'data_reader_writer',
'data_reader',
'data_writer',
'fts_admin',
'fts_searcher',
'query_delete',
'query_insert',
'query_select',
'query_update',
'views_admin',
):
if role in existing_roles:
roles.append(role + '[{bucket}]')
return roles
def delete_rbac_users(self):
if not self.is_compatible(min_release='5.0'):
return
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
self.rest.delete_rbac_user(host=master, bucket=bucket)
def add_rbac_users(self):
if not self.rest.supports_rbac(self.master_node):
logger.info('RBAC not supported - skipping adding RBAC users')
return
if self.rest.is_community(self.master_node):
roles = self.generate_ce_roles()
else:
roles = self.generate_ee_roles()
for master in self.cluster_spec.masters:
admin_user, admin_password = self.cluster_spec.rest_credentials
self.rest.add_rbac_user(
host=master,
user=admin_user,
password=admin_password,
roles=['admin'],
)
for bucket in self.test_config.buckets:
bucket_roles = [role.format(bucket=bucket) for role in roles]
bucket_roles.append("admin")
self.rest.add_rbac_user(
host=master,
user=bucket, # Backward compatibility
password=self.test_config.bucket.password,
roles=bucket_roles,
)
def throttle_cpu(self):
if self.remote.os == 'Cygwin':
return
self.remote.enable_cpu()
if self.test_config.cluster.online_cores:
self.remote.disable_cpu(self.test_config.cluster.online_cores)
def tune_memory_settings(self):
kernel_memory = self.test_config.cluster.kernel_mem_limit
if kernel_memory:
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.tune_memory_settings(host_string=server,
size=kernel_memory)
self.monitor.wait_for_servers()
def reset_memory_settings(self):
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.reset_memory_settings(host_string=server)
self.monitor.wait_for_servers()
def flush_iptables(self):
self.remote.flush_iptables()
def clear_login_history(self):
self.remote.clear_wtmp()
def disable_wan(self):
self.remote.disable_wan()
def enable_ipv6(self):
if self.test_config.cluster.ipv6:
self.remote.enable_ipv6()
def set_x509_certificates(self):
logger.info('Setting x509 settings')
if self.test_config.access_settings.ssl_mode == "auth":
self.remote.setup_x509()
for host in self.cluster_spec.servers:
self.rest.upload_cluster_certificate(host)
for host in self.cluster_spec.servers:
self.rest.reload_cluster_certificate(host)
self.rest.enable_certificate_auth(host)
def set_cipher_suite(self):
if self.test_config.access_settings.cipher_list:
check_cipher_suit = self.rest.get_cipher_suite(self.master_node)
logger.info('current cipher suit: {}'.format(check_cipher_suit))
self.rest.set_cipher_suite(
self.master_node, self.test_config.access_settings.cipher_list)
check_cipher_suit = self.rest.get_cipher_suite(self.master_node)
logger.info('new cipher suit: {}'.format(check_cipher_suit))
def set_min_tls_version(self):
if self.test_config.access_settings.min_tls_version:
check_tls_version = self.rest.get_minimum_tls_version(
self.master_node)
logger.info('current tls version: {}'.format(check_tls_version))
self.rest.set_minimum_tls_version(
self.master_node,
self.test_config.access_settings.min_tls_version)
check_tls_version = self.rest.get_minimum_tls_version(
self.master_node)
logger.info('new tls version: {}'.format(check_tls_version))
class ClusterManager:
def __init__(self,
cluster_spec: ClusterSpec,
test_config: TestConfig,
verbose: bool = False):
self.cluster_spec = cluster_spec
self.test_config = test_config
self.dynamic_infra = self.cluster_spec.dynamic_infrastructure
self.rest = RestHelper(cluster_spec)
self.remote = RemoteHelper(cluster_spec, verbose)
self.monitor = Monitor(cluster_spec, test_config, verbose)
self.memcached = MemcachedHelper(test_config)
self.master_node = next(self.cluster_spec.masters)
self.initial_nodes = test_config.cluster.initial_nodes
self.build = self.rest.get_version(self.master_node)
def is_compatible(self, min_release: str) -> bool:
for master in self.cluster_spec.masters:
version = self.rest.get_version(master)
return version >= min_release
def set_data_path(self):
if self.dynamic_infra:
return
for server in self.cluster_spec.servers:
self.remote.change_owner(server, self.cluster_spec.data_path)
self.rest.set_data_path(server, self.cluster_spec.data_path)
def set_index_path(self):
if self.dynamic_infra:
return
for server in self.cluster_spec.servers:
self.remote.change_owner(server, self.cluster_spec.index_path)
self.rest.set_index_path(server, self.cluster_spec.index_path)
def set_analytics_path(self):
if self.dynamic_infra:
return
paths = []
for path in self.cluster_spec.analytics_paths:
for i in range(self.test_config.analytics_settings.num_io_devices):
io_device = '{}/dev{}'.format(path, i)
paths.append(io_device)
for server in self.cluster_spec.servers_by_role('cbas'):
for path in self.cluster_spec.analytics_paths:
self.remote.change_owner(server, path)
self.rest.set_analytics_paths(server, paths)
def rename(self):
if self.dynamic_infra:
return
else:
for server in self.cluster_spec.servers:
self.rest.rename(server)
def set_auth(self):
if self.dynamic_infra:
return
else:
for server in self.cluster_spec.servers:
self.rest.set_auth(server)
def set_mem_quotas(self):
if self.dynamic_infra:
cluster = self.remote.get_cluster()
cluster['spec']['cluster']['dataServiceMemoryQuota'] = \
'{}Mi'.format(self.test_config.cluster.mem_quota)
cluster['spec']['cluster']['indexServiceMemoryQuota'] = \
'{}Mi'.format(self.test_config.cluster.index_mem_quota)
if self.test_config.cluster.fts_index_mem_quota:
cluster['spec']['cluster']['searchServiceMemoryQuota'] = \
'{}Mi'.format(self.test_config.cluster.fts_index_mem_quota)
if self.test_config.cluster.analytics_mem_quota:
cluster['spec']['cluster']['analyticsServiceMemoryQuota'] = \
'{}Mi'.format(self.test_config.cluster.analytics_mem_quota)
if self.test_config.cluster.eventing_mem_quota:
cluster['spec']['cluster']['eventingServiceMemoryQuota'] = \
'{}Mi'.format(self.test_config.cluster.eventing_mem_quota)
self.remote.update_cluster_config(cluster)
else:
for master in self.cluster_spec.masters:
self.rest.set_mem_quota(master,
self.test_config.cluster.mem_quota)
self.rest.set_index_mem_quota(
master, self.test_config.cluster.index_mem_quota)
if self.test_config.cluster.fts_index_mem_quota:
self.rest.set_fts_index_mem_quota(
master, self.test_config.cluster.fts_index_mem_quota)
if self.test_config.cluster.analytics_mem_quota:
self.rest.set_analytics_mem_quota(
master, self.test_config.cluster.analytics_mem_quota)
if self.test_config.cluster.eventing_mem_quota:
self.rest.set_eventing_mem_quota(
master, self.test_config.cluster.eventing_mem_quota)
def set_query_settings(self):
logger.info('Setting query settings')
if self.dynamic_infra:
return
query_nodes = self.cluster_spec.servers_by_role('n1ql')
if query_nodes:
settings = self.test_config.n1ql_settings.cbq_settings
if settings:
self.rest.set_query_settings(query_nodes[0], settings)
settings = self.rest.get_query_settings(query_nodes[0])
settings = pretty_dict(settings)
logger.info('Query settings: {}'.format(settings))
def set_index_settings(self):
logger.info('Setting index settings')
index_nodes = self.cluster_spec.servers_by_role('index')
if index_nodes:
settings = self.test_config.gsi_settings.settings
if settings:
if self.dynamic_infra:
cluster = self.remote.get_cluster()
cluster['spec']['cluster']['indexStorageSetting'] = \
settings['indexer.settings.storage_mode']
self.remote.update_cluster_config(cluster,
timeout=300,
reboot=True)
logger.info('Index settings: {}'.format(settings))
else:
self.rest.set_index_settings(index_nodes[0], settings)
settings = self.rest.get_index_settings(index_nodes[0])
settings = pretty_dict(settings)
logger.info('Index settings: {}'.format(settings))
def set_services(self):
if self.dynamic_infra:
cluster = self.remote.get_cluster()
server_types = dict()
server_roles = self.cluster_spec.roles
for server, role in server_roles.items():
role = role\
.replace('kv', 'data')\
.replace('n1ql', 'query')
server_type_count = server_types.get(role, 0)
server_types[role] = server_type_count + 1
istio = 'false'
if self.cluster_spec.istio_enabled(cluster_name='k8s_cluster_1'):
istio = 'true'
cluster_servers = []
operator_version = self.remote.get_operator_version()
operator_major = int(operator_version.split(".")[0])
operator_minor = int(operator_version.split(".")[1])
for server_role, server_role_count in server_types.items():
node_selector = {
'{}_enabled'.format(
service.replace('data', 'kv').replace('query',
'n1ql')): 'true'
for service in server_role.split(",")
}
node_selector['NodeRoles'] = 'couchbase1'
spec = {
'imagePullSecrets': [{
'name': 'regcred'
}],
'nodeSelector': node_selector,
}
if (operator_major, operator_minor) <= (2, 1):
spec['containers'] = []
pod_def =\
{
'spec': spec,
'metadata':
{
'annotations': {'sidecar.istio.io/inject': istio}
}
}
server_def = \
{
'name': server_role.replace(",", "-"),
'services': server_role.split(","),
'pod': pod_def,
'size': server_role_count,
'volumeMounts': {'default': 'couchbase_kv'}
}
cluster_servers.append(server_def)
cluster['spec']['servers'] = cluster_servers
self.remote.update_cluster_config(cluster,
timeout=300,
reboot=True)
else:
if not self.is_compatible(min_release='4.0.0'):
return
for master in self.cluster_spec.masters:
roles = self.cluster_spec.roles[master]
self.rest.set_services(master, roles)
def add_nodes(self):
if self.dynamic_infra:
return
for (_, servers), initial_nodes \
in zip(self.cluster_spec.clusters, self.initial_nodes):
if initial_nodes < 2: # Single-node cluster
continue
master = servers[0]
for node in servers[1:initial_nodes]:
roles = self.cluster_spec.roles[node]
self.rest.add_node(master, node, roles)
def rebalance(self):
if self.dynamic_infra:
return
for (_, servers), initial_nodes \
in zip(self.cluster_spec.clusters, self.initial_nodes):
master = servers[0]
known_nodes = servers[:initial_nodes]
ejected_nodes = []
self.rest.rebalance(master, known_nodes, ejected_nodes)
self.monitor.monitor_rebalance(master)
self.wait_until_healthy()
def increase_bucket_limit(self, num_buckets: int):
if self.dynamic_infra:
return
for master in self.cluster_spec.masters:
self.rest.increase_bucket_limit(master, num_buckets)
def flush_buckets(self):
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.flush_bucket(host=master, bucket=bucket_name)
def delete_buckets(self):
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.delete_bucket(host=master, name=bucket_name)
def create_buckets(self):
mem_quota = self.test_config.cluster.mem_quota
if self.test_config.cluster.num_buckets > 7:
self.increase_bucket_limit(self.test_config.cluster.num_buckets +
3)
if self.test_config.cluster.eventing_metadata_bucket_mem_quota:
mem_quota -= (
self.test_config.cluster.eventing_metadata_bucket_mem_quota +
self.test_config.cluster.eventing_bucket_mem_quota)
per_bucket_quota = mem_quota // self.test_config.cluster.num_buckets
if self.dynamic_infra:
self.remote.delete_all_buckets()
for bucket_name in self.test_config.buckets:
self.remote.create_bucket(bucket_name, per_bucket_quota,
self.test_config.bucket)
else:
if self.test_config.bucket.backend_storage == 'magma':
self.enable_developer_preview()
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.create_bucket(
host=master,
name=bucket_name,
ram_quota=per_bucket_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.
eviction_policy,
bucket_type=self.test_config.bucket.bucket_type,
backend_storage=self.test_config.bucket.
backend_storage,
conflict_resolution_type=self.test_config.bucket.
conflict_resolution_type,
compression_mode=self.test_config.bucket.
compression_mode,
)
def create_collections(self):
if self.dynamic_infra:
return
collection_map = self.test_config.collection.collection_map
for master in self.cluster_spec.masters:
if collection_map is not None:
if self.test_config.collection.use_bulk_api:
for bucket in collection_map.keys():
create_scopes = []
for scope in collection_map[bucket]:
scope_collections = []
for collection in collection_map[bucket][scope]:
scope_collections.append({"name": collection})
create_scopes.append({
"name":
scope,
"collections":
scope_collections
})
self.rest.set_collection_map(master, bucket,
{"scopes": create_scopes})
else:
for bucket in collection_map.keys():
delete_default = True
for scope in collection_map[bucket]:
if scope == '_default':
for collection in collection_map[bucket][
scope]:
if collection == "_default":
delete_default = False
if delete_default:
self.rest.delete_collection(
master, bucket, '_default', '_default')
for bucket in collection_map.keys():
for scope in collection_map[bucket]:
if scope != '_default':
self.rest.create_scope(master, bucket, scope)
for collection in collection_map[bucket][scope]:
if collection != '_default':
self.rest.create_collection(
master, bucket, scope, collection)
def create_eventing_buckets(self):
if not self.test_config.cluster.eventing_bucket_mem_quota:
return
if self.dynamic_infra:
return
per_bucket_quota = \
self.test_config.cluster.eventing_bucket_mem_quota \
// self.test_config.cluster.eventing_buckets
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.eventing_buckets:
self.rest.create_bucket(
host=master,
name=bucket_name,
ram_quota=per_bucket_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.eviction_policy,
bucket_type=self.test_config.bucket.bucket_type,
conflict_resolution_type=self.test_config.bucket.
conflict_resolution_type,
)
def create_eventing_metadata_bucket(self):
if not self.test_config.cluster.eventing_metadata_bucket_mem_quota:
return
if self.dynamic_infra:
return
for master in self.cluster_spec.masters:
self.rest.create_bucket(
host=master,
name=self.test_config.cluster.EVENTING_METADATA_BUCKET_NAME,
ram_quota=self.test_config.cluster.
eventing_metadata_bucket_mem_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.EVICTION_POLICY,
bucket_type=self.test_config.bucket.BUCKET_TYPE,
)
def configure_auto_compaction(self):
compaction_settings = self.test_config.compaction
if self.dynamic_infra:
cluster = self.remote.get_cluster()
db = int(compaction_settings.db_percentage)
view = int(compaction_settings.view_percentage)
para = bool(str(compaction_settings.parallel).lower())
auto_compaction = cluster['spec']['cluster']\
.get('autoCompaction',
{'databaseFragmentationThreshold': {'percent': 30},
'viewFragmentationThreshold': {'percent': 30},
'parallelCompaction': False})
db_percent = auto_compaction.get('databaseFragmentationThreshold',
{'percent': 30})
db_percent['percent'] = db
auto_compaction['databaseFragmentationThreshold'] = db_percent
views_percent = auto_compaction.get('viewFragmentationThreshold',
{'percent': 30})
views_percent['percent'] = view
auto_compaction['viewFragmentationThreshold'] = views_percent
auto_compaction['parallelCompaction'] = para
self.remote.update_cluster_config(cluster)
else:
for master in self.cluster_spec.masters:
self.rest.configure_auto_compaction(master,
compaction_settings)
settings = self.rest.get_auto_compaction_settings(master)
logger.info('Auto-compaction settings: {}'.format(
pretty_dict(settings)))
def configure_internal_settings(self):
internal_settings = self.test_config.internal_settings
for master in self.cluster_spec.masters:
for parameter, value in internal_settings.items():
if self.dynamic_infra:
raise Exception(
'not supported for dynamic infrastructure yet')
else:
self.rest.set_internal_settings(
master, {parameter: maybe_atoi(value)})
def configure_xdcr_settings(self):
xdcr_cluster_settings = self.test_config.xdcr_cluster_settings
if self.dynamic_infra:
return
for master in self.cluster_spec.masters:
for parameter, value in xdcr_cluster_settings.items():
self.rest.set_xdcr_cluster_settings(
master, {parameter: maybe_atoi(value)})
def tweak_memory(self):
if self.dynamic_infra:
return
self.remote.reset_swap()
self.remote.drop_caches()
self.remote.set_swappiness()
self.remote.disable_thp()
def enable_n2n_encryption(self):
if self.dynamic_infra:
return
if self.test_config.cluster.enable_n2n_encryption:
for master in self.cluster_spec.masters:
self.remote.enable_n2n_encryption(
master, self.test_config.cluster.enable_n2n_encryption)
def restart_with_alternative_num_vbuckets(self):
num_vbuckets = self.test_config.cluster.num_vbuckets
if num_vbuckets is not None:
if self.dynamic_infra:
raise Exception('not supported for dynamic infrastructure yet')
else:
self.remote.restart_with_alternative_num_vbuckets(num_vbuckets)
def restart_with_alternative_bucket_options(self):
"""Apply custom buckets settings.
Tune bucket settings (e.g., max_num_shards or max_num_auxio) using
"/diag/eval" and restart the entire cluster.
"""
if self.dynamic_infra:
return
if self.test_config.bucket_extras:
self.remote.enable_nonlocal_diag_eval()
cmd = 'ns_bucket:update_bucket_props("{}", ' \
'[{{extra_config_string, "{}={}"}}]).'
for option, value in self.test_config.bucket_extras.items():
if re.search("^num_.*_threads$", option):
self.rest.set_num_threads(self.master_node, option, value)
else:
logger.info('Changing {} to {}'.format(option, value))
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
diag_eval = cmd.format(bucket, option, value)
self.rest.run_diag_eval(master, diag_eval)
if self.test_config.bucket_extras:
self.disable_auto_failover()
self.remote.restart()
self.wait_until_healthy()
self.enable_auto_failover()
def tune_logging(self):
if self.dynamic_infra:
return
self.remote.tune_log_rotation()
self.remote.restart()
def enable_auto_failover(self):
enabled = self.test_config.bucket.autofailover_enabled
failover_min = self.test_config.bucket.failover_min
failover_max = self.test_config.bucket.failover_max
if self.dynamic_infra:
cluster = self.remote.get_cluster()
cluster['spec']['cluster']['autoFailoverMaxCount'] = 1
cluster['spec']['cluster']['autoFailoverServerGroup'] = bool(
enabled)
cluster['spec']['cluster']['autoFailoverOnDataDiskIssues'] = bool(
enabled)
cluster['spec']['cluster']['autoFailoverOnDataDiskIssuesTimePeriod'] = \
'{}s'.format(10)
cluster['spec']['cluster']['autoFailoverTimeout'] = \
'{}s'.format(failover_max)
self.remote.update_cluster_config(cluster)
else:
for master in self.cluster_spec.masters:
self.rest.set_auto_failover(master, enabled, failover_min,
failover_max)
def disable_auto_failover(self):
enabled = 'false'
failover_min = self.test_config.bucket.failover_min
failover_max = self.test_config.bucket.failover_max
if self.dynamic_infra:
cluster = self.remote.get_cluster()
cluster['spec']['cluster']['autoFailoverMaxCount'] = 1
cluster['spec']['cluster']['autoFailoverServerGroup'] = bool(
enabled)
cluster['spec']['cluster']['autoFailoverOnDataDiskIssues'] = bool(
enabled)
cluster['spec']['cluster']['autoFailoverOnDataDiskIssuesTimePeriod'] = \
'{}s'.format(10)
cluster['spec']['cluster']['autoFailoverTimeout'] = \
'{}s'.format(failover_max)
self.remote.update_cluster_config(cluster)
else:
for master in self.cluster_spec.masters:
self.rest.set_auto_failover(master, enabled, failover_min,
failover_max)
def wait_until_warmed_up(self):
if self.test_config.bucket.bucket_type in ('ephemeral', 'memcached'):
return
if self.dynamic_infra:
self.remote.wait_for_cluster_ready()
else:
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
self.monitor.monitor_warmup(self.memcached, master, bucket)
def wait_until_healthy(self):
if self.dynamic_infra:
self.remote.wait_for_cluster_ready()
else:
for master in self.cluster_spec.masters:
self.monitor.monitor_node_health(master)
for analytics_node in self.rest.get_active_nodes_by_role(
master, 'cbas'):
self.monitor.monitor_analytics_node_active(analytics_node)
def gen_disabled_audit_events(self, master: str) -> List[str]:
curr_settings = self.rest.get_audit_settings(master)
curr_disabled = {str(event) for event in curr_settings['disabled']}
disabled = curr_disabled - self.test_config.audit_settings.extra_events
return list(disabled)
def enable_audit(self):
if self.dynamic_infra:
return
if not self.is_compatible(min_release='4.0.0') or \
self.rest.is_community(self.master_node):
return
if not self.test_config.audit_settings.enabled:
return
for master in self.cluster_spec.masters:
disabled = []
if self.test_config.audit_settings.extra_events:
disabled = self.gen_disabled_audit_events(master)
self.rest.enable_audit(master, disabled)
def generate_ce_roles(self) -> List[str]:
return ['admin']
def generate_ee_roles(self) -> List[str]:
existing_roles = {
r['role']
for r in self.rest.get_rbac_roles(self.master_node)
}
roles = []
for role in (
'bucket_admin',
'data_dcp_reader',
'data_monitoring',
'data_reader_writer',
'data_reader',
'data_writer',
'fts_admin',
'fts_searcher',
'query_delete',
'query_insert',
'query_select',
'query_update',
'views_admin',
):
if role in existing_roles:
roles.append(role + '[{bucket}]')
return roles
def delete_rbac_users(self):
if not self.is_compatible(min_release='5.0'):
return
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
self.rest.delete_rbac_user(host=master, bucket=bucket)
def add_rbac_users(self):
if self.dynamic_infra:
self.remote.create_from_file(
"cloud/operator/2/1/user-password-secret.yaml")
# self.remote.create_from_file("cloud/operator/2/1/admin-user.yaml")
self.remote.create_from_file("cloud/operator/2/1/bucket-user.yaml")
self.remote.create_from_file(
"cloud/operator/2/1/rbac-admin-group.yaml")
self.remote.create_from_file(
"cloud/operator/2/1/rbac-admin-role-binding.yaml")
else:
if not self.rest.supports_rbac(self.master_node):
logger.info('RBAC not supported - skipping adding RBAC users')
return
if self.rest.is_community(self.master_node):
roles = self.generate_ce_roles()
else:
roles = self.generate_ee_roles()
for master in self.cluster_spec.masters:
admin_user, admin_password = self.cluster_spec.rest_credentials
self.rest.add_rbac_user(
host=master,
user=admin_user,
password=admin_password,
roles=['admin'],
)
buckets = self.test_config.buckets + self.test_config.eventing_buckets
for bucket in buckets:
bucket_roles = [
role.format(bucket=bucket) for role in roles
]
bucket_roles.append("admin")
self.rest.add_rbac_user(
host=master,
user=bucket, # Backward compatibility
password=self.test_config.bucket.password,
roles=bucket_roles,
)
def add_extra_rbac_users(self, num_users):
if not self.rest.supports_rbac(self.master_node):
logger.info('RBAC not supported - skipping adding RBAC users')
return
if self.rest.is_community(self.master_node):
roles = self.generate_ce_roles()
else:
roles = self.generate_ee_roles()
for master in self.cluster_spec.masters:
admin_user, admin_password = self.cluster_spec.rest_credentials
self.rest.add_rbac_user(
host=master,
user=admin_user,
password=admin_password,
roles=['admin'],
)
for bucket in self.test_config.buckets:
bucket_roles = [role.format(bucket=bucket) for role in roles]
bucket_roles.append("admin")
for i in range(1, num_users + 1):
user = '******'.format(user_number=str(i))
self.rest.add_rbac_user(
host=master,
user=user,
password=self.test_config.bucket.password,
roles=bucket_roles,
)
def throttle_cpu(self):
if self.dynamic_infra:
cluster = self.remote.get_cluster()
if self.test_config.cluster.enable_cpu_cores:
server_groups = cluster['spec']['servers']
updated_server_groups = []
default_cpu = 80
for server_group in server_groups:
resources = server_group.get('resources', {})
limits = resources.get('limits', {})
limits['cpu'] = default_cpu
resources['limits'] = limits
server_group['resources'] = resources
updated_server_groups.append(server_group)
cluster['spec']['servers'] = updated_server_groups
if self.test_config.cluster.online_cores:
server_groups = cluster['spec']['servers']
updated_server_groups = []
online_vcpus = self.test_config.cluster.online_cores * 2
for server_group in server_groups:
resources = server_group.get('resources', {})
limits = resources.get('limits', {})
limits['cpu'] = online_vcpus
resources['limits'] = limits
server_group['resources'] = resources
updated_server_groups.append(server_group)
cluster['spec']['servers'] = updated_server_groups
self.remote.update_cluster_config(cluster,
timeout=300,
reboot=True)
else:
if self.remote.os == 'Cygwin':
return
if self.test_config.cluster.enable_cpu_cores:
self.remote.enable_cpu()
if self.test_config.cluster.online_cores:
self.remote.disable_cpu(self.test_config.cluster.online_cores)
def tune_memory_settings(self):
kernel_memory = self.test_config.cluster.kernel_mem_limit
if kernel_memory:
if self.dynamic_infra:
cluster = self.remote.get_cluster()
server_groups = cluster['spec']['servers']
tune_services = set()
# CAO uses different service names than perfrunner
for service in self.test_config.cluster.kernel_mem_limit_services:
if service == 'kv':
service = 'data'
elif service == 'n1ql':
service = 'query'
elif service == 'fts':
service = 'search'
elif service == 'cbas':
service = 'analytics'
tune_services.add(service)
updated_server_groups = []
default_mem = '128Gi'
for server_group in server_groups:
services_in_group = set(server_group['services'])
resources = server_group.get('resources', {})
limits = resources.get('limits', {})
mem_limit = limits.get('memory', default_mem)
if services_in_group.intersection(
tune_services) and kernel_memory != 0:
mem_limit = '{}Mi'.format(kernel_memory)
limits['memory'] = mem_limit
resources['limits'] = limits
server_group['resources'] = resources
updated_server_groups.append(server_group)
cluster['spec']['servers'] = updated_server_groups
self.remote.update_cluster_config(cluster,
timeout=300,
reboot=True)
else:
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.tune_memory_settings(host_string=server,
size=kernel_memory)
self.monitor.wait_for_servers()
def reset_memory_settings(self):
if self.dynamic_infra:
return
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.reset_memory_settings(host_string=server)
self.monitor.wait_for_servers()
def flush_iptables(self):
if self.dynamic_infra:
return
self.remote.flush_iptables()
def clear_login_history(self):
if self.dynamic_infra:
return
self.remote.clear_wtmp()
def disable_wan(self):
if self.dynamic_infra:
return
self.remote.disable_wan()
def enable_ipv6(self):
if self.dynamic_infra:
return
if self.test_config.cluster.ipv6:
version, build_number = self.build.split('-')
build = tuple(map(int, version.split('.'))) + (int(build_number), )
if build < (6, 5, 0, 0):
self.remote.update_ip_family_rest()
else:
self.remote.update_ip_family_cli()
self.remote.enable_ipv6()
def set_x509_certificates(self):
if self.dynamic_infra:
return
logger.info('Setting x509 settings')
if self.test_config.access_settings.ssl_mode == "auth":
self.remote.setup_x509()
for host in self.cluster_spec.servers:
self.rest.upload_cluster_certificate(host)
for host in self.cluster_spec.servers:
self.rest.reload_cluster_certificate(host)
self.rest.enable_certificate_auth(host)
def set_cipher_suite(self):
if self.dynamic_infra:
return
if self.test_config.access_settings.cipher_list:
check_cipher_suit = self.rest.get_cipher_suite(self.master_node)
logger.info('current cipher suit: {}'.format(check_cipher_suit))
self.rest.set_cipher_suite(
self.master_node, self.test_config.access_settings.cipher_list)
check_cipher_suit = self.rest.get_cipher_suite(self.master_node)
logger.info('new cipher suit: {}'.format(check_cipher_suit))
def set_min_tls_version(self):
if self.dynamic_infra:
return
if self.test_config.access_settings.min_tls_version:
check_tls_version = self.rest.get_minimum_tls_version(
self.master_node)
logger.info('current tls version: {}'.format(check_tls_version))
self.rest.set_minimum_tls_version(
self.master_node,
self.test_config.access_settings.min_tls_version)
check_tls_version = self.rest.get_minimum_tls_version(
self.master_node)
logger.info('new tls version: {}'.format(check_tls_version))
def get_debug_rpm_url(self):
release, build_number = self.build.split('-')
build = tuple(map(int, release.split('.'))) + (int(build_number), )
if build > (7, 0, 0, 0):
release = 'cheshire-cat'
elif build > (6, 5, 0, 0) and build < (7, 0, 0, 0):
release = 'mad-hatter'
elif build < (6, 5, 0, 0):
release = 'alice'
centos_version = self.remote.detect_centos_release()
rpm_url = 'http://latestbuilds.service.couchbase.com/builds/' \
'latestbuilds/couchbase-server/{}/{}/' \
'couchbase-server-enterprise-debuginfo-{}-centos{}.x86_64.rpm' \
''.format(release, build_number, self.build, centos_version)
return rpm_url
def install_cb_debug_rpm(self):
self.remote.install_cb_debug_rpm(url=self.get_debug_rpm_url())
def enable_developer_preview(self):
release, build_number = self.build.split('-')
build = tuple(map(int, release.split('.'))) + (int(build_number), )
if build > (7, 0, 0, 4698) or build < (0, 0, 0, 9999):
self.remote.enable_developer_preview()
class Profiler:
DEBUG_PORTS = {
'fts': 8094,
'index': 9102,
'kv': 9998, # goxdcr
'n1ql': 8093,
}
ENDPOINTS = {
'cpu': 'http://127.0.0.1:{}/debug/pprof/profile',
'heap': 'http://127.0.0.1:{}/debug/pprof/heap',
'goroutine': 'http://127.0.0.1:{}/debug/pprof/goroutine?debug=2',
}
def __init__(self, cluster_spec: ClusterSpec, test_config: TestConfig):
self.test_config = test_config
self.rest = RestHelper(cluster_spec)
self.master_node = next(cluster_spec.masters)
self.ssh_username, self.ssh_password = cluster_spec.ssh_credentials
def new_tunnel(self, host: str, port: int) -> SSHTunnelForwarder:
return SSHTunnelForwarder(
ssh_address_or_host=host,
ssh_username=self.ssh_username,
ssh_password=self.ssh_password,
remote_bind_address=('127.0.0.1', port),
)
def save(self, host: str, service: str, profile: str, content: bytes):
fname = '{}_{}_{}_{}.pprof'.format(host, service, profile, uhex()[:6])
with open(fname, 'wb') as fh:
fh.write(content)
def profile(self, host: str, service: str, profile: str):
logger.info('Collecting {} profile on {}'.format(profile, host))
endpoint = self.ENDPOINTS[profile]
port = self.DEBUG_PORTS[service]
with self.new_tunnel(host, port) as tunnel:
url = endpoint.format(tunnel.local_bind_port)
response = requests.get(url=url)
self.save(host, service, profile, response.content)
def timer(self, **kwargs):
timer = Timer(
function=self.profile,
interval=self.test_config.profiling_settings.interval,
num_runs=self.test_config.profiling_settings.num_profiles,
kwargs=kwargs,
)
timer.start()
def schedule(self):
for service in self.test_config.profiling_settings.services:
logger.info(
'Scheduling profiling of "{}" services'.format(service))
for server in self.rest.get_active_nodes_by_role(self.master_node,
role=service):
for profile in self.test_config.profiling_settings.profiles:
self.timer(host=server, service=service, profile=profile)
class PerfTest:
COLLECTORS = {}
ROOT_CERTIFICATE = 'root.pem'
def __init__(self, cluster_spec: ClusterSpec, test_config: TestConfig,
verbose: bool):
self.cluster_spec = cluster_spec
self.test_config = test_config
self.dynamic_infra = self.cluster_spec.dynamic_infrastructure
self.target_iterator = TargetIterator(cluster_spec, test_config)
self.cluster = ClusterManager(cluster_spec, test_config)
self.remote = RemoteHelper(cluster_spec, verbose)
self.profiler = Profiler(cluster_spec, test_config)
self.master_node = next(cluster_spec.masters)
self.memcached = MemcachedHelper(test_config)
self.monitor = Monitor(cluster_spec, test_config, verbose)
self.rest = RestHelper(cluster_spec, test_config)
self.build = self.rest.get_version(self.master_node)
self.metrics = MetricHelper(self)
self.reporter = ShowFastReporter(cluster_spec, test_config, self.build)
self.cbmonitor_snapshots = []
self.cbmonitor_clusters = []
if self.test_config.test_case.use_workers:
self.worker_manager = WorkerManager(cluster_spec, test_config,
verbose)
if self.test_config.cluster.enable_n2n_encryption:
self.download_certificate()
def __enter__(self):
return self
def __exit__(self, exc_type, exc_val, exc_tb):
failure = self.debug()
self.tear_down()
if exc_type == KeyboardInterrupt:
logger.warn('The test was interrupted')
return True
if failure:
logger.interrupt(failure)
@property
def query_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'n1ql')
@property
def index_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'index')
@property
def fts_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'fts')
@property
def analytics_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'cbas')
@property
def eventing_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'eventing')
def tear_down(self):
if self.test_config.profiling_settings.linux_perf_profile_flag:
self.collect_linux_perf_profiles()
if self.test_config.test_case.use_workers:
self.worker_manager.download_celery_logs()
self.worker_manager.terminate()
if self.test_config.cluster.online_cores:
self.remote.enable_cpu()
if self.test_config.cluster.kernel_mem_limit:
self.collect_logs()
self.cluster.reset_memory_settings()
def collect_linux_perf_profiles(self):
self.remote.generate_linux_perf_script()
self.remote.get_linuxperf_files()
def collect_logs(self):
self.remote.collect_info()
for hostname in self.cluster_spec.servers:
for fname in glob.glob('{}/*.zip'.format(hostname)):
shutil.move(fname, '{}.zip'.format(hostname))
def reset_memory_settings(self):
if self.test_config.cluster.kernel_mem_limit:
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.reset_memory_settings(host_string=server)
self.monitor.wait_for_servers()
def debug(self) -> str:
failure = self.check_core_dumps()
failure = self.check_rebalance() or failure
return self.check_failover() or failure
def download_certificate(self):
cert = self.rest.get_certificate(self.master_node)
with open(self.ROOT_CERTIFICATE, 'w') as fh:
fh.write(cert)
def check_rebalance(self) -> str:
if self.dynamic_infra:
pass
else:
for master in self.cluster_spec.masters:
if self.rest.is_not_balanced(master):
return 'The cluster is not balanced'
def check_failover(self) -> Optional[str]:
if self.dynamic_infra:
return
if hasattr(self, 'rebalance_settings'):
if self.rebalance_settings.failover or \
self.rebalance_settings.graceful_failover:
return
for master in self.cluster_spec.masters:
num_failovers = self.rest.get_failover_counter(master)
if num_failovers:
return 'Failover happened {} time(s)'.format(num_failovers)
def check_core_dumps(self) -> str:
dumps_per_host = self.remote.detect_core_dumps()
core_dumps = {
host: dumps
for host, dumps in dumps_per_host.items() if dumps
}
if core_dumps:
return pretty_dict(core_dumps)
def restore(self):
logger.info('Restoring data')
self.remote.purge_restore_progress(
self.test_config.restore_settings.backup_storage,
self.test_config.restore_settings.backup_repo)
self.remote.restore_data(
self.test_config.restore_settings.backup_storage,
self.test_config.restore_settings.backup_repo,
)
def fts_collections_restore(self):
restore_mapping = None
collection_map = self.test_config.collection.collection_map
for target in self.target_iterator:
if not collection_map.get(target.bucket, {}).get(
"_default", {}).get("_default", {}).get('load', 0):
restore_mapping = \
"{0}._default._default={0}.scope-1.collection-1"\
.format(target.bucket)
logger.info('Restoring data')
self.remote.purge_restore_progress(
self.test_config.restore_settings.backup_storage,
self.test_config.restore_settings.backup_repo)
self.remote.restore_data(
self.test_config.restore_settings.backup_storage,
self.test_config.restore_settings.backup_repo,
map_data=restore_mapping)
def fts_cbimport(self):
logger.info('Restoring data into collections')
num_collections = self.test_config.jts_access_settings.collections_number
scope_prefix = self.test_config.jts_access_settings.scope_prefix
collection_prefix = self.test_config.jts_access_settings.collection_prefix
scope = self.test_config.jts_access_settings.scope_number
name_of_backup = self.test_config.restore_settings.backup_repo
self.remote.export_data(num_collections, collection_prefix,
scope_prefix, scope, name_of_backup)
def restore_local(self):
logger.info('Restoring data')
local.extract_cb(filename='couchbase.rpm')
local.purge_restore_progress(
self.cluster_spec,
archive=self.test_config.restore_settings.backup_storage,
repo=self.test_config.restore_settings.backup_repo)
local.cbbackupmgr_restore(
master_node=self.master_node,
cluster_spec=self.cluster_spec,
threads=self.test_config.restore_settings.threads,
archive=self.test_config.restore_settings.backup_storage,
repo=self.test_config.restore_settings.backup_repo,
include_data=self.test_config.backup_settings.include_data,
map_data=self.test_config.restore_settings.map_data,
use_tls=self.test_config.restore_settings.use_tls)
def load_tpcds_json_data(self):
logger.info('Importing data')
if self.test_config.collection.collection_map is not None:
cm = self.test_config.collection.collection_map
for bucket in self.test_config.buckets:
for scope in cm[bucket]:
for collection in cm[bucket][scope]:
if cm[bucket][scope][collection]['load'] == 1:
self.remote.load_tpcds_data_json_collection(
self.test_config.import_settings.import_file,
bucket, scope, collection, self.test_config.
import_settings.docs_per_collections)
else:
for bucket in self.test_config.buckets:
self.remote.load_tpcds_data_json(
self.test_config.import_settings.import_file,
bucket,
)
def compact_bucket(self, wait: bool = True):
for target in self.target_iterator:
self.rest.trigger_bucket_compaction(target.node, target.bucket)
if wait:
for target in self.target_iterator:
self.monitor.monitor_task(target.node, 'bucket_compaction')
def wait_for_persistence(self):
for target in self.target_iterator:
self.monitor.monitor_disk_queues(target.node, target.bucket)
self.monitor.monitor_dcp_queues(target.node, target.bucket)
self.monitor.monitor_replica_count(target.node, target.bucket)
def wait_for_indexing(self):
if self.test_config.index_settings.statements:
for server in self.index_nodes:
self.monitor.monitor_indexing(server)
def check_num_items(self,
bucket_items: dict = None,
max_retry: int = None):
if bucket_items:
for target in self.target_iterator:
num_items = bucket_items.get(target.bucket, None)
if num_items:
num_items = num_items * (
1 + self.test_config.bucket.replica_number)
self.monitor.monitor_num_items(target.node,
target.bucket,
num_items,
max_retry=max_retry)
elif getattr(self.test_config.load_settings, 'collections', None):
for target in self.target_iterator:
num_load_targets = 0
target_scope_collections = self.test_config.load_settings.collections[
target.bucket]
for scope in target_scope_collections.keys():
for collection in target_scope_collections[scope].keys():
if target_scope_collections[scope][collection][
'load'] == 1:
num_load_targets += 1
num_items = \
(self.test_config.load_settings.items // num_load_targets) * \
num_load_targets * \
(1 + self.test_config.bucket.replica_number)
self.monitor.monitor_num_items(target.node,
target.bucket,
num_items,
max_retry=max_retry)
else:
num_items = self.test_config.load_settings.items * (
1 + self.test_config.bucket.replica_number)
for target in self.target_iterator:
self.monitor.monitor_num_items(target.node,
target.bucket,
num_items,
max_retry=max_retry)
def reset_kv_stats(self):
master_node = next(self.cluster_spec.masters)
if self.test_config.cluster.enable_n2n_encryption:
local.get_cbstats(self.master_node, 11210, "reset",
self.cluster_spec)
else:
for bucket in self.test_config.buckets:
for server in self.rest.get_server_list(master_node, bucket):
port = self.rest.get_memcached_port(server)
self.memcached.reset_stats(server, port, bucket)
def create_indexes(self):
logger.info('Creating and building indexes')
if not self.test_config.index_settings.couchbase_fts_index_name:
create_statements = []
build_statements = []
for statement in self.test_config.index_settings.statements:
check_stmt = statement.replace(" ", "").upper()
if 'CREATEINDEX' in check_stmt \
or 'CREATEPRIMARYINDEX' in check_stmt:
create_statements.append(statement)
elif 'BUILDINDEX' in check_stmt:
build_statements.append(statement)
for statement in create_statements:
logger.info('Creating index: ' + statement)
self.rest.exec_n1ql_statement(self.query_nodes[0], statement)
cont = False
while not cont:
building = 0
index_status = self.rest.get_index_status(
self.index_nodes[0])
index_list = index_status['status']
for index in index_list:
if index['status'] != "Ready" and index[
'status'] != "Created":
building += 1
if building < 10:
cont = True
else:
time.sleep(10)
for statement in build_statements:
logger.info('Building index: ' + statement)
self.rest.exec_n1ql_statement(self.query_nodes[0], statement)
cont = False
while not cont:
building = 0
index_status = self.rest.get_index_status(
self.index_nodes[0])
index_list = index_status['status']
for index in index_list:
if index['status'] != "Ready" and index[
'status'] != "Created":
building += 1
if building < 10:
cont = True
else:
time.sleep(10)
logger.info('Index Create and Build Complete')
else:
self.create_fts_index_n1ql()
def create_fts_index_n1ql(self):
logger.info("Creating FTS index")
definition = read_json(
self.test_config.index_settings.couchbase_fts_index_configfile)
bucket_name = self.test_config.buckets[0]
definition.update(
{'name': self.test_config.index_settings.couchbase_fts_index_name})
if self.test_config.collection.collection_map:
collection_map = self.test_config.collection.collection_map
definition["params"]["doc_config"][
"mode"] = "scope.collection.type_field"
scope_name = list(collection_map[bucket_name].keys())[1:][0]
collection_name = list(
collection_map[bucket_name][scope_name].keys())[0]
ind_type_mapping = \
copy.deepcopy(definition["params"]["mapping"]["default_mapping"])
definition["params"]["mapping"]["default_mapping"][
"enabled"] = False
new_type_mapping_name = "{}.{}".format(scope_name, collection_name)
definition["params"]["mapping"]["types"] = {
new_type_mapping_name: ind_type_mapping
}
logger.info('Index definition: {}'.format(pretty_dict(definition)))
self.rest.create_fts_index(
self.fts_nodes[0],
self.test_config.index_settings.couchbase_fts_index_name,
definition)
self.monitor.monitor_fts_indexing_queue(
self.fts_nodes[0],
self.test_config.index_settings.couchbase_fts_index_name,
int(self.test_config.access_settings.items))
def create_functions(self):
logger.info('Creating n1ql functions')
for statement in self.test_config.n1ql_function_settings.statements:
self.rest.exec_n1ql_statement(self.query_nodes[0], statement)
def sleep(self):
access_settings = self.test_config.access_settings
logger.info('Running phase for {} seconds'.format(
access_settings.time))
time.sleep(access_settings.time)
def run_phase(self,
phase: str,
task: Callable,
settings: PhaseSettings,
target_iterator: Iterable,
timer: int = None,
wait: bool = True):
logger.info('Running {}: {}'.format(phase, pretty_dict(settings)))
self.worker_manager.run_tasks(task, settings, target_iterator, timer)
if wait:
self.worker_manager.wait_for_workers()
def load(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.load_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('load phase', task, settings, target_iterator)
def hot_load(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.hot_load_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('hot load phase', task, settings, target_iterator)
def xattr_load(self,
task: Callable = spring_task,
target_iterator: Iterable = None):
if target_iterator is None:
target_iterator = self.target_iterator
settings = self.test_config.xattr_load_settings
self.run_phase('xattr phase', task, settings, target_iterator)
def access(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.access_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('access phase',
task,
settings,
target_iterator,
timer=settings.time)
def access_bg(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.access_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('background access phase',
task,
settings,
target_iterator,
timer=settings.time,
wait=False)
def report_kpi(self, *args, **kwargs):
if self.test_config.stats_settings.enabled:
self._report_kpi(*args, **kwargs)
def _report_kpi(self, *args, **kwargs):
pass
def _measure_curr_ops(self) -> int:
ops = 0
for bucket in self.test_config.buckets:
for server in self.rest.get_active_nodes_by_role(
self.master_node, "kv"):
port = self.rest.get_memcached_port(server)
stats = self.memcached.get_stats(server, port, bucket)
for stat in 'cmd_get', 'cmd_set':
ops += int(stats[stat])
return ops
def _measure_disk_ops(self):
ret_stats = dict()
for bucket in self.test_config.buckets:
for server in self.rest.get_active_nodes_by_role(
self.master_node, "kv"):
ret_stats[server] = dict()
port = self.rest.get_memcached_port(server)
stats = self.memcached.get_stats(server, port, bucket)
ret_stats[server]["get_ops"] = int(stats['ep_bg_fetched'])
sets = \
int(stats['vb_active_ops_create']) \
+ int(stats['vb_replica_ops_create']) \
+ int(stats['vb_pending_ops_create']) \
+ int(stats['vb_active_ops_update']) \
+ int(stats['vb_replica_ops_update']) \
+ int(stats['vb_pending_ops_update'])
ret_stats[server]["set_ops"] = sets
return ret_stats
class PerfTest:
COLLECTORS = {}
ROOT_CERTIFICATE = 'root.pem'
def __init__(self,
cluster_spec: ClusterSpec,
test_config: TestConfig,
verbose: bool):
self.cluster_spec = cluster_spec
self.test_config = test_config
self.target_iterator = TargetIterator(cluster_spec, test_config)
self.cluster = ClusterManager(cluster_spec, test_config)
self.memcached = MemcachedHelper(test_config)
self.monitor = Monitor(cluster_spec, test_config, verbose)
self.rest = RestHelper(cluster_spec)
self.remote = RemoteHelper(cluster_spec, verbose)
self.profiler = Profiler(cluster_spec, test_config)
self.master_node = next(cluster_spec.masters)
self.build = self.rest.get_version(self.master_node)
self.metrics = MetricHelper(self)
self.reporter = ShowFastReporter(cluster_spec, test_config, self.build)
self.cbmonitor_snapshots = []
self.cbmonitor_clusters = []
if self.test_config.test_case.use_workers:
self.worker_manager = WorkerManager(cluster_spec, test_config,
verbose)
def __enter__(self):
return self
def __exit__(self, exc_type, exc_val, exc_tb):
failure = self.debug()
self.tear_down()
if exc_type == KeyboardInterrupt:
logger.warn('The test was interrupted')
return True
if failure:
logger.interrupt(failure)
@property
def query_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'n1ql')
@property
def index_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'index')
@property
def fts_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'fts')
@property
def analytics_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'cbas')
@property
def eventing_nodes(self) -> List[str]:
return self.rest.get_active_nodes_by_role(self.master_node, 'eventing')
def tear_down(self):
if self.test_config.test_case.use_workers:
self.worker_manager.download_celery_logs()
self.worker_manager.terminate()
if self.test_config.cluster.online_cores:
self.remote.enable_cpu()
if self.test_config.cluster.kernel_mem_limit:
self.collect_logs()
self.cluster.reset_memory_settings()
def collect_logs(self):
self.remote.collect_info()
for hostname in self.cluster_spec.servers:
for fname in glob.glob('{}/*.zip'.format(hostname)):
shutil.move(fname, '{}.zip'.format(hostname))
def reset_memory_settings(self):
if self.test_config.cluster.kernel_mem_limit:
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.reset_memory_settings(host_string=server)
self.monitor.wait_for_servers()
def debug(self) -> str:
failure = self.check_core_dumps()
failure = self.check_rebalance() or failure
return self.check_failover() or failure
def download_certificate(self):
cert = self.rest.get_certificate(self.master_node)
with open(self.ROOT_CERTIFICATE, 'w') as fh:
fh.write(cert)
def check_rebalance(self) -> str:
for master in self.cluster_spec.masters:
if self.rest.is_not_balanced(master):
return 'The cluster is not balanced'
def check_failover(self) -> Optional[str]:
if hasattr(self, 'rebalance_settings'):
if self.rebalance_settings.failover or \
self.rebalance_settings.graceful_failover:
return
for master in self.cluster_spec.masters:
num_failovers = self.rest.get_failover_counter(master)
if num_failovers:
return 'Failover happened {} time(s)'.format(num_failovers)
def check_core_dumps(self) -> str:
dumps_per_host = self.remote.detect_core_dumps()
core_dumps = {
host: dumps for host, dumps in dumps_per_host.items() if dumps
}
if core_dumps:
return pretty_dict(core_dumps)
def restore(self):
logger.info('Restoring data')
self.remote.restore_data(
self.test_config.restore_settings.backup_storage,
self.test_config.restore_settings.backup_repo,
)
def restore_local(self):
logger.info('Restoring data')
local.extract_cb(filename='couchbase.rpm')
local.cbbackupmgr_restore(
master_node=self.master_node,
cluster_spec=self.cluster_spec,
threads=self.test_config.restore_settings.threads,
archive=self.test_config.restore_settings.backup_storage,
repo=self.test_config.restore_settings.backup_repo,
)
def import_data(self):
logger.info('Importing data')
for bucket in self.test_config.buckets:
self.remote.import_data(
self.test_config.restore_settings.import_file, bucket,
)
def compact_bucket(self, wait: bool = True):
for target in self.target_iterator:
self.rest.trigger_bucket_compaction(target.node, target.bucket)
if wait:
for target in self.target_iterator:
self.monitor.monitor_task(target.node, 'bucket_compaction')
def wait_for_persistence(self):
for target in self.target_iterator:
self.monitor.monitor_disk_queues(target.node, target.bucket)
self.monitor.monitor_dcp_queues(target.node, target.bucket)
def wait_for_indexing(self):
if self.test_config.index_settings.statements:
for server in self.index_nodes:
self.monitor.monitor_indexing(server)
def check_num_items(self):
num_items = self.test_config.load_settings.items * (
1 + self.test_config.bucket.replica_number
)
for target in self.target_iterator:
self.monitor.monitor_num_items(target.node, target.bucket,
num_items)
def reset_kv_stats(self):
master_node = next(self.cluster_spec.masters)
for bucket in self.test_config.buckets:
for server in self.rest.get_server_list(master_node, bucket):
port = self.rest.get_memcached_port(server)
self.memcached.reset_stats(server, port, bucket)
def create_indexes(self):
logger.info('Creating and building indexes')
for statement in self.test_config.index_settings.statements:
self.rest.exec_n1ql_statement(self.query_nodes[0], statement)
def create_functions(self):
logger.info('Creating n1ql functions')
for statement in self.test_config.n1ql_function_settings.statements:
self.rest.exec_n1ql_statement(self.query_nodes[0], statement)
def sleep(self):
access_settings = self.test_config.access_settings
logger.info('Running phase for {} seconds'.format(access_settings.time))
time.sleep(access_settings.time)
def run_phase(self,
phase: str,
task: Callable, settings: PhaseSettings,
target_iterator: Iterable,
timer: int = None,
wait: bool = True):
logger.info('Running {}: {}'.format(phase, pretty_dict(settings)))
self.worker_manager.run_tasks(task, settings, target_iterator, timer)
if wait:
self.worker_manager.wait_for_workers()
def load(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.load_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('load phase',
task, settings, target_iterator)
def hot_load(self,
task: Callable = spring_task):
settings = self.test_config.hot_load_settings
self.run_phase('hot load phase',
task, settings, self.target_iterator)
def xattr_load(self,
task: Callable = spring_task,
target_iterator: Iterable = None):
if target_iterator is None:
target_iterator = self.target_iterator
settings = self.test_config.xattr_load_settings
self.run_phase('xattr phase',
task, settings, target_iterator)
def access(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.access_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('access phase',
task, settings, target_iterator,
timer=settings.time)
def access_bg(self,
task: Callable = spring_task,
settings: PhaseSettings = None,
target_iterator: Iterable = None):
if settings is None:
settings = self.test_config.access_settings
if target_iterator is None:
target_iterator = self.target_iterator
self.run_phase('background access phase',
task, settings, target_iterator,
timer=settings.time, wait=False)
def report_kpi(self, *args, **kwargs):
if self.test_config.stats_settings.enabled:
self._report_kpi(*args, **kwargs)
def _report_kpi(self, *args, **kwargs):
pass
class Profiler:
DEBUG_PORTS = {
'fts': 8094,
'index': 9102,
'kv': 9998, # goxdcr
'n1ql': 8093,
}
ENDPOINTS = {
'cpu': 'http://127.0.0.1:{}/debug/pprof/profile',
'heap': 'http://127.0.0.1:{}/debug/pprof/heap',
'goroutine': 'http://127.0.0.1:{}/debug/pprof/goroutine?debug=2',
}
def __init__(self, cluster_spec: ClusterSpec, test_config: TestConfig):
self.test_config = test_config
self.rest = RestHelper(cluster_spec)
self.master_node = next(cluster_spec.masters)
self.ssh_username, self.ssh_password = cluster_spec.ssh_credentials
def new_tunnel(self, host: str, port: int) -> SSHTunnelForwarder:
return SSHTunnelForwarder(
ssh_address_or_host=host,
ssh_username=self.ssh_username,
ssh_password=self.ssh_password,
remote_bind_address=('127.0.0.1', port),
)
def save(self, host: str, service: str, profile: str, content: bytes):
fname = '{}_{}_{}_{}.pprof'.format(host, service, profile, uhex()[:6])
with open(fname, 'wb') as fh:
fh.write(content)
def profile(self, host: str, service: str, profile: str):
logger.info('Collecting {} profile on {}'.format(profile, host))
endpoint = self.ENDPOINTS[profile]
port = self.DEBUG_PORTS[service]
with self.new_tunnel(host, port) as tunnel:
url = endpoint.format(tunnel.local_bind_port)
response = requests.get(url=url, auth=self.rest.auth)
self.save(host, service, profile, response.content)
def timer(self, **kwargs):
timer = Timer(
function=self.profile,
interval=self.test_config.profiling_settings.interval,
num_runs=self.test_config.profiling_settings.num_profiles,
kwargs=kwargs,
)
timer.start()
def schedule(self):
for service in self.test_config.profiling_settings.services:
logger.info('Scheduling profiling of "{}" services'.format(service))
for server in self.rest.get_active_nodes_by_role(self.master_node,
role=service):
for profile in self.test_config.profiling_settings.profiles:
self.timer(host=server, service=service, profile=profile)
class ClusterManager:
def __init__(self, cluster_spec: ClusterSpec, test_config: TestConfig,
verbose: bool = False):
self.cluster_spec = cluster_spec
self.test_config = test_config
self.rest = RestHelper(cluster_spec)
self.remote = RemoteHelper(cluster_spec, verbose)
self.monitor = Monitor(cluster_spec, test_config, verbose)
self.memcached = MemcachedHelper(test_config)
self.master_node = next(self.cluster_spec.masters)
self.initial_nodes = test_config.cluster.initial_nodes
def is_compatible(self, min_release: str) -> bool:
for master in self.cluster_spec.masters:
version = self.rest.get_version(master)
return version >= min_release
def set_data_path(self):
for server in self.cluster_spec.servers:
self.remote.change_owner(server, self.cluster_spec.data_path)
self.rest.set_data_path(server, self.cluster_spec.data_path)
def set_index_path(self):
for server in self.cluster_spec.servers:
self.remote.change_owner(server, self.cluster_spec.index_path)
self.rest.set_index_path(server, self.cluster_spec.index_path)
def set_analytics_path(self):
paths = []
for path in self.cluster_spec.analytics_paths:
for i in range(self.test_config.analytics_settings.num_io_devices):
io_device = '{}/dev{}'.format(path, i)
paths.append(io_device)
for server in self.cluster_spec.servers_by_role('cbas'):
for path in self.cluster_spec.analytics_paths:
self.remote.change_owner(server, path)
self.rest.set_analytics_paths(server, paths)
def rename(self):
for server in self.cluster_spec.servers:
self.rest.rename(server)
def set_auth(self):
for server in self.cluster_spec.servers:
self.rest.set_auth(server)
def set_mem_quotas(self):
for master in self.cluster_spec.masters:
self.rest.set_mem_quota(master,
self.test_config.cluster.mem_quota)
self.rest.set_index_mem_quota(master,
self.test_config.cluster.index_mem_quota)
if self.test_config.cluster.fts_index_mem_quota:
self.rest.set_fts_index_mem_quota(master,
self.test_config.cluster.fts_index_mem_quota)
if self.test_config.cluster.analytics_mem_quota:
self.rest.set_analytics_mem_quota(master,
self.test_config.cluster.analytics_mem_quota)
if self.test_config.cluster.eventing_mem_quota:
self.rest.set_eventing_mem_quota(master,
self.test_config.cluster.eventing_mem_quota)
def set_query_settings(self):
logger.info('Setting query settings')
query_nodes = self.cluster_spec.servers_by_role('n1ql')
if query_nodes:
settings = self.test_config.n1ql_settings.cbq_settings
if settings:
self.rest.set_query_settings(query_nodes[0], settings)
settings = self.rest.get_query_settings(query_nodes[0])
settings = pretty_dict(settings)
logger.info('Query settings: {}'.format(settings))
def set_index_settings(self):
logger.info('Setting index settings')
index_nodes = self.cluster_spec.servers_by_role('index')
if index_nodes:
settings = self.test_config.gsi_settings.settings
if settings:
self.rest.set_index_settings(index_nodes[0], settings)
settings = self.rest.get_index_settings(index_nodes[0])
settings = pretty_dict(settings)
logger.info('Index settings: {}'.format(settings))
def set_services(self):
if not self.is_compatible(min_release='4.0.0'):
return
for master in self.cluster_spec.masters:
roles = self.cluster_spec.roles[master]
self.rest.set_services(master, roles)
def add_nodes(self):
for (_, servers), initial_nodes in zip(self.cluster_spec.clusters,
self.initial_nodes):
if initial_nodes < 2: # Single-node cluster
continue
master = servers[0]
for node in servers[1:initial_nodes]:
roles = self.cluster_spec.roles[node]
self.rest.add_node(master, node, roles)
def rebalance(self):
for (_, servers), initial_nodes in zip(self.cluster_spec.clusters,
self.initial_nodes):
master = servers[0]
known_nodes = servers[:initial_nodes]
ejected_nodes = []
self.rest.rebalance(master, known_nodes, ejected_nodes)
self.monitor.monitor_rebalance(master)
self.wait_until_healthy()
def increase_bucket_limit(self, num_buckets: int):
for master in self.cluster_spec.masters:
self.rest.increase_bucket_limit(master, num_buckets)
def flush_buckets(self):
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.flush_bucket(host=master,
bucket=bucket_name)
def delete_buckets(self):
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.delete_bucket(host=master,
name=bucket_name)
def create_buckets(self):
mem_quota = self.test_config.cluster.mem_quota
if self.test_config.cluster.num_buckets > 7:
self.increase_bucket_limit(self.test_config.cluster.num_buckets + 3)
if self.test_config.cluster.eventing_metadata_bucket_mem_quota:
mem_quota -= (self.test_config.cluster.eventing_metadata_bucket_mem_quota +
self.test_config.cluster.eventing_bucket_mem_quota)
per_bucket_quota = mem_quota // self.test_config.cluster.num_buckets
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.buckets:
self.rest.create_bucket(
host=master,
name=bucket_name,
ram_quota=per_bucket_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.eviction_policy,
bucket_type=self.test_config.bucket.bucket_type,
conflict_resolution_type=self.test_config.bucket.conflict_resolution_type,
compression_mode=self.test_config.bucket.compression_mode,
)
def create_eventing_buckets(self):
if not self.test_config.cluster.eventing_bucket_mem_quota:
return
per_bucket_quota = \
self.test_config.cluster.eventing_bucket_mem_quota \
// self.test_config.cluster.eventing_buckets
for master in self.cluster_spec.masters:
for bucket_name in self.test_config.eventing_buckets:
self.rest.create_bucket(
host=master,
name=bucket_name,
ram_quota=per_bucket_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.eviction_policy,
bucket_type=self.test_config.bucket.bucket_type,
conflict_resolution_type=self.test_config.bucket.conflict_resolution_type,
)
def create_eventing_metadata_bucket(self):
if not self.test_config.cluster.eventing_metadata_bucket_mem_quota:
return
for master in self.cluster_spec.masters:
self.rest.create_bucket(
host=master,
name=self.test_config.cluster.EVENTING_METADATA_BUCKET_NAME,
ram_quota=self.test_config.cluster.eventing_metadata_bucket_mem_quota,
password=self.test_config.bucket.password,
replica_number=self.test_config.bucket.replica_number,
replica_index=self.test_config.bucket.replica_index,
eviction_policy=self.test_config.bucket.EVICTION_POLICY,
bucket_type=self.test_config.bucket.BUCKET_TYPE,
)
def configure_auto_compaction(self):
compaction_settings = self.test_config.compaction
for master in self.cluster_spec.masters:
self.rest.configure_auto_compaction(master, compaction_settings)
settings = self.rest.get_auto_compaction_settings(master)
logger.info('Auto-compaction settings: {}'
.format(pretty_dict(settings)))
def configure_internal_settings(self):
internal_settings = self.test_config.internal_settings
for master in self.cluster_spec.masters:
for parameter, value in internal_settings.items():
self.rest.set_internal_settings(master,
{parameter: maybe_atoi(value)})
def configure_xdcr_settings(self):
xdcr_cluster_settings = self.test_config.xdcr_cluster_settings
for master in self.cluster_spec.masters:
for parameter, value in xdcr_cluster_settings.items():
self.rest.set_xdcr_cluster_settings(master,
{parameter: maybe_atoi(value)})
def tweak_memory(self):
self.remote.reset_swap()
self.remote.drop_caches()
self.remote.set_swappiness()
self.remote.disable_thp()
def restart_with_alternative_num_vbuckets(self):
num_vbuckets = self.test_config.cluster.num_vbuckets
if num_vbuckets is not None:
self.remote.restart_with_alternative_num_vbuckets(num_vbuckets)
def restart_with_alternative_bucket_options(self):
"""Apply custom buckets settings.
Tune bucket settings (e.g., max_num_shards or max_num_auxio) using
"/diag/eval" and restart the entire cluster.
"""
if self.test_config.bucket_extras:
self.remote.enable_nonlocal_diag_eval()
cmd = 'ns_bucket:update_bucket_props("{}", ' \
'[{{extra_config_string, "{}={}"}}]).'
for option, value in self.test_config.bucket_extras.items():
logger.info('Changing {} to {}'.format(option, value))
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
diag_eval = cmd.format(bucket, option, value)
self.rest.run_diag_eval(master, diag_eval)
if self.test_config.bucket_extras:
self.remote.restart()
self.wait_until_healthy()
def tune_logging(self):
self.remote.tune_log_rotation()
self.remote.restart()
def enable_auto_failover(self):
failover_min = self.test_config.bucket.failover_min
failover_max = self.test_config.bucket.failover_max
for master in self.cluster_spec.masters:
self.rest.enable_auto_failover(master, failover_min, failover_max)
def wait_until_warmed_up(self):
if self.test_config.bucket.bucket_type in ('ephemeral', 'memcached'):
return
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
self.monitor.monitor_warmup(self.memcached, master, bucket)
def wait_until_healthy(self):
for master in self.cluster_spec.masters:
self.monitor.monitor_node_health(master)
for analytics_node in self.rest.get_active_nodes_by_role(master,
'cbas'):
self.monitor.monitor_analytics_node_active(analytics_node)
def gen_disabled_audit_events(self, master: str) -> List[str]:
curr_settings = self.rest.get_audit_settings(master)
curr_disabled = {str(event) for event in curr_settings['disabled']}
disabled = curr_disabled - self.test_config.audit_settings.extra_events
return list(disabled)
def enable_audit(self):
if not self.is_compatible(min_release='4.0.0') or \
self.rest.is_community(self.master_node):
return
if not self.test_config.audit_settings.enabled:
return
for master in self.cluster_spec.masters:
disabled = []
if self.test_config.audit_settings.extra_events:
disabled = self.gen_disabled_audit_events(master)
self.rest.enable_audit(master, disabled)
def generate_ce_roles(self) -> List[str]:
return ['admin']
def generate_ee_roles(self) -> List[str]:
existing_roles = {r['role']
for r in self.rest.get_rbac_roles(self.master_node)}
roles = []
for role in (
'bucket_admin',
'data_dcp_reader',
'data_monitoring',
'data_reader_writer',
'data_reader',
'data_writer',
'fts_admin',
'fts_searcher',
'query_delete',
'query_insert',
'query_select',
'query_update',
'views_admin',
):
if role in existing_roles:
roles.append(role + '[{bucket}]')
return roles
def delete_rbac_users(self):
if not self.is_compatible(min_release='5.0'):
return
for master in self.cluster_spec.masters:
for bucket in self.test_config.buckets:
self.rest.delete_rbac_user(
host=master,
bucket=bucket
)
def add_rbac_users(self):
if not self.rest.supports_rbac(self.master_node):
logger.info('RBAC not supported - skipping adding RBAC users')
return
if self.rest.is_community(self.master_node):
roles = self.generate_ce_roles()
else:
roles = self.generate_ee_roles()
for master in self.cluster_spec.masters:
admin_user, admin_password = self.cluster_spec.rest_credentials
self.rest.add_rbac_user(
host=master,
user=admin_user,
password=admin_password,
roles=['admin'],
)
for bucket in self.test_config.buckets:
bucket_roles = [role.format(bucket=bucket) for role in roles]
bucket_roles.append("admin")
self.rest.add_rbac_user(
host=master,
user=bucket, # Backward compatibility
password=self.test_config.bucket.password,
roles=bucket_roles,
)
def throttle_cpu(self):
if self.remote.os == 'Cygwin':
return
self.remote.enable_cpu()
if self.test_config.cluster.online_cores:
self.remote.disable_cpu(self.test_config.cluster.online_cores)
def tune_memory_settings(self):
kernel_memory = self.test_config.cluster.kernel_mem_limit
if kernel_memory:
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.tune_memory_settings(host_string=server,
size=kernel_memory)
self.monitor.wait_for_servers()
def reset_memory_settings(self):
for service in self.test_config.cluster.kernel_mem_limit_services:
for server in self.cluster_spec.servers_by_role(service):
self.remote.reset_memory_settings(host_string=server)
self.monitor.wait_for_servers()
def flush_iptables(self):
self.remote.flush_iptables()
def clear_login_history(self):
self.remote.clear_wtmp()
def disable_wan(self):
self.remote.disable_wan()
def enable_ipv6(self):
if self.test_config.cluster.ipv6:
self.remote.enable_ipv6()
def set_x509_certificates(self):
logger.info('Setting x509 settings')
if self.test_config.access_settings.ssl_mode == "auth":
self.remote.setup_x509()
for host in self.cluster_spec.servers:
self.rest.upload_cluster_certificate(host)
for host in self.cluster_spec.servers:
self.rest.reload_cluster_certificate(host)
self.rest.enable_certificate_auth(host)