def _submit_and_await_termination(
client: skein.Client,
spec: skein.ApplicationSpec,
tasks: typing.List[str],
poll_every_secs: int = 10
):
app = client.submit_and_connect(spec)
events: typing.Dict[str, typing.Dict[str, str]] = {task: {} for task in tasks}
event_listener = Thread(target=_aggregate_events, args=(app.kv, events))
event_listener.start()
with _shutdown_on_exception(app):
state = None
while True:
report = client.application_report(app.id)
logger.info(
f"Application report for {app.id} (state: {report.state})")
if state != report.state:
logger.info(_format_app_report(report))
if report.final_status != "undefined":
event_listener.join()
logger.info(_format_run_summary(events))
if report.final_status == "failed":
raise RunFailed
else:
break
time.sleep(poll_every_secs)
state = report.state
def _setup_skein_cluster(pyenvs: Dict[NodeLabel, PythonEnvDescription],
task_specs: Dict[str, TaskSpec] = TASK_SPEC_NONE,
*,
skein_client: skein.Client = None,
files: Dict[str, str] = None,
env: Dict[str, str] = {},
queue: str = "default",
acls: ACLs = None,
file_systems: List[str] = None,
log_conf_file: str = None,
standalone_client_mode: bool = False) -> SkeinCluster:
os.environ["JAVA_TOOL_OPTIONS"] = \
"-XX:ParallelGCThreads=1 -XX:CICompilerCount=2 "\
f"{os.environ.get('JAVA_TOOL_OPTIONS', '')}"
with tempfile.TemporaryDirectory() as tempdir:
task_files, task_env = _setup_task_env(tempdir, files, env)
services = {}
for task_type, task_spec in list(task_specs.items()):
pyenv = pyenvs[task_spec.label]
service_env = task_env.copy()
if task_spec.termination_timeout_seconds >= 0:
_add_to_env(service_env, "SERVICE_TERMINATION_TIMEOUT_SECONDS",
str(task_spec.termination_timeout_seconds))
services[task_type] = skein.Service(
script=gen_task_cmd(pyenv, log_conf_file),
resources=skein.model.Resources(task_spec.memory,
task_spec.vcores),
max_restarts=0,
instances=task_spec.instances,
node_label=task_spec.label.value,
files={
**task_files, pyenv.dest_path: pyenv.path_to_archive
},
env=service_env)
spec = skein.ApplicationSpec(services,
queue=queue,
acls=acls,
file_systems=file_systems)
if skein_client is None:
skein_client = skein.Client()
task_instances = [(task_type, spec.instances)
for task_type, spec in task_specs.items()]
events: Dict[str, Dict[str, str]] = \
{task: {} for task in iter_tasks(task_instances)}
app = skein_client.submit_and_connect(spec)
# Start a thread which collects all events posted by all tasks in kv store
event_listener = Thread(target=_aggregate_events,
args=(app.kv, events))
event_listener.start()
cluster_spec = _setup_cluster_tasks(task_instances, app,
standalone_client_mode)
return SkeinCluster(skein_client, app, task_instances, cluster_spec,
event_listener, events)
def wait_for_finished(client: skein.Client, app_id: str):
logger.info(f"application_id: {app_id}")
while True:
report = client.application_report(app_id)
logger.info(report)
if report.final_status != "undefined":
logger.info(report.final_status)
break
time.sleep(3)
def _get_app_logs(
client: skein.Client, app: skein.ApplicationClient,
wait_for_nb_logs: int) -> Optional[skein.model.ApplicationLogs]:
for ind in range(YARN_LOG_TRIES):
try:
logs = client.application_logs(app.id)
nb_keys = len(logs.keys())
logger.info(f"Got {nb_keys}/{wait_for_nb_logs} log files")
if nb_keys == wait_for_nb_logs:
return logs
except Exception:
logger.warn(
f"Cannot collect logs (attempt {ind+1}/{YARN_LOG_TRIES})",
exc_info=True)
time.sleep(3)
return None
def get_application_logs(
client: skein.Client,
app_id: str,
wait_for_nb_logs: Optional[int] = None,
log_tries: int = 15) -> Optional[skein.model.ApplicationLogs]:
for ind in range(log_tries):
try:
logs = client.application_logs(app_id)
nb_keys = len(logs.keys())
logger.info(f"Got {nb_keys}/{wait_for_nb_logs} log files")
if not wait_for_nb_logs or nb_keys == wait_for_nb_logs:
return logs
except Exception:
logger.warning(
f"Cannot collect logs (attempt {ind+1}/{log_tries})")
time.sleep(3)
return None
def _setup_skein_cluster(
pyenvs: Dict[topologies.NodeLabel, _env.PythonEnvDescription],
task_specs: Dict[str, topologies.TaskSpec] = TASK_SPEC_NONE,
*,
custom_task_module: Optional[str] = None,
skein_client: skein.Client = None,
files: Dict[str, str] = None,
env: Dict[str, str] = {},
queue: str = "default",
acls: ACLs = None,
file_systems: List[str] = None,
name: str = "RunOnYarn",
n_try: int = 0,
pre_script_hook: Optional[str] = None
) -> SkeinCluster:
os.environ["JAVA_TOOL_OPTIONS"] = \
"-XX:ParallelGCThreads=1 -XX:CICompilerCount=2 "\
f"{os.environ.get('JAVA_TOOL_OPTIONS', '')}"
pre_script_hook = pre_script_hook if pre_script_hook else ""
with tempfile.TemporaryDirectory() as tempdir:
task_files, task_env = _setup_task_env(tempdir, files, env, n_try)
services = {}
for task_type, task_spec in list(task_specs.items()):
pyenv = pyenvs[task_spec.label]
service_env = task_env.copy()
if task_spec.tb_termination_timeout_seconds >= 0:
service_env["TB_TERMINATION_TIMEOUT_SECONDS"] = \
str(task_spec.tb_termination_timeout_seconds)
if task_spec.tb_model_dir:
service_env["TB_MODEL_DIR"] = str(task_spec.tb_model_dir)
if task_spec.tb_extra_args:
service_env["TB_EXTRA_ARGS"] = str(task_spec.tb_extra_args)
services[task_type] = skein.Service(
script=f'''
set -x
{pre_script_hook}
{_env.gen_task_cmd(
pyenv,
task_type,
custom_task_module)}
''',
resources=skein.model.Resources(task_spec.memory, task_spec.vcores),
max_restarts=0,
instances=task_spec.instances,
node_label=task_spec.label.value,
files={
**task_files,
pyenv.dest_path: pyenv.path_to_archive
},
env=service_env)
# on the cluster we don't ask again for delegation tokens
if "HADOOP_TOKEN_FILE_LOCATION" in os.environ:
file_systems = None
spec = skein.ApplicationSpec(
services,
queue=queue,
acls=acls,
file_systems=file_systems,
name=name
)
if skein_client is None:
skein_client = skein.Client()
task_instances = [(task_type, spec.instances) for task_type, spec in task_specs.items()]
events: Dict[str, Dict[str, str]] = \
{task: {} for task in _internal.iter_tasks(task_instances)}
app = skein_client.submit_and_connect(spec)
# Start a thread which collects all events posted by all tasks in kv store
event_listener = Thread(target=_aggregate_events, args=(app.kv, events))
event_listener.start()
return SkeinCluster(skein_client, app, task_instances, event_listener, events)
def submit(
skein_client: skein.Client,
module_name: str,
additional_files: Optional[List[str]] = None,
archive_hdfs: Optional[str] = None,
args: Optional[List[str]] = None,
env_vars: Optional[Dict[str, str]] = None,
hadoop_file_systems: Tuple[str, ...] = (),
max_attempts: int = 1,
max_restarts: int = 0,
memory: str = "1 GiB",
name: str = "yarn_launcher",
node_label: Optional[str] = None,
num_containers: int = 1,
num_cores: int = 1,
pre_script_hook: Optional[str] = None,
queue: Optional[str] = None,
user: Optional[str] = None,
) -> str:
"""Submit application via skein."""
with tempfile.TemporaryDirectory() as tmp_dir:
# Update Environment Variables and script hook
env = dict(env_vars) if env_vars else dict()
pre_script_hook = pre_script_hook if pre_script_hook else ""
env.update({"SKEIN_CONFIG": "./.skein", "GIT_PYTHON_REFRESH": "quiet"})
# Create Skein Config, Service and Spec
skein_config = skein_config_builder.build(
module_name,
args=args if args else [],
package_path=archive_hdfs,
additional_files=additional_files,
tmp_dir=tmp_dir,
)
skein_service = skein.Service(
resources=skein.model.Resources(memory, num_cores),
instances=num_containers,
files=skein_config.files,
env=env,
script=f"""
set -x
env
{pre_script_hook}
{skein_config.script}
""",
max_restarts=max_restarts,
)
skein_spec = skein.ApplicationSpec(
name=name,
file_systems=list(hadoop_file_systems),
services={name: skein_service},
acls=skein.model.ACLs(enable=True,
ui_users=["*"],
view_users=["*"]),
max_attempts=max_attempts,
)
# Activate impersonation only if user to run the job is not
# the current user (yarn issue)
if user and user != getpass.getuser():
skein_spec.user = user
if queue:
skein_spec.queue = queue
if node_label:
skein_service.node_label = node_label
return skein_client.submit(skein_spec)
