def version():
release = __version__
# noinspection PyBroadException
try:
if git.is_working_copy(io.normalize_path("%s/.." % paths.rally_root())):
revision = git.head_revision(paths.rally_root())
return "%s (git revision: %s)" % (release, revision.strip())
except BaseException:
pass
# cannot determine head revision so user has probably installed Rally via pip instead of git clone
return release
def revision():
"""
:return: The current git revision if Rally is installed in development mode or ``None``.
"""
# noinspection PyBroadException
try:
if git.is_working_copy(io.normalize_path("%s/.." % paths.rally_root())):
raw_revision = git.head_revision(paths.rally_root())
return raw_revision.strip()
except BaseException:
pass
return None
def version():
"""
:return: The release version string and an optional suffix for the current git revision if Rally is installed in development mode.
"""
release = __version__
# noinspection PyBroadException
try:
if git.is_working_copy(io.normalize_path("%s/.." %
paths.rally_root())):
revision = git.head_revision(paths.rally_root())
return "%s (git revision: %s)" % (release, revision.strip())
except BaseException:
pass
# cannot determine head revision so user has probably installed Rally via pip instead of git clone
return release
def receiveMsg_StartNodes(self, msg, sender):
try:
self.host = msg.ip
if msg.external:
self.logger.info(
"Connecting to externally provisioned nodes on [%s].",
msg.ip)
else:
self.logger.info("Starting node(s) %s on [%s].", msg.node_ids,
msg.ip)
# Load node-specific configuration
self.config = config.auto_load_local_config(
msg.cfg,
additional_sections=[
# only copy the relevant bits
"track",
"mechanic",
"client",
# allow metrics store to extract race meta-data
"race",
"source"
])
# set root path (normally done by the main entry point)
self.config.add(config.Scope.application, "node", "rally.root",
paths.rally_root())
if not msg.external:
self.config.add(config.Scope.benchmark, "provisioning",
"node.ip", msg.ip)
# we need to override the port with the value that the user has specified instead of using the default value (39200)
self.config.add(config.Scope.benchmark, "provisioning",
"node.http.port", msg.port)
self.config.add(config.Scope.benchmark, "provisioning",
"node.ids", msg.node_ids)
cls = metrics.metrics_store_class(self.config)
self.metrics_store = cls(self.config)
self.metrics_store.open(ctx=msg.open_metrics_context)
# avoid follow-up errors in case we receive an unexpected ActorExitRequest due to an early failure in a parent actor.
self.metrics_store.lap = 0
self.mechanic = create(self.config, self.metrics_store,
msg.all_node_ips, msg.cluster_settings,
msg.sources, msg.build, msg.distribution,
msg.external, msg.docker)
nodes = self.mechanic.start_engine()
self.running = True
self.send(
getattr(msg, "reply_to", sender),
NodesStarted([NodeMetaInfo(node) for node in nodes],
self.metrics_store.meta_info))
except Exception:
self.logger.exception("Cannot process message [%s]", msg)
# avoid "can't pickle traceback objects"
import traceback
ex_type, ex_value, ex_traceback = sys.exc_info()
self.send(getattr(msg, "reply_to", sender),
actor.BenchmarkFailure(ex_value, traceback.format_exc()))
def main():
check_python_version()
log.install_default_log_config()
log.configure_logging()
logger = logging.getLogger(__name__)
start = time.time()
# Early init of console output so we start to show everything consistently.
console.init(quiet=False)
arg_parser = create_arg_parser()
args = arg_parser.parse_args()
console.init(quiet=args.quiet)
console.println(BANNER)
cfg = config.Config(config_name=args.configuration_name)
if not cfg.config_present():
cfg.install_default_config()
cfg.load_config(auto_upgrade=True)
cfg.add(config.Scope.application, "system", "time.start", datetime.datetime.utcnow())
# Local config per node
cfg.add(config.Scope.application, "node", "rally.root", paths.rally_root())
cfg.add(config.Scope.application, "node", "rally.cwd", os.getcwd())
logger.info("OS [%s]", str(platform.uname()))
logger.info("Python [%s]", str(sys.implementation))
logger.info("Rally version [%s]", version.version())
logger.debug("Command line arguments: %s", args)
# Configure networking
net.init()
if not args.offline:
probing_url = cfg.opts("system", "probing.url", default_value="https://github.com", mandatory=False)
if not net.has_internet_connection(probing_url):
console.warn("No Internet connection detected. Automatic download of track data sets etc. is disabled.", logger=logger)
cfg.add(config.Scope.applicationOverride, "system", "offline.mode", True)
else:
logger.info("Detected a working Internet connection.")
result = dispatch_sub_command(arg_parser, args, cfg)
end = time.time()
if result == ExitStatus.SUCCESSFUL:
console.println("")
console.info("SUCCESS (took %d seconds)" % (end - start), overline="-", underline="-")
elif result == ExitStatus.INTERRUPTED:
console.println("")
console.info("ABORTED (took %d seconds)" % (end - start), overline="-", underline="-")
sys.exit(130)
elif result == ExitStatus.ERROR:
console.println("")
console.info("FAILURE (took %d seconds)" % (end - start), overline="-", underline="-")
sys.exit(64)
def receiveMsg_StartNodes(self, msg, sender):
try:
self.host = msg.ip
if msg.external:
self.logger.info("Connecting to externally provisioned nodes on [%s].", msg.ip)
else:
self.logger.info("Starting node(s) %s on [%s].", msg.node_ids, msg.ip)
# Load node-specific configuration
cfg = config.auto_load_local_config(
msg.cfg,
additional_sections=[
# only copy the relevant bits
"track",
"mechanic",
"client",
"telemetry",
# allow metrics store to extract race meta-data
"race",
"source",
],
)
# set root path (normally done by the main entry point)
cfg.add(config.Scope.application, "node", "rally.root", paths.rally_root())
if not msg.external:
cfg.add(config.Scope.benchmark, "provisioning", "node.ids", msg.node_ids)
cls = metrics.metrics_store_class(cfg)
metrics_store = cls(cfg)
metrics_store.open(ctx=msg.open_metrics_context)
# avoid follow-up errors in case we receive an unexpected ActorExitRequest due to an early failure in a parent actor.
self.mechanic = create(
cfg,
metrics_store,
msg.ip,
msg.port,
msg.all_node_ips,
msg.all_node_ids,
msg.sources,
msg.distribution,
msg.external,
msg.docker,
)
self.mechanic.start_engine()
self.wakeupAfter(METRIC_FLUSH_INTERVAL_SECONDS)
self.send(getattr(msg, "reply_to", sender), NodesStarted())
except Exception:
self.logger.exception("Cannot process message [%s]", msg)
# avoid "can't pickle traceback objects"
_, ex_value, _ = sys.exc_info()
self.send(getattr(msg, "reply_to", sender), actor.BenchmarkFailure(ex_value, traceback.format_exc()))
def main():
check_python_version()
log.install_default_log_config()
log.configure_logging()
logger = logging.getLogger(__name__)
start = time.time()
# Early init of console output so we start to show everything consistently.
console.init(quiet=False)
arg_parser = create_arg_parser()
args = arg_parser.parse_args()
console.init(quiet=args.quiet)
console.println(BANNER)
cfg = config.Config(config_name=args.configuration_name)
sub_command = derive_sub_command(args, cfg)
ensure_configuration_present(cfg, args, sub_command)
if args.effective_start_date:
cfg.add(config.Scope.application, "system", "time.start",
args.effective_start_date)
cfg.add(config.Scope.application, "system", "time.start.user_provided",
True)
else:
cfg.add(config.Scope.application, "system", "time.start",
datetime.datetime.utcnow())
cfg.add(config.Scope.application, "system", "time.start.user_provided",
False)
cfg.add(config.Scope.applicationOverride, "system", "trial.id",
str(uuid.uuid4()))
cfg.add(config.Scope.applicationOverride, "system", "quiet.mode",
args.quiet)
cfg.add(config.Scope.applicationOverride, "system", "offline.mode",
args.offline)
# Local config per node
cfg.add(config.Scope.application, "node", "rally.root", paths.rally_root())
cfg.add(config.Scope.application, "node", "rally.cwd", os.getcwd())
cfg.add(config.Scope.applicationOverride, "mechanic", "source.revision",
args.revision)
if args.distribution_version:
cfg.add(config.Scope.applicationOverride, "mechanic",
"distribution.version", args.distribution_version)
cfg.add(config.Scope.applicationOverride, "mechanic",
"distribution.repository", args.distribution_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "car.names",
opts.csv_to_list(args.car))
if args.team_path:
cfg.add(config.Scope.applicationOverride, "mechanic", "team.path",
os.path.abspath(io.normalize_path(args.team_path)))
cfg.add(config.Scope.applicationOverride, "mechanic",
"repository.name", None)
else:
cfg.add(config.Scope.applicationOverride, "mechanic",
"repository.name", args.team_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "car.plugins",
opts.csv_to_list(args.elasticsearch_plugins))
cfg.add(config.Scope.applicationOverride, "mechanic", "car.params",
opts.to_dict(args.car_params))
cfg.add(config.Scope.applicationOverride, "mechanic", "plugin.params",
opts.to_dict(args.plugin_params))
if args.keep_cluster_running:
cfg.add(config.Scope.applicationOverride, "mechanic", "keep.running",
True)
# force-preserve the cluster nodes.
cfg.add(config.Scope.applicationOverride, "mechanic",
"preserve.install", True)
else:
cfg.add(config.Scope.applicationOverride, "mechanic", "keep.running",
False)
cfg.add(config.Scope.applicationOverride, "mechanic",
"preserve.install", convert.to_bool(args.preserve_install))
cfg.add(config.Scope.applicationOverride, "mechanic", "runtime.jdk",
args.runtime_jdk)
cfg.add(config.Scope.applicationOverride, "mechanic", "telemetry.devices",
opts.csv_to_list(args.telemetry))
cfg.add(config.Scope.applicationOverride, "mechanic", "telemetry.params",
opts.to_dict(args.telemetry_params))
cfg.add(config.Scope.applicationOverride, "race", "pipeline",
args.pipeline)
cfg.add(config.Scope.applicationOverride, "race", "laps", args.laps)
cfg.add(config.Scope.applicationOverride, "race", "user.tag",
args.user_tag)
# We can assume here that if a track-path is given, the user did not specify a repository either (although argparse sets it to
# its default value)
if args.track_path:
cfg.add(config.Scope.applicationOverride, "track", "track.path",
os.path.abspath(io.normalize_path(args.track_path)))
cfg.add(config.Scope.applicationOverride, "track", "repository.name",
None)
if args.track:
# stay as close as possible to argparse errors although we have a custom validation.
arg_parser.error(
"argument --track not allowed with argument --track-path")
# cfg.add(config.Scope.applicationOverride, "track", "track.name", None)
else:
# cfg.add(config.Scope.applicationOverride, "track", "track.path", None)
cfg.add(config.Scope.applicationOverride, "track", "repository.name",
args.track_repository)
# set the default programmatically because we need to determine whether the user has provided a value
chosen_track = args.track if args.track else "geonames"
cfg.add(config.Scope.applicationOverride, "track", "track.name",
chosen_track)
cfg.add(config.Scope.applicationOverride, "track", "params",
opts.to_dict(args.track_params))
cfg.add(config.Scope.applicationOverride, "track", "challenge.name",
args.challenge)
cfg.add(config.Scope.applicationOverride, "track", "include.tasks",
opts.csv_to_list(args.include_tasks))
cfg.add(config.Scope.applicationOverride, "track", "test.mode.enabled",
args.test_mode)
cfg.add(config.Scope.applicationOverride, "reporting", "format",
args.report_format)
cfg.add(config.Scope.applicationOverride, "reporting", "values",
args.show_in_report)
cfg.add(config.Scope.applicationOverride, "reporting", "output.path",
args.report_file)
if sub_command == "compare":
cfg.add(config.Scope.applicationOverride, "reporting",
"baseline.timestamp", args.baseline)
cfg.add(config.Scope.applicationOverride, "reporting",
"contender.timestamp", args.contender)
if sub_command == "generate":
cfg.add(config.Scope.applicationOverride, "generator", "chart.type",
args.chart_type)
cfg.add(config.Scope.applicationOverride, "generator", "output.path",
args.output_path)
if args.chart_spec_path and (args.track or args.challenge or args.car
or args.node_count):
console.println(
"You need to specify either --chart-spec-path or --track, --challenge, --car and "
"--node-count but not both.")
exit(1)
if args.chart_spec_path:
cfg.add(config.Scope.applicationOverride, "generator",
"chart.spec.path", args.chart_spec_path)
else:
# other options are stored elsewhere already
cfg.add(config.Scope.applicationOverride, "generator",
"node.count", args.node_count)
cfg.add(config.Scope.applicationOverride, "driver", "profiling",
args.enable_driver_profiling)
cfg.add(config.Scope.applicationOverride, "driver", "on.error",
args.on_error)
cfg.add(config.Scope.applicationOverride, "driver", "load_driver_hosts",
opts.csv_to_list(args.load_driver_hosts))
if sub_command != "list":
# Also needed by mechanic (-> telemetry) - duplicate by module?
target_hosts = opts.TargetHosts(args.target_hosts)
cfg.add(config.Scope.applicationOverride, "client", "hosts",
target_hosts)
client_options = opts.ClientOptions(args.client_options,
target_hosts=target_hosts)
cfg.add(config.Scope.applicationOverride, "client", "options",
client_options)
if "timeout" not in client_options.default:
console.info(
"You did not provide an explicit timeout in the client options. Assuming default of 10 seconds."
)
if list(target_hosts.all_hosts) != list(
client_options.all_client_options):
console.println(
"--target-hosts and --client-options must define the same keys for multi cluster setups."
)
exit(1)
# split by component?
if sub_command == "list":
cfg.add(config.Scope.applicationOverride, "system",
"list.config.option", args.configuration)
cfg.add(config.Scope.applicationOverride, "system",
"list.races.max_results", args.limit)
logger.info("OS [%s]", str(os.uname()))
logger.info("Python [%s]", str(sys.implementation))
logger.info("Rally version [%s]", version.version())
logger.info("Command line arguments: %s", args)
# Configure networking
net.init()
if not args.offline:
if not net.has_internet_connection():
console.warn(
"No Internet connection detected. Automatic download of track data sets etc. is disabled.",
logger=logger)
cfg.add(config.Scope.applicationOverride, "system", "offline.mode",
True)
else:
logger.info("Detected a working Internet connection.")
success = dispatch_sub_command(cfg, sub_command)
end = time.time()
if success:
console.println("")
console.info("SUCCESS (took %d seconds)" % (end - start),
overline="-",
underline="-")
else:
console.println("")
console.info("FAILURE (took %d seconds)" % (end - start),
overline="-",
underline="-")
sys.exit(64)
def receiveMessage(self, msg, sender):
# at the moment, we implement all message handling blocking. This is not ideal but simple to get started with. Besides, the caller
# needs to block anyway. The only reason we implement mechanic as an actor is to distribute them.
# noinspection PyBroadException
try:
logger.debug(
"NodeMechanicActor#receiveMessage(msg = [%s] sender = [%s])" %
(str(type(msg)), str(sender)))
if isinstance(msg, StartNodes):
self.host = msg.ip
if msg.external:
logger.info(
"Connecting to externally provisioned nodes on [%s]." %
msg.ip)
else:
logger.info("Starting node(s) %s on [%s]." %
(msg.node_ids, msg.ip))
# Load node-specific configuration
self.config = config.auto_load_local_config(
msg.cfg,
additional_sections=[
# only copy the relevant bits
"track",
"mechanic",
"client",
# allow metrics store to extract race meta-data
"race",
"source"
])
# set root path (normally done by the main entry point)
self.config.add(config.Scope.application, "node", "rally.root",
paths.rally_root())
if not msg.external:
self.config.add(config.Scope.benchmark, "provisioning",
"node.ip", msg.ip)
# we need to override the port with the value that the user has specified instead of using the default value (39200)
self.config.add(config.Scope.benchmark, "provisioning",
"node.http.port", msg.port)
self.config.add(config.Scope.benchmark, "provisioning",
"node.ids", msg.node_ids)
self.metrics_store = metrics.InMemoryMetricsStore(self.config)
self.metrics_store.open(ctx=msg.open_metrics_context)
# avoid follow-up errors in case we receive an unexpected ActorExitRequest due to an early failure in a parent actor.
self.metrics_store.lap = 0
self.mechanic = create(self.config, self.metrics_store,
msg.all_node_ips, msg.cluster_settings,
msg.sources, msg.build,
msg.distribution, msg.external,
msg.docker)
nodes = self.mechanic.start_engine()
self.running = True
self.send(
sender,
NodesStarted([NodeMetaInfo(node) for node in nodes],
self.metrics_store.meta_info))
elif isinstance(msg, ApplyMetricsMetaInfo):
self.metrics_store.merge_meta_info(msg.meta_info)
self.send(sender, MetricsMetaInfoApplied())
elif isinstance(msg, ResetRelativeTime):
logger.info(
"Resetting relative time of system metrics store on host [%s]."
% self.host)
self.metrics_store.reset_relative_time()
elif isinstance(msg, OnBenchmarkStart):
self.metrics_store.lap = msg.lap
self.mechanic.on_benchmark_start()
self.send(sender, BenchmarkStarted())
elif isinstance(msg, OnBenchmarkStop):
self.mechanic.on_benchmark_stop()
# clear metrics store data to not send duplicate system metrics data
self.send(
sender,
BenchmarkStopped(
self.metrics_store.to_externalizable(clear=True)))
elif isinstance(msg, StopNodes):
logger.info("Stopping nodes %s." % self.mechanic.nodes)
self.mechanic.stop_engine()
self.send(sender,
NodesStopped(self.metrics_store.to_externalizable()))
# clear all state as the mechanic might get reused later
self.running = False
self.config = None
self.mechanic = None
self.metrics_store = None
elif isinstance(msg, thespian.actors.ActorExitRequest):
if self.running:
logger.info("Stopping nodes %s (due to ActorExitRequest)" %
self.mechanic.nodes)
self.mechanic.stop_engine()
self.running = False
except BaseException:
self.running = False
logger.exception("Cannot process message [%s]" % msg)
# avoid "can't pickle traceback objects"
import traceback
ex_type, ex_value, ex_traceback = sys.exc_info()
self.send(sender,
actor.BenchmarkFailure(ex_value, traceback.format_exc()))
def main():
check_python_version()
start = time.time()
# Early init of console output so we start to show everything consistently.
console.init(quiet=False)
# allow to see a thread-dump on SIGQUIT
faulthandler.register(signal.SIGQUIT, file=sys.stderr)
pre_configure_logging()
args = parse_args()
console.init(quiet=args.quiet)
console.println(BANNER)
cfg = config.Config(config_name=args.configuration_name)
sub_command = derive_sub_command(args, cfg)
ensure_configuration_present(cfg, args, sub_command)
if args.effective_start_date:
cfg.add(config.Scope.application, "system", "time.start",
args.effective_start_date)
cfg.add(config.Scope.application, "system", "time.start.user_provided",
True)
else:
cfg.add(config.Scope.application, "system", "time.start",
datetime.datetime.utcnow())
cfg.add(config.Scope.application, "system", "time.start.user_provided",
False)
cfg.add(config.Scope.applicationOverride, "system", "quiet.mode",
args.quiet)
# per node?
cfg.add(config.Scope.applicationOverride, "system", "offline.mode",
args.offline)
cfg.add(config.Scope.applicationOverride, "system", "logging.output",
args.logging)
# only temporary to ignore unknown actor messages
cfg.add(config.Scope.applicationOverride, "system",
"ignore.unknown.return", args.ignore_unknown_return_values)
# Local config per node
cfg.add(config.Scope.application, "node", "rally.root", paths.rally_root())
cfg.add(config.Scope.application, "node", "rally.cwd", os.getcwd())
cfg.add(config.Scope.applicationOverride, "mechanic", "source.revision",
args.revision)
#TODO dm: Consider renaming this one. It's used by different modules
if args.distribution_version:
cfg.add(config.Scope.applicationOverride, "mechanic",
"distribution.version", args.distribution_version)
cfg.add(config.Scope.applicationOverride, "mechanic",
"distribution.repository", args.distribution_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "repository.name",
args.team_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "car.name", args.car)
cfg.add(config.Scope.applicationOverride, "mechanic", "car.plugins",
csv_to_list(args.elasticsearch_plugins))
cfg.add(config.Scope.applicationOverride, "mechanic", "node.datapaths",
csv_to_list(args.data_paths))
cfg.add(config.Scope.applicationOverride, "mechanic", "preserve.install",
convert.to_bool(args.preserve_install))
cfg.add(config.Scope.applicationOverride, "mechanic", "telemetry.devices",
csv_to_list(args.telemetry))
if args.override_src_dir is not None:
cfg.add(config.Scope.applicationOverride, "source", "local.src.dir",
args.override_src_dir)
cfg.add(config.Scope.applicationOverride, "race", "pipeline",
args.pipeline)
cfg.add(config.Scope.applicationOverride, "race", "laps", args.laps)
cfg.add(config.Scope.applicationOverride, "race", "user.tag",
args.user_tag)
cfg.add(config.Scope.applicationOverride, "track", "repository.name",
args.track_repository)
cfg.add(config.Scope.applicationOverride, "track", "track.name",
args.track)
cfg.add(config.Scope.applicationOverride, "track", "challenge.name",
args.challenge)
cfg.add(config.Scope.applicationOverride, "track", "test.mode.enabled",
args.test_mode)
cfg.add(config.Scope.applicationOverride, "track", "auto_manage_indices",
to_bool(args.auto_manage_indices))
cfg.add(config.Scope.applicationOverride, "reporting", "format",
args.report_format)
cfg.add(config.Scope.applicationOverride, "reporting", "output.path",
args.report_file)
if sub_command == "compare":
cfg.add(config.Scope.applicationOverride, "reporting",
"baseline.timestamp", args.baseline)
cfg.add(config.Scope.applicationOverride, "reporting",
"contender.timestamp", args.contender)
################################
# new section name: driver
################################
cfg.add(config.Scope.applicationOverride, "benchmarks", "cluster.health",
args.cluster_health)
cfg.add(config.Scope.applicationOverride, "driver", "profiling",
args.enable_driver_profiling)
if sub_command != "list":
# Also needed by mechanic (-> telemetry) - duplicate by module?
cfg.add(config.Scope.applicationOverride, "client", "hosts",
_normalize_hosts(csv_to_list(args.target_hosts)))
client_options = kv_to_map(csv_to_list(args.client_options))
cfg.add(config.Scope.applicationOverride, "client", "options",
client_options)
if "timeout" not in client_options:
console.info(
"You did not provide an explicit timeout in the client options. Assuming default of 10 seconds."
)
# split by component?
if sub_command == "list":
cfg.add(config.Scope.applicationOverride, "system",
"list.config.option", args.configuration)
cfg.add(config.Scope.applicationOverride, "system",
"list.races.max_results", args.limit)
configure_logging(cfg)
logger.info("OS [%s]" % str(os.uname()))
logger.info("Python [%s]" % str(sys.implementation))
logger.info("Rally version [%s]" % version.version())
logger.info("Command line arguments: %s" % args)
# Configure networking
net.init()
if not args.offline:
if not net.has_internet_connection():
console.warn(
"No Internet connection detected. Automatic download of track data sets etc. is disabled.",
logger=logger)
cfg.add(config.Scope.applicationOverride, "system", "offline.mode",
True)
else:
logger.info("Detected a working Internet connection.")
# Kill any lingering Rally processes before attempting to continue - the actor system needs to be a singleton on this machine
# noinspection PyBroadException
try:
process.kill_running_rally_instances()
except BaseException:
logger.exception(
"Could not terminate potentially running Rally instances correctly. Attempting to go on anyway."
)
success = dispatch_sub_command(cfg, sub_command)
end = time.time()
if success:
console.println("")
console.info("SUCCESS (took %d seconds)" % (end - start),
overline="-",
underline="-")
else:
console.println("")
console.info("FAILURE (took %d seconds)" % (end - start),
overline="-",
underline="-")
sys.exit(64)
def receiveMessage(self, msg, sender):
# at the moment, we implement all message handling blocking. This is not ideal but simple to get started with. Besides, the caller
# needs to block anyway. The only reason we implement mechanic as an actor is to distribute them.
# noinspection PyBroadException
try:
logger.debug(
"NodeMechanicActor#receiveMessage(msg = [%s] sender = [%s])" %
(str(type(msg)), str(sender)))
if isinstance(msg, StartEngine):
logger.info("Starting engine")
# Load node-specific configuration
self.config = config.Config(config_name=msg.cfg.name)
self.config.load_config()
self.config.add(config.Scope.application, "node", "rally.root",
paths.rally_root())
# copy only the necessary configuration sections
self.config.add_all(msg.cfg, "system")
self.config.add_all(msg.cfg, "client")
self.config.add_all(msg.cfg, "track")
self.config.add_all(msg.cfg, "mechanic")
if msg.port is not None:
# we need to override the port with the value that the user has specified instead of using the default value (39200)
self.config.add(config.Scope.benchmark, "provisioning",
"node.http.port", msg.port)
self.metrics_store = metrics.InMemoryMetricsStore(self.config)
self.metrics_store.open(ctx=msg.open_metrics_context)
self.mechanic = create(self.config, self.metrics_store,
self.single_machine, msg.sources,
msg.build, msg.distribution,
msg.external, msg.docker)
cluster = self.mechanic.start_engine()
self.send(
sender,
EngineStarted(
ClusterMetaInfo(cluster.hosts, cluster.source_revision,
cluster.distribution_version),
self.metrics_store.meta_info))
elif isinstance(msg, OnBenchmarkStart):
self.metrics_store.lap = msg.lap
self.mechanic.on_benchmark_start()
self.send(sender, Success())
elif isinstance(msg, OnBenchmarkStop):
self.mechanic.on_benchmark_stop()
# clear metrics store data to not send duplicate system metrics data
self.send(
sender,
BenchmarkStopped(
self.metrics_store.to_externalizable(clear=True)))
elif isinstance(msg, StopEngine):
logger.info("Stopping engine")
self.mechanic.stop_engine()
self.send(
sender,
EngineStopped(self.metrics_store.to_externalizable()))
# clear all state as the mechanic might get reused later
self.config = None
self.mechanic = None
self.metrics_store = None
except BaseException:
logger.exception("Cannot process message [%s]" % msg)
# avoid "can't pickle traceback objects"
import traceback
ex_type, ex_value, ex_traceback = sys.exc_info()
self.send(sender, Failure(ex_value, traceback.format_exc()))
def receiveMessage(self, msg, sender):
# at the moment, we implement all message handling blocking. This is not ideal but simple to get started with. Besides, the caller
# needs to block anyway. The only reason we implement mechanic as an actor is to distribute them.
# noinspection PyBroadException
try:
logger.debug("NodeMechanicActor#receiveMessage(msg = [%s] sender = [%s])" % (str(type(msg)), str(sender)))
if isinstance(msg, StartNodes):
self.host = msg.ip
if msg.external:
logger.info("Connecting to externally provisioned nodes on [%s]." % msg.ip)
else:
logger.info("Starting node(s) %s on [%s]." % (msg.node_ids, msg.ip))
# Load node-specific configuration
self.config = config.auto_load_local_config(msg.cfg, additional_sections=[
# only copy the relevant bits
"track", "mechanic", "client",
# allow metrics store to extract race meta-data
"race",
"source"
])
# set root path (normally done by the main entry point)
self.config.add(config.Scope.application, "node", "rally.root", paths.rally_root())
if not msg.external:
self.config.add(config.Scope.benchmark, "provisioning", "node.ip", msg.ip)
# we need to override the port with the value that the user has specified instead of using the default value (39200)
self.config.add(config.Scope.benchmark, "provisioning", "node.http.port", msg.port)
self.config.add(config.Scope.benchmark, "provisioning", "node.ids", msg.node_ids)
self.metrics_store = metrics.InMemoryMetricsStore(self.config)
self.metrics_store.open(ctx=msg.open_metrics_context)
# avoid follow-up errors in case we receive an unexpected ActorExitRequest due to an early failure in a parent actor.
self.metrics_store.lap = 0
self.mechanic = create(self.config, self.metrics_store, msg.all_node_ips, msg.cluster_settings, msg.sources, msg.build,
msg.distribution, msg.external, msg.docker)
nodes = self.mechanic.start_engine()
self.running = True
self.send(sender, NodesStarted([NodeMetaInfo(node) for node in nodes], self.metrics_store.meta_info))
elif isinstance(msg, ApplyMetricsMetaInfo):
self.metrics_store.merge_meta_info(msg.meta_info)
self.send(sender, MetricsMetaInfoApplied())
elif isinstance(msg, ResetRelativeTime):
logger.info("Resetting relative time of system metrics store on host [%s]." % self.host)
self.metrics_store.reset_relative_time()
elif isinstance(msg, OnBenchmarkStart):
self.metrics_store.lap = msg.lap
self.mechanic.on_benchmark_start()
self.send(sender, BenchmarkStarted())
elif isinstance(msg, OnBenchmarkStop):
self.mechanic.on_benchmark_stop()
# clear metrics store data to not send duplicate system metrics data
self.send(sender, BenchmarkStopped(self.metrics_store.to_externalizable(clear=True)))
elif isinstance(msg, StopNodes):
logger.info("Stopping nodes %s." % self.mechanic.nodes)
self.mechanic.stop_engine()
self.send(sender, NodesStopped(self.metrics_store.to_externalizable()))
# clear all state as the mechanic might get reused later
self.running = False
self.config = None
self.mechanic = None
self.metrics_store = None
elif isinstance(msg, thespian.actors.ActorExitRequest):
if self.running:
logger.info("Stopping nodes %s (due to ActorExitRequest)" % self.mechanic.nodes)
self.mechanic.stop_engine()
self.running = False
except BaseException:
self.running = False
logger.exception("Cannot process message [%s]" % msg)
# avoid "can't pickle traceback objects"
import traceback
ex_type, ex_value, ex_traceback = sys.exc_info()
self.send(sender, actor.BenchmarkFailure(ex_value, traceback.format_exc()))
def main():
check_python_version()
start = time.time()
# Early init of console output so we start to show everything consistently.
console.init(quiet=False)
# allow to see a thread-dump on SIGQUIT
faulthandler.register(signal.SIGQUIT, file=sys.stderr)
pre_configure_logging()
arg_parser = create_arg_parser()
args = arg_parser.parse_args()
console.init(quiet=args.quiet)
console.println(BANNER)
cfg = config.Config(config_name=args.configuration_name)
sub_command = derive_sub_command(args, cfg)
ensure_configuration_present(cfg, args, sub_command)
if args.effective_start_date:
cfg.add(config.Scope.application, "system", "time.start",
args.effective_start_date)
cfg.add(config.Scope.application, "system", "time.start.user_provided",
True)
else:
cfg.add(config.Scope.application, "system", "time.start",
datetime.datetime.utcnow())
cfg.add(config.Scope.application, "system", "time.start.user_provided",
False)
cfg.add(config.Scope.applicationOverride, "system", "quiet.mode",
args.quiet)
# per node?
cfg.add(config.Scope.applicationOverride, "system", "offline.mode",
args.offline)
cfg.add(config.Scope.applicationOverride, "system", "logging.output",
args.logging)
# Local config per node
cfg.add(config.Scope.application, "node", "rally.root", paths.rally_root())
cfg.add(config.Scope.application, "node", "rally.cwd", os.getcwd())
cfg.add(config.Scope.applicationOverride, "mechanic", "source.revision",
args.revision)
if args.distribution_version:
cfg.add(config.Scope.applicationOverride, "mechanic",
"distribution.version", args.distribution_version)
cfg.add(config.Scope.applicationOverride, "mechanic",
"distribution.repository", args.distribution_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "repository.name",
args.team_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "car.names",
csv_to_list(args.car))
cfg.add(config.Scope.applicationOverride, "mechanic", "car.plugins",
csv_to_list(args.elasticsearch_plugins))
cfg.add(config.Scope.applicationOverride, "mechanic", "node.datapaths",
csv_to_list(args.data_paths))
if args.keep_cluster_running:
cfg.add(config.Scope.applicationOverride, "mechanic", "keep.running",
True)
# force-preserve the cluster nodes.
cfg.add(config.Scope.applicationOverride, "mechanic",
"preserve.install", True)
else:
cfg.add(config.Scope.applicationOverride, "mechanic", "keep.running",
False)
cfg.add(config.Scope.applicationOverride, "mechanic",
"preserve.install", convert.to_bool(args.preserve_install))
cfg.add(config.Scope.applicationOverride, "mechanic", "telemetry.devices",
csv_to_list(args.telemetry))
cfg.add(config.Scope.applicationOverride, "race", "pipeline",
args.pipeline)
cfg.add(config.Scope.applicationOverride, "race", "laps", args.laps)
cfg.add(config.Scope.applicationOverride, "race", "user.tag",
args.user_tag)
# We can assume here that if a track-path is given, the user did not specify a repository either (although argparse sets it to
# its default value)
if args.track_path:
cfg.add(config.Scope.applicationOverride, "track", "track.path",
os.path.abspath(io.normalize_path(args.track_path)))
cfg.add(config.Scope.applicationOverride, "track", "repository.name",
None)
if args.track:
# stay as close as possible to argparse errors although we have a custom validation.
arg_parser.error(
"argument --track not allowed with argument --track-path")
# cfg.add(config.Scope.applicationOverride, "track", "track.name", None)
else:
# cfg.add(config.Scope.applicationOverride, "track", "track.path", None)
cfg.add(config.Scope.applicationOverride, "track", "repository.name",
args.track_repository)
# set the default programmatically because we need to determine whether the user has provided a value
chosen_track = args.track if args.track else "geonames"
cfg.add(config.Scope.applicationOverride, "track", "track.name",
chosen_track)
cfg.add(config.Scope.applicationOverride, "track", "params",
kv_to_map(csv_to_list(args.track_params)))
cfg.add(config.Scope.applicationOverride, "track", "challenge.name",
args.challenge)
cfg.add(config.Scope.applicationOverride, "track", "include.tasks",
csv_to_list(args.include_tasks))
cfg.add(config.Scope.applicationOverride, "track", "test.mode.enabled",
args.test_mode)
cfg.add(config.Scope.applicationOverride, "track", "auto_manage_indices",
to_bool(args.auto_manage_indices))
cfg.add(config.Scope.applicationOverride, "reporting", "format",
args.report_format)
cfg.add(config.Scope.applicationOverride, "reporting", "values",
args.show_in_report)
cfg.add(config.Scope.applicationOverride, "reporting", "output.path",
args.report_file)
if sub_command == "compare":
cfg.add(config.Scope.applicationOverride, "reporting",
"baseline.timestamp", args.baseline)
cfg.add(config.Scope.applicationOverride, "reporting",
"contender.timestamp", args.contender)
cfg.add(config.Scope.applicationOverride, "driver", "cluster.health",
args.cluster_health)
if args.cluster_health != "green":
console.warn(
"--cluster-health is deprecated and will be removed in a future version of Rally."
)
cfg.add(config.Scope.applicationOverride, "driver", "profiling",
args.enable_driver_profiling)
cfg.add(config.Scope.applicationOverride, "driver", "on.error",
args.on_error)
cfg.add(config.Scope.applicationOverride, "driver", "load_driver_hosts",
csv_to_list(args.load_driver_hosts))
if sub_command != "list":
# Also needed by mechanic (-> telemetry) - duplicate by module?
cfg.add(config.Scope.applicationOverride, "client", "hosts",
_normalize_hosts(csv_to_list(args.target_hosts)))
client_options = kv_to_map(csv_to_list(args.client_options))
cfg.add(config.Scope.applicationOverride, "client", "options",
client_options)
if "timeout" not in client_options:
console.info(
"You did not provide an explicit timeout in the client options. Assuming default of 10 seconds."
)
# split by component?
if sub_command == "list":
cfg.add(config.Scope.applicationOverride, "system",
"list.config.option", args.configuration)
cfg.add(config.Scope.applicationOverride, "system",
"list.races.max_results", args.limit)
configure_logging(cfg)
logger.info("OS [%s]" % str(os.uname()))
logger.info("Python [%s]" % str(sys.implementation))
logger.info("Rally version [%s]" % version.version())
logger.info("Command line arguments: %s" % args)
# Configure networking
net.init()
if not args.offline:
if not net.has_internet_connection():
console.warn(
"No Internet connection detected. Automatic download of track data sets etc. is disabled.",
logger=logger)
cfg.add(config.Scope.applicationOverride, "system", "offline.mode",
True)
else:
logger.info("Detected a working Internet connection.")
# Kill any lingering Rally processes before attempting to continue - the actor system needs to be a singleton on this machine
# noinspection PyBroadException
try:
process.kill_running_rally_instances()
except BaseException:
logger.exception(
"Could not terminate potentially running Rally instances correctly. Attempting to go on anyway."
)
success = dispatch_sub_command(cfg, sub_command)
end = time.time()
if success:
console.println("")
console.info("SUCCESS (took %d seconds)" % (end - start),
overline="-",
underline="-")
else:
console.println("")
console.info("FAILURE (took %d seconds)" % (end - start),
overline="-",
underline="-")
sys.exit(64)
def main():
check_python_version()
start = time.time()
# Early init of console output so we start to show everything consistently.
console.init(quiet=False)
# allow to see a thread-dump on SIGQUIT
faulthandler.register(signal.SIGQUIT, file=sys.stderr)
pre_configure_logging()
args = parse_args()
console.init(quiet=args.quiet)
console.println(BANNER)
cfg = config.Config(config_name=args.configuration_name)
sub_command = derive_sub_command(args, cfg)
ensure_configuration_present(cfg, args, sub_command)
if args.effective_start_date:
cfg.add(config.Scope.application, "system", "time.start", args.effective_start_date)
cfg.add(config.Scope.application, "system", "time.start.user_provided", True)
else:
cfg.add(config.Scope.application, "system", "time.start", datetime.datetime.utcnow())
cfg.add(config.Scope.application, "system", "time.start.user_provided", False)
cfg.add(config.Scope.applicationOverride, "system", "quiet.mode", args.quiet)
# per node?
cfg.add(config.Scope.applicationOverride, "system", "offline.mode", args.offline)
cfg.add(config.Scope.applicationOverride, "system", "logging.output", args.logging)
# Local config per node
cfg.add(config.Scope.application, "node", "rally.root", paths.rally_root())
cfg.add(config.Scope.application, "node", "rally.cwd", os.getcwd())
cfg.add(config.Scope.applicationOverride, "mechanic", "source.revision", args.revision)
#TODO dm: Consider renaming this one. It's used by different modules
if args.distribution_version:
cfg.add(config.Scope.applicationOverride, "mechanic", "distribution.version", args.distribution_version)
cfg.add(config.Scope.applicationOverride, "mechanic", "distribution.repository", args.distribution_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "repository.name", args.team_repository)
cfg.add(config.Scope.applicationOverride, "mechanic", "car.names", csv_to_list(args.car))
cfg.add(config.Scope.applicationOverride, "mechanic", "car.plugins", csv_to_list(args.elasticsearch_plugins))
cfg.add(config.Scope.applicationOverride, "mechanic", "node.datapaths", csv_to_list(args.data_paths))
if args.keep_cluster_running:
cfg.add(config.Scope.applicationOverride, "mechanic", "keep.running", True)
# force-preserve the cluster nodes.
cfg.add(config.Scope.applicationOverride, "mechanic", "preserve.install", True)
else:
cfg.add(config.Scope.applicationOverride, "mechanic", "keep.running", False)
cfg.add(config.Scope.applicationOverride, "mechanic", "preserve.install", convert.to_bool(args.preserve_install))
cfg.add(config.Scope.applicationOverride, "mechanic", "telemetry.devices", csv_to_list(args.telemetry))
cfg.add(config.Scope.applicationOverride, "race", "pipeline", args.pipeline)
cfg.add(config.Scope.applicationOverride, "race", "laps", args.laps)
cfg.add(config.Scope.applicationOverride, "race", "user.tag", args.user_tag)
cfg.add(config.Scope.applicationOverride, "track", "repository.name", args.track_repository)
cfg.add(config.Scope.applicationOverride, "track", "track.name", args.track)
cfg.add(config.Scope.applicationOverride, "track", "challenge.name", args.challenge)
cfg.add(config.Scope.applicationOverride, "track", "include.tasks", csv_to_list(args.include_tasks))
cfg.add(config.Scope.applicationOverride, "track", "test.mode.enabled", args.test_mode)
cfg.add(config.Scope.applicationOverride, "track", "auto_manage_indices", to_bool(args.auto_manage_indices))
cfg.add(config.Scope.applicationOverride, "reporting", "format", args.report_format)
cfg.add(config.Scope.applicationOverride, "reporting", "output.path", args.report_file)
if sub_command == "compare":
cfg.add(config.Scope.applicationOverride, "reporting", "baseline.timestamp", args.baseline)
cfg.add(config.Scope.applicationOverride, "reporting", "contender.timestamp", args.contender)
################################
# new section name: driver
################################
cfg.add(config.Scope.applicationOverride, "driver", "cluster.health", args.cluster_health)
cfg.add(config.Scope.applicationOverride, "driver", "profiling", args.enable_driver_profiling)
cfg.add(config.Scope.applicationOverride, "driver", "load_driver_hosts", csv_to_list(args.load_driver_hosts))
if sub_command != "list":
# Also needed by mechanic (-> telemetry) - duplicate by module?
cfg.add(config.Scope.applicationOverride, "client", "hosts", _normalize_hosts(csv_to_list(args.target_hosts)))
client_options = kv_to_map(csv_to_list(args.client_options))
cfg.add(config.Scope.applicationOverride, "client", "options", client_options)
if "timeout" not in client_options:
console.info("You did not provide an explicit timeout in the client options. Assuming default of 10 seconds.")
# split by component?
if sub_command == "list":
cfg.add(config.Scope.applicationOverride, "system", "list.config.option", args.configuration)
cfg.add(config.Scope.applicationOverride, "system", "list.races.max_results", args.limit)
configure_logging(cfg)
logger.info("OS [%s]" % str(os.uname()))
logger.info("Python [%s]" % str(sys.implementation))
logger.info("Rally version [%s]" % version.version())
logger.info("Command line arguments: %s" % args)
# Configure networking
net.init()
if not args.offline:
if not net.has_internet_connection():
console.warn("No Internet connection detected. Automatic download of track data sets etc. is disabled.",
logger=logger)
cfg.add(config.Scope.applicationOverride, "system", "offline.mode", True)
else:
logger.info("Detected a working Internet connection.")
# Kill any lingering Rally processes before attempting to continue - the actor system needs to be a singleton on this machine
# noinspection PyBroadException
try:
process.kill_running_rally_instances()
except BaseException:
logger.exception("Could not terminate potentially running Rally instances correctly. Attempting to go on anyway.")
success = dispatch_sub_command(cfg, sub_command)
end = time.time()
if success:
console.println("")
console.info("SUCCESS (took %d seconds)" % (end - start), overline="-", underline="-")
else:
console.println("")
console.info("FAILURE (took %d seconds)" % (end - start), overline="-", underline="-")
sys.exit(64)