def report_engine_event(
self,
message,
pipeline_run,
engine_event_data=None,
cls=None,
step_key=None,
):
"""
Report a EngineEvent that occurred outside of a pipeline execution context.
"""
from dagster.core.events import EngineEventData, DagsterEvent, DagsterEventType
from dagster.core.events.log import DagsterEventRecord
check.class_param(cls, "cls")
check.str_param(message, "message")
check.inst_param(pipeline_run, "pipeline_run", PipelineRun)
engine_event_data = check.opt_inst_param(
engine_event_data,
"engine_event_data",
EngineEventData,
EngineEventData([]),
)
if cls:
message = "[{}] {}".format(cls.__name__, message)
log_level = logging.INFO
if engine_event_data and engine_event_data.error:
log_level = logging.ERROR
dagster_event = DagsterEvent(
event_type_value=DagsterEventType.ENGINE_EVENT.value,
pipeline_name=pipeline_run.pipeline_name,
message=message,
event_specific_data=engine_event_data,
)
event_record = DagsterEventRecord(
message=message,
user_message=message,
level=log_level,
pipeline_name=pipeline_run.pipeline_name,
run_id=pipeline_run.run_id,
error_info=None,
timestamp=time.time(),
step_key=step_key,
dagster_event=dagster_event,
)
self.handle_new_event(event_record)
return dagster_event
def terminate(self, run_id):
check.str_param(run_id, "run_id")
run = self._instance.get_run_by_id(run_id)
if not run:
return False
can_terminate = self.can_terminate(run_id)
if not can_terminate:
self._instance.report_engine_event(
message=
"Unable to terminate pipeline: can_terminate returned {}.".
format(can_terminate),
pipeline_run=run,
cls=self.__class__,
)
return False
job_name = get_job_name_from_run_id(run_id)
job_namespace = self.get_namespace_from_run_config(run_id)
self._instance.report_run_canceling(run)
try:
termination_result = delete_job(job_name=job_name,
namespace=job_namespace)
if termination_result:
self._instance.report_engine_event(
message="Pipeline was terminated successfully.",
pipeline_run=run,
cls=self.__class__,
)
else:
self._instance.report_engine_event(
message=
"Pipeline was not terminated successfully; delete_job returned {}"
.format(termination_result),
pipeline_run=run,
cls=self.__class__,
)
return termination_result
except Exception: # pylint: disable=broad-except
self._instance.report_engine_event(
message=
"Pipeline was not terminated successfully; encountered error in delete_job",
pipeline_run=run,
engine_event_data=EngineEventData.engine_error(
serializable_error_info_from_exc_info(sys.exc_info())),
cls=self.__class__,
)
def core_execute_run(recon_pipeline, pipeline_run, instance):
check.inst_param(recon_pipeline, "recon_pipeline", ReconstructablePipeline)
check.inst_param(pipeline_run, "pipeline_run", PipelineRun)
check.inst_param(instance, "instance", DagsterInstance)
# try to load the pipeline definition early
try:
recon_pipeline.get_definition()
except Exception: # pylint: disable=broad-except
yield instance.report_engine_event(
"Could not load pipeline definition.",
pipeline_run,
EngineEventData.engine_error(
serializable_error_info_from_exc_info(sys.exc_info())),
)
yield from _report_run_failed_if_not_finished(instance,
pipeline_run.run_id)
return
try:
yield from execute_run_iterator(recon_pipeline, pipeline_run, instance)
except (KeyboardInterrupt, DagsterExecutionInterruptedError):
yield from _report_run_failed_if_not_finished(instance,
pipeline_run.run_id)
yield instance.report_engine_event(
message="Pipeline execution terminated by interrupt",
pipeline_run=pipeline_run,
)
except Exception: # pylint: disable=broad-except
yield instance.report_engine_event(
"An exception was thrown during execution that is likely a framework error, "
"rather than an error in user code.",
pipeline_run,
EngineEventData.engine_error(
serializable_error_info_from_exc_info(sys.exc_info())),
)
yield from _report_run_failed_if_not_finished(instance,
pipeline_run.run_id)
def launch_step(
self,
step_handler_context: StepHandlerContext) -> List[DagsterEvent]:
container_context = self._get_docker_container_context(
step_handler_context)
client = self._get_client(container_context)
step_image = self._get_image(step_handler_context)
validate_docker_image(step_image)
try:
step_container = self._create_step_container(
client, container_context, step_image,
step_handler_context.execute_step_args)
except docker.errors.ImageNotFound:
client.images.pull(step_image)
step_container = self._create_step_container(
client, container_context, step_image,
step_handler_context.execute_step_args)
if len(container_context.networks) > 1:
for network_name in container_context.networks[1:]:
network = client.networks.get(network_name)
network.connect(step_container)
assert (len(
step_handler_context.execute_step_args.step_keys_to_execute) == 1
), "Launching multiple steps is not currently supported"
step_key = step_handler_context.execute_step_args.step_keys_to_execute[
0]
events = [
DagsterEvent(
event_type_value=DagsterEventType.ENGINE_EVENT.value,
pipeline_name=step_handler_context.execute_step_args.
pipeline_origin.pipeline_name,
step_key=step_key,
message="Launching step in Docker container",
event_specific_data=EngineEventData([
MetadataEntry("Step key", value=step_key),
MetadataEntry("Docker container id",
value=step_container.id),
], ),
)
]
step_container.start()
return events
def _launch_k8s_job_with_args(self, job_name, args, run):
container_context = self.get_container_context_for_run(run)
pod_name = job_name
pipeline_origin = run.pipeline_code_origin
user_defined_k8s_config = get_user_defined_k8s_config(
frozentags(run.tags))
repository_origin = pipeline_origin.repository_origin
job_config = container_context.get_k8s_job_config(
job_image=repository_origin.container_image, run_launcher=self)
self._instance.add_run_tags(
run.run_id,
{DOCKER_IMAGE_TAG: job_config.job_image},
)
job = construct_dagster_k8s_job(
job_config=job_config,
args=args,
job_name=job_name,
pod_name=pod_name,
component="run_worker",
user_defined_k8s_config=user_defined_k8s_config,
labels={
"dagster/job": pipeline_origin.pipeline_name,
"dagster/run-id": run.run_id,
},
)
self._instance.report_engine_event(
"Creating Kubernetes run worker job",
run,
EngineEventData([
MetadataEntry("Kubernetes Job name", value=job_name),
MetadataEntry("Kubernetes Namespace",
value=container_context.namespace),
MetadataEntry("Run ID", value=run.run_id),
]),
cls=self.__class__,
)
self._batch_api.create_namespaced_job(
body=job, namespace=container_context.namespace)
self._instance.report_engine_event(
"Kubernetes run worker job created",
run,
cls=self.__class__,
)
def evt(name):
return EventRecord(
None,
name,
"debug",
"",
"foo",
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
"nonce",
event_specific_data=EngineEventData.in_process(999),
),
)
def create_test_event_log_record(message: str, run_id: str = DEFAULT_RUN_ID):
return EventRecord(
None,
message,
"debug",
"",
run_id,
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
"nonce",
event_specific_data=EngineEventData.in_process(999),
),
)
def _execute_plan(_self, instance_ref_dict, handle_dict, run_id, step_keys, retries_dict):
check.dict_param(instance_ref_dict, 'instance_ref_dict')
check.dict_param(handle_dict, 'handle_dict')
check.str_param(run_id, 'run_id')
check.list_param(step_keys, 'step_keys', of_type=str)
check.dict_param(retries_dict, 'retries_dict')
instance_ref = InstanceRef.from_dict(instance_ref_dict)
instance = DagsterInstance.from_ref(instance_ref)
handle = ExecutionTargetHandle.from_dict(handle_dict)
retries = Retries.from_config(retries_dict)
pipeline_run = instance.get_run_by_id(run_id)
check.invariant(pipeline_run, 'Could not load run {}'.format(run_id))
pipeline_def = handle.build_pipeline_definition().build_sub_pipeline(
pipeline_run.selector.solid_subset
)
step_keys_str = ", ".join(step_keys)
execution_plan = create_execution_plan(
pipeline_def,
pipeline_run.environment_dict,
mode=pipeline_run.mode,
step_keys_to_execute=pipeline_run.step_keys_to_execute,
).build_subset_plan(step_keys)
engine_event = instance.report_engine_event(
'Executing steps {} in celery worker'.format(step_keys_str),
pipeline_run,
EngineEventData(
[EventMetadataEntry.text(step_keys_str, 'step_keys'),], marker_end=DELEGATE_MARKER,
),
CeleryEngine,
step_key=execution_plan.step_key_for_single_step_plans(),
)
events = [engine_event]
for step_event in execute_plan_iterator(
execution_plan,
pipeline_run=pipeline_run,
environment_dict=pipeline_run.environment_dict,
instance=instance,
retries=retries,
):
events.append(step_event)
serialized_events = [serialize_dagster_namedtuple(event) for event in events]
return serialized_events
def create_event(count: int, run_id: str = RUN_ID):
return EventLogEntry(
None,
str(count),
"debug",
"",
run_id,
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
"nonce",
event_specific_data=EngineEventData.in_process(999),
),
)
def evt(name):
return DagsterEventRecord(
None,
name,
'debug',
'',
'foo',
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
'nonce',
event_specific_data=EngineEventData.in_process(999),
),
)
def execute(pipeline_context, execution_plan):
check.inst_param(pipeline_context, 'pipeline_context',
SystemPipelineExecutionContext)
check.inst_param(execution_plan, 'execution_plan', ExecutionPlan)
step_keys_to_execute = execution_plan.step_keys_to_execute
yield DagsterEvent.engine_event(
pipeline_context,
'Executing steps in process (pid: {pid})'.format(pid=os.getpid()),
event_specific_data=EngineEventData.in_process(
os.getpid(),
step_keys_to_execute,
),
)
with time_execution_scope() as timer_result:
check.param_invariant(
isinstance(pipeline_context.executor_config, ExecutorConfig),
'pipeline_context',
'Expected executor_config to be ExecutorConfig got {}'.format(
pipeline_context.executor_config),
)
for event in inner_plan_execution_iterator(
pipeline_context, execution_plan,
pipeline_context.executor_config.retries):
yield event
yield DagsterEvent.engine_event(
pipeline_context,
'Finished steps in process (pid: {pid}) in {duration_ms}'.format(
pid=os.getpid(),
duration_ms=format_duration(timer_result.millis)),
event_specific_data=EngineEventData.in_process(
os.getpid(), step_keys_to_execute),
)
def execute_step_out_of_process(self, step_context, step, errors,
term_events):
command = InProcessExecutorChildProcessCommand(
run_config=step_context.run_config,
pipeline_run=step_context.pipeline_run,
step_key=step.key,
instance_ref=step_context.instance.get_ref(),
term_event=term_events[step.key],
recon_pipeline=self.pipeline,
retries=self.retries,
)
yield DagsterEvent.engine_event(
step_context,
"Launching subprocess for {}".format(step.key),
EngineEventData(marker_start=DELEGATE_MARKER),
step_key=step.key,
)
for ret in execute_child_process_command(command):
if ret is None or isinstance(ret, DagsterEvent):
yield ret
elif isinstance(ret, ChildProcessEvent):
if isinstance(ret, ChildProcessSystemErrorEvent):
errors[ret.pid] = ret.error_info
elif isinstance(ret, KeyboardInterrupt):
yield DagsterEvent.engine_event(
step_context,
"Multiprocess engine: received KeyboardInterrupt - forwarding to active child processes",
EngineEventData.interrupted(list(term_events.keys())),
)
for term_event in term_events.values():
term_event.set()
else:
check.failed(
"Unexpected return value from child process {}".format(
type(ret)))
def execute(self, plan_context, execution_plan):
check.inst_param(plan_context, "plan_context", PlanOrchestrationContext)
check.inst_param(execution_plan, "execution_plan", ExecutionPlan)
step_keys_to_execute = execution_plan.step_keys_to_execute
yield DagsterEvent.engine_event(
plan_context,
"Executing steps in process (pid: {pid})".format(pid=os.getpid()),
event_specific_data=EngineEventData.in_process(os.getpid(), step_keys_to_execute),
)
with time_execution_scope() as timer_result:
yield from iter(
ExecuteRunWithPlanIterable(
execution_plan=plan_context.execution_plan,
iterator=inner_plan_execution_iterator,
execution_context_manager=PlanExecutionContextManager(
pipeline=plan_context.pipeline,
retry_mode=plan_context.retry_mode,
execution_plan=plan_context.execution_plan,
run_config=plan_context.run_config,
pipeline_run=plan_context.pipeline_run,
instance=plan_context.instance,
raise_on_error=plan_context.raise_on_error,
output_capture=plan_context.output_capture,
),
)
)
yield DagsterEvent.engine_event(
plan_context,
"Finished steps in process (pid: {pid}) in {duration_ms}".format(
pid=os.getpid(), duration_ms=format_duration(timer_result.millis)
),
event_specific_data=EngineEventData.in_process(os.getpid(), step_keys_to_execute),
)
def launch_run(self, instance, run, external_pipeline):
check.inst_param(run, 'run', PipelineRun)
check.inst_param(external_pipeline, 'external_pipeline',
ExternalPipeline)
job_name = 'dagster-run-{}'.format(run.run_id)
pod_name = job_name
resources = get_k8s_resource_requirements(
frozentags(external_pipeline.tags))
job = construct_dagster_k8s_job(
job_config=self.job_config,
command=['dagster-graphql'],
args=[
'-p',
'executeRunInProcess',
'-v',
seven.json.dumps({
'runId':
run.run_id,
'repositoryName':
external_pipeline.handle.repository_name,
'repositoryLocationName':
external_pipeline.handle.location_name,
}),
'--remap-sigterm',
],
job_name=job_name,
pod_name=pod_name,
component='runmaster',
resources=resources,
)
self._batch_api.create_namespaced_job(body=job,
namespace=self.job_namespace)
self._instance.report_engine_event(
'Kubernetes runmaster job launched',
run,
EngineEventData([
EventMetadataEntry.text(job_name, 'Kubernetes Job name'),
EventMetadataEntry.text(pod_name, 'Kubernetes Pod name'),
EventMetadataEntry.text(self.job_namespace,
'Kubernetes Namespace'),
EventMetadataEntry.text(run.run_id, 'Run ID'),
]),
cls=K8sRunLauncher,
)
return run
def launch_run(self, instance, run, external_pipeline):
check.inst_param(run, "run", PipelineRun)
check.inst_param(external_pipeline, "external_pipeline",
ExternalPipeline)
job_name = "dagster-run-{}".format(run.run_id)
pod_name = job_name
user_defined_k8s_config = get_user_defined_k8s_config(
frozentags(run.tags))
pipeline_origin = external_pipeline.get_python_origin()
repository_origin = pipeline_origin.repository_origin
job_config = (self._get_grpc_job_config(
repository_origin.container_image)
if repository_origin.container_image else
self.get_static_job_config())
input_json = serialize_dagster_namedtuple(
ExecuteRunArgs(
pipeline_origin=pipeline_origin,
pipeline_run_id=run.run_id,
instance_ref=None,
))
job = construct_dagster_k8s_job(
job_config=job_config,
args=["dagster", "api", "execute_run", input_json],
job_name=job_name,
pod_name=pod_name,
component="run_coordinator",
user_defined_k8s_config=user_defined_k8s_config,
)
self._batch_api.create_namespaced_job(body=job,
namespace=self.job_namespace)
self._instance.report_engine_event(
"Kubernetes run worker job launched",
run,
EngineEventData([
EventMetadataEntry.text(job_name, "Kubernetes Job name"),
EventMetadataEntry.text(self.job_namespace,
"Kubernetes Namespace"),
EventMetadataEntry.text(run.run_id, "Run ID"),
]),
cls=self.__class__,
)
return run
def launch_run(self, instance, run, external_pipeline):
if isinstance(
external_pipeline.get_external_origin().
external_repository_origin.repository_location_origin,
GrpcServerRepositoryLocationOrigin,
):
repository_location_handle = (
external_pipeline.repository_handle.repository_location_handle)
if not isinstance(repository_location_handle,
GrpcServerRepositoryLocationHandle):
raise DagsterInvariantViolationError(
"Expected RepositoryLocationHandle to be of type "
"GrpcServerRepositoryLocationHandle but found type {}".
format(type(repository_location_handle)))
repository_name = external_pipeline.repository_handle.repository_name
location_name = external_pipeline.repository_handle.repository_location_handle.location_name
pipeline_origin = PipelinePythonOrigin(
pipeline_name=external_pipeline.name,
repository_origin=repository_location_handle.
get_repository_python_origin(repository_name),
)
else:
location_name = 'local'
pipeline_origin = external_pipeline.get_python_origin()
input_json = serialize_dagster_namedtuple(
ExecuteRunArgs(
pipeline_origin=pipeline_origin,
pipeline_run_id=run.run_id,
instance_ref=None,
))
app = self._get_app(location_name)
sig = app.signature('launch_run',
args=(input_json, ),
queue=f"{location_name}-pipelines")
result = sig.delay()
instance.report_engine_event(
"Started Celery task for pipeline (task id: {result.id}).".format(
result=result),
run,
EngineEventData(metadata_entries=[
EventMetadataEntry.text(result.id, "task_id"),
]),
)
return run
def _execute_run_command_body(recon_pipeline, pipeline_run_id, instance,
write_stream_fn, set_exit_code_on_failure):
if instance.should_start_background_run_thread:
cancellation_thread, cancellation_thread_shutdown_event = start_run_cancellation_thread(
instance, pipeline_run_id)
pipeline_run = instance.get_run_by_id(pipeline_run_id)
pid = os.getpid()
instance.report_engine_event(
"Started process for run (pid: {pid}).".format(pid=pid),
pipeline_run,
EngineEventData.in_process(pid, marker_end="cli_api_subprocess_init"),
)
run_worker_failed = 0
try:
for event in core_execute_run(
recon_pipeline,
pipeline_run,
instance,
):
write_stream_fn(event)
if event.event_type == DagsterEventType.PIPELINE_FAILURE:
run_worker_failed = True
except:
# relies on core_execute_run writing failures to the event log before raising
run_worker_failed = True
finally:
if instance.should_start_background_run_thread:
cancellation_thread_shutdown_event.set()
if cancellation_thread.is_alive():
cancellation_thread.join(timeout=15)
if cancellation_thread.is_alive():
instance.report_engine_event(
"Cancellation thread did not shutdown gracefully",
pipeline_run,
)
instance.report_engine_event(
"Process for run exited (pid: {pid}).".format(pid=pid),
pipeline_run,
)
return 1 if (run_worker_failed and set_exit_code_on_failure) else 0
def _execute_schedule(graphene_info, external_pipeline, execution_params,
errors):
check.inst_param(external_pipeline, 'external_pipeline', ExternalPipeline)
instance = graphene_info.context.instance
mode, environment_dict = execution_params.mode, execution_params.environment_dict
validation_result = validate_config_from_snap(
external_pipeline.config_schema_snapshot,
external_pipeline.root_config_key_for_mode(mode),
environment_dict,
)
if validation_result.success:
external_execution_plan = graphene_info.context.get_external_execution_plan(
external_pipeline, environment_dict, mode,
execution_params.step_keys)
pipeline_run = instance.create_run(
pipeline_name=external_pipeline.name,
environment_dict=environment_dict,
mode=mode,
solid_subset=(execution_params.selector.solid_subset
if execution_params.selector is not None else None),
tags=execution_params.execution_metadata.tags,
pipeline_snapshot=external_pipeline.pipeline_snapshot,
execution_plan_snapshot=external_execution_plan.
execution_plan_snapshot,
)
# Inject errors into event log at this point
if len(errors) > 0:
for error in errors:
instance.report_engine_event(error.message, pipeline_run,
EngineEventData.engine_error(error))
# Launch run if run launcher is defined
run_launcher = graphene_info.context.instance.run_launcher
if run_launcher:
result = _launch_pipeline_execution_for_created_run(
graphene_info, pipeline_run.run_id)
else:
result = _start_pipeline_execution_for_created_run(
graphene_info, pipeline_run.run_id)
return pipeline_run, result
def cli_api_execute_run(output_file, instance, repository_handle,
pipeline_run):
check.str_param(output_file, 'output_file')
check.inst_param(instance, 'instance', DagsterInstance)
check.inst_param(repository_handle, 'repository_handle', RepositoryHandle)
check.inst_param(pipeline_run, 'pipeline_run', PipelineRun)
pointer = repository_handle.get_pointer()
location_handle = repository_handle.repository_location_handle
if isinstance(location_handle, PythonEnvRepositoryLocationHandle):
executable_path = location_handle.executable_path
elif isinstance(location_handle, InProcessRepositoryLocationHandle):
# [legacy] default to using sys.executable for the in process
# location handle
executable_path = sys.executable
else:
raise DagsterInvariantViolationError(
"Unable to resolve executable_path for repository location handle of type {}"
.format(location_handle.__class__))
executable_path = (location_handle.executable_path if isinstance(
location_handle, PythonEnvRepositoryLocationHandle) else
sys.executable)
parts = (
[executable_path, '-m', 'dagster', 'api', 'execute_run', output_file] +
xplat_shlex_split(pointer.get_cli_args()) + [
'--instance-ref',
'{instance_ref}'.format(
instance_ref=serialize_dagster_namedtuple(instance.get_ref())),
'--pipeline-run-id',
'{pipeline_run_id}'.format(pipeline_run_id=pipeline_run.run_id),
])
instance.report_engine_event(
'About to start process for pipeline "{pipeline_name}" (run_id: {run_id}).'
.format(pipeline_name=pipeline_run.pipeline_name,
run_id=pipeline_run.run_id),
pipeline_run,
engine_event_data=EngineEventData(
marker_start='cli_api_subprocess_init'),
)
return open_ipc_subprocess(parts)
def cli_api_execute_run(output_file, instance, pipeline_origin, pipeline_run):
check.str_param(output_file, 'output_file')
check.inst_param(instance, 'instance', DagsterInstance)
check.inst_param(pipeline_origin, 'pipeline_origin', PipelinePythonOrigin)
check.inst_param(pipeline_run, 'pipeline_run', PipelineRun)
from dagster.cli.api import ExecuteRunArgsLoadComplete
with safe_tempfile_path() as input_file:
write_unary_input(
input_file,
ExecuteRunArgs(
pipeline_origin=pipeline_origin,
pipeline_run_id=pipeline_run.run_id,
instance_ref=instance.get_ref(),
),
)
parts = [
pipeline_origin.executable_path,
'-m',
'dagster',
'api',
'execute_run',
input_file,
output_file,
]
instance.report_engine_event(
'About to start process for pipeline "{pipeline_name}" (run_id: {run_id}).'
.format(pipeline_name=pipeline_run.pipeline_name,
run_id=pipeline_run.run_id),
pipeline_run,
engine_event_data=EngineEventData(
marker_start='cli_api_subprocess_init'),
)
process = open_ipc_subprocess(parts)
# we need to process this event in order to ensure that the called process loads the input
event = next(ipc_read_event_stream(output_file))
check.inst(event, ExecuteRunArgsLoadComplete)
return process
def _execute_schedule(graphene_info, pipeline_def, execution_params, errors):
instance = graphene_info.context.instance
execution_plan = None
if is_config_valid(pipeline_def, execution_params.environment_dict, execution_params.mode):
execution_plan = create_execution_plan(
pipeline_def, execution_params.environment_dict, mode=execution_params.mode,
)
execution_plan_snapshot = None
if execution_plan:
execution_plan_snapshot = snapshot_from_execution_plan(
execution_plan, pipeline_def.get_pipeline_snapshot_id()
)
pipeline_run = instance.create_run(
pipeline_name=pipeline_def.name,
environment_dict=execution_params.environment_dict,
mode=execution_params.mode,
solid_subset=(
execution_params.selector.solid_subset
if execution_params.selector is not None
else None
),
tags=execution_params.execution_metadata.tags,
pipeline_snapshot=pipeline_def.get_pipeline_snapshot(),
execution_plan_snapshot=execution_plan_snapshot,
)
# Inject errors into event log at this point
if len(errors) > 0:
for error in errors:
instance.report_engine_event(
error.message, pipeline_run, EngineEventData.engine_error(error)
)
# Launch run if run launcher is defined
run_launcher = graphene_info.context.instance.run_launcher
if run_launcher:
result = _launch_pipeline_execution_for_created_run(graphene_info, pipeline_run.run_id)
else:
result = _start_pipeline_execution_for_created_run(graphene_info, pipeline_run.run_id)
return pipeline_run, result
def launch_run(self, instance, run, external_pipeline):
check.inst_param(instance, 'instance', DagsterInstance)
check.inst_param(run, 'run', PipelineRun)
check.inst_param(external_pipeline, 'external_pipeline', ExternalPipeline)
job_name = 'dagster-run-{}'.format(run.run_id)
pod_name = job_name
job = construct_dagster_graphql_k8s_job(
self.job_config,
args=[
'-p',
'executeRunInProcess',
'-v',
seven.json.dumps(
{
'runId': run.run_id,
'repositoryName': external_pipeline.handle.repository_name,
'repositoryLocationName': external_pipeline.handle.location_name,
}
),
],
job_name=job_name,
pod_name=pod_name,
component='runmaster',
)
api = kubernetes.client.BatchV1Api()
api.create_namespaced_job(body=job, namespace=self.job_namespace)
instance.report_engine_event(
'Kubernetes runmaster job launched',
run,
EngineEventData(
[
EventMetadataEntry.text(job_name, 'Kubernetes Job name'),
EventMetadataEntry.text(pod_name, 'Kubernetes Pod name'),
EventMetadataEntry.text(self.job_namespace, 'Kubernetes Namespace'),
EventMetadataEntry.text(run.run_id, 'Run ID'),
]
),
cls=K8sRunLauncher,
)
return run
def report_engine_event(self,
cls,
message,
pipeline_run,
engine_event_data=None):
'''
Report a EngineEvent that occurred outside of a pipeline execution context.
'''
from dagster.core.events import EngineEventData, DagsterEvent, DagsterEventType
from dagster.core.events.log import DagsterEventRecord
check.class_param(cls, 'cls')
check.str_param(message, 'message')
check.inst_param(pipeline_run, 'pipeline_run', PipelineRun)
engine_event_data = check.opt_inst_param(
engine_event_data,
'engine_event_data',
EngineEventData,
EngineEventData([]),
)
message = "[{}] {}".format(cls.__name__, message)
log_level = logging.INFO
if engine_event_data and engine_event_data.error:
log_level = logging.ERROR
event_record = DagsterEventRecord(
message=message,
user_message=message,
level=log_level,
pipeline_name=pipeline_run.pipeline_name,
run_id=pipeline_run.run_id,
error_info=None,
timestamp=time.time(),
dagster_event=DagsterEvent(
event_type_value=DagsterEventType.ENGINE_EVENT.value,
pipeline_name=pipeline_run.pipeline_name,
message=message,
event_specific_data=engine_event_data,
),
)
self.handle_new_event(event_record)
def _execute_run(request):
try:
execute_run_args = deserialize_json_to_dagster_namedtuple(
request.serialized_execute_run_args)
check.inst_param(execute_run_args, 'execute_run_args', ExecuteRunArgs)
recon_pipeline = recon_pipeline_from_origin(
execute_run_args.pipeline_origin)
instance = DagsterInstance.from_ref(execute_run_args.instance_ref)
pipeline_run = instance.get_run_by_id(execute_run_args.pipeline_run_id)
pid = os.getpid()
except: # pylint: disable=bare-except
yield IPCErrorMessage(
serializable_error_info=serializable_error_info_from_exc_info(
sys.exc_info()),
message='Error during RPC setup for ExecuteRun',
)
return
yield instance.report_engine_event(
'Started process for pipeline (pid: {pid}).'.format(pid=pid),
pipeline_run,
EngineEventData.in_process(pid, marker_end='cli_api_subprocess_init'),
)
# This is so nasty but seemingly unavoidable
# https://amir.rachum.com/blog/2017/03/03/generator-cleanup/
closed = False
try:
for event in _core_execute_run(recon_pipeline, pipeline_run, instance):
yield event
except GeneratorExit:
closed = True
raise
finally:
if not closed:
yield instance.report_engine_event(
'Process for pipeline exited (pid: {pid}).'.format(pid=pid),
pipeline_run,
)
def test_event_log_storage_store_events_and_wipe(self, storage):
assert len(storage.get_logs_for_run(DEFAULT_RUN_ID)) == 0
storage.store_event(
EventRecord(
None,
"Message2",
"debug",
"",
DEFAULT_RUN_ID,
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
"nonce",
event_specific_data=EngineEventData.in_process(999),
),
))
assert len(storage.get_logs_for_run(DEFAULT_RUN_ID)) == 1
assert storage.get_stats_for_run(DEFAULT_RUN_ID)
storage.wipe()
assert len(storage.get_logs_for_run(DEFAULT_RUN_ID)) == 0
def test_in_memory_event_log_storage_store_events_and_wipe():
storage = InMemoryEventLogStorage()
assert len(storage.get_logs_for_run('foo')) == 0
storage.store_event(
DagsterEventRecord(
None,
'Message2',
'debug',
'',
'foo',
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
'nonce',
event_specific_data=EngineEventData.in_process(999),
),
))
assert len(storage.get_logs_for_run('foo')) == 1
storage.wipe()
assert len(storage.get_logs_for_run('foo')) == 0
def test_event_log_storage_store_events_and_wipe(event_storage_factory_cm_fn):
with event_storage_factory_cm_fn() as storage:
assert len(storage.get_logs_for_run('foo')) == 0
storage.store_event(
DagsterEventRecord(
None,
'Message2',
'debug',
'',
'foo',
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
'nonce',
event_specific_data=EngineEventData.in_process(999),
),
))
assert len(storage.get_logs_for_run('foo')) == 1
assert storage.get_stats_for_run('foo')
storage.wipe()
assert len(storage.get_logs_for_run('foo')) == 0
def test_filesystem_event_log_storage_store_events_and_wipe():
with seven.TemporaryDirectory() as tmpdir_path:
storage = SqliteEventLogStorage(tmpdir_path)
assert len(storage.get_logs_for_run('foo')) == 0
storage.store_event(
DagsterEventRecord(
None,
'Message2',
'debug',
'',
'foo',
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
'nonce',
event_specific_data=EngineEventData.in_process(999),
),
))
assert len(storage.get_logs_for_run('foo')) == 1
storage.wipe()
assert len(storage.get_logs_for_run('foo')) == 0
def test_event_log_delete(event_storage_factory_cm_fn):
with event_storage_factory_cm_fn() as storage:
assert len(storage.get_logs_for_run("foo")) == 0
storage.store_event(
EventRecord(
None,
"Message2",
"debug",
"",
"foo",
time.time(),
dagster_event=DagsterEvent(
DagsterEventType.ENGINE_EVENT.value,
"nonce",
event_specific_data=EngineEventData.in_process(999),
),
))
assert len(storage.get_logs_for_run("foo")) == 1
assert storage.get_stats_for_run("foo")
storage.delete_events("foo")
assert len(storage.get_logs_for_run("foo")) == 0
def start_run(self, execute_run_args):
check.inst_param(execute_run_args, "execute_run_args", ExecuteExternalPipelineArgs)
with DagsterInstance.from_ref(execute_run_args.instance_ref) as instance:
try:
res = self._query(
"StartRun",
api_pb2.StartRunRequest,
serialized_execute_run_args=serialize_dagster_namedtuple(execute_run_args),
)
return deserialize_json_to_dagster_namedtuple(res.serialized_start_run_result)
except Exception: # pylint: disable=bare-except
pipeline_run = instance.get_run_by_id(execute_run_args.pipeline_run_id)
instance.report_engine_event(
message="Unexpected error in IPC client",
pipeline_run=pipeline_run,
engine_event_data=EngineEventData.engine_error(
serializable_error_info_from_exc_info(sys.exc_info())
),
)
raise
