def test_report_with_failed_finish_message_argument(unused_tcp_port):
host = "localhost"
url = f"ws://{host}:{unused_tcp_port}"
reporter = Event(evaluator_url=url)
job1 = Job({"name": "job1", "stdout": "stdout", "stderr": "stderr"}, 0)
lines = []
with _mock_ws_thread(host, unused_tcp_port, lines):
reporter.report(Init([job1], 1, 19, ee_id="ee_id", real_id=0, step_id=0))
reporter.report(Running(job1, 100, 10))
reporter.report(Finish().with_error("massive_failure"))
assert len(lines) == 1
def test_report_only_job_running_for_successful_run(unused_tcp_port):
host = "localhost"
url = f"ws://{host}:{unused_tcp_port}"
reporter = Event(evaluator_url=url)
job1 = Job({"name": "job1", "stdout": "stdout", "stderr": "stderr"}, 0)
lines = []
with _mock_ws_thread(host, unused_tcp_port, lines):
reporter.report(Init([job1], 1, 19, ee_id="ee_id", real_id=0, step_id=0))
reporter.report(Running(job1, 100, 10))
reporter.report(Finish())
assert len(lines) == 1
def test_report_with_successful_finish_message_argument(tmpdir):
reporter = Event(event_log=tmpdir / "event_log")
job1 = Job({"name": "job1", "stdout": "stdout", "stderr": "stderr"}, 0)
reporter.report(Init([job1], 1, 19, ee_id="ee_id", real_id=0, stage_id=0))
reporter.report(Running(job1, 100, 10))
reporter.report(Finish())
with open(reporter._event_log, "r") as f:
lines = f.readlines()
assert len(lines) == 3
event = json.loads(lines[2])
assert event["type"] == _FM_STEP_SUCCESS
def test_report_with_failed_finish_message_argument(tmpdir):
reporter = Event(event_log=tmpdir / "event_log")
job1 = Job({"name": "job1", "stdout": "stdout", "stderr": "stderr"}, 0)
reporter.report(Init([job1], 1, 19, ee_id="ee_id", real_id=0, stage_id=0))
reporter.report(Running(job1, 100, 10))
reporter.report(Finish().with_error("massive_failure"))
with open(reporter._event_log, "r") as f:
lines = f.readlines()
assert len(lines) == 3
event = json.loads(lines[2])
assert event["type"] == _FM_STEP_FAILURE
assert event["data"]["error_msg"] == "massive_failure"
def test_report_with_running_message_argument(tmpdir):
reporter = Event(event_log=tmpdir / "event_log")
job1 = Job({"name": "job1", "stdout": "stdout", "stderr": "stderr"}, 0)
reporter.report(Init([job1], 1, 19, ee_id="ee_id", real_id=0, stage_id=0))
reporter.report(Running(job1, 100, 10))
with open(reporter._event_log, "r") as f:
lines = f.readlines()
assert len(lines) == 2
event = json.loads(lines[1])
assert event["type"] == _FM_JOB_RUNNING
assert event["data"]["max_memory_usage"] == 100
assert event["data"]["current_memory_usage"] == 10
def test_report_with_running_message_argument(self):
msg = Running(Job({"name": "job1"}, 0), 100, 10)
self.reporter.status_dict = self.reporter._init_job_status_dict(
msg.timestamp, 0, [msg.job])
self.reporter.report(msg)
with open(self.reporter.STATUS_json, "r") as f:
contents = "".join(f.readlines())
self.assertIn('"status": "Running"', contents,
"status.json missing status")
self.assertIn('"max_memory_usage": 100', contents,
"status.json missing max_memory_usage")
self.assertIn('"current_memory_usage": 10', contents,
"status.json missing current_memory_usage")
def test_report_with_running_message_argument(unused_tcp_port):
host = "localhost"
url = f"ws://{host}:{unused_tcp_port}"
reporter = Event(evaluator_url=url)
job1 = Job({"name": "job1", "stdout": "stdout", "stderr": "stderr"}, 0)
lines = []
with _mock_ws_thread(host, unused_tcp_port, lines):
reporter.report(Init([job1], 1, 19, ee_id="ee_id", real_id=0, step_id=0))
reporter.report(Running(job1, 100, 10))
reporter.report(Finish())
assert len(lines) == 1
event = json.loads(lines[0])
assert event["type"] == _FM_JOB_RUNNING
assert event["data"]["max_memory_usage"] == 100
assert event["data"]["current_memory_usage"] == 10
def run(self):
start_message = Start(self)
errors = self._check_job_files()
errors.extend(self._assert_arg_list())
self._dump_exec_env()
if errors:
yield start_message.with_error("\n".join(errors))
return
yield start_message
executable = self.job_data.get("executable")
assert_file_executable(executable)
arg_list = [executable]
if self.job_data.get("argList"):
arg_list += self.job_data["argList"]
if self.job_data.get("stdin"):
stdin = open(self.job_data.get("stdin"))
else:
stdin = None
if self.std_err:
stderr = open(self.std_err, "w")
else:
stderr = None
if self.std_out:
stdout = open(self.std_out, "w")
else:
stdout = None
if self.job_data.get("target_file"):
target_file_mtime = 0
if os.path.exists(self.job_data["target_file"]):
stat = os.stat(self.job_data["target_file"])
target_file_mtime = stat.st_mtime
exec_env = self.job_data.get("exec_env")
if exec_env:
exec_name, _ = os.path.splitext(
os.path.basename(self.job_data.get("executable"))
)
with open("%s_exec_env.json" % exec_name, "w") as f:
f.write(json.dumps(exec_env))
max_running_minutes = self.job_data.get("max_running_minutes")
run_start_time = dt.now()
proc = Popen(
arg_list,
stdin=stdin,
stdout=stdout,
stderr=stderr,
env=self.job_data.get("environment"),
)
exit_code = None
process = Process(proc.pid)
max_memory_usage = 0
while exit_code is None:
try:
memory = process.memory_info().rss
except (NoSuchProcess, AccessDenied, ZombieProcess):
"""In case of a process that has died and is in some
transitional state, we ignore any failures. Only seen on OSX
thus far.
See https://github.com/giampaolo/psutil/issues/1044#issuecomment-298745532
"""
memory = 0
if memory > max_memory_usage:
max_memory_usage = memory
yield Running(self, max_memory_usage, memory)
try:
exit_code = process.wait(timeout=self.MEMORY_POLL_PERIOD)
except TimeoutExpired:
run_time = dt.now() - run_start_time
if (
max_running_minutes is not None
and run_time.seconds > max_running_minutes * 60
):
"""
If the spawned process is not in the same process group
as the callee (job_dispatch), we will kill the process
group explicitly.
Propagating the unsuccessful Exited message will kill the
callee group. See job_dispatch.py.
"""
process_group_id = os.getpgid(proc.pid)
this_group_id = os.getpgid(os.getpid())
if process_group_id != this_group_id:
os.killpg(process_group_id, signal.SIGKILL)
yield Exited(self, exit_code).with_error(
"Job:{} has been running for more than {} minutes - explicitly killed.".format(
self.name(), max_running_minutes
)
)
return
exited_message = Exited(self, exit_code)
if exit_code != 0:
yield exited_message.with_error(
"Process exited with status code {}".format(exit_code)
)
return
# exit_code is 0
if self.job_data.get("error_file"):
if os.path.exists(self.job_data["error_file"]):
yield exited_message.with_error(
"Found the error file:{} - job failed.".format(
self.job_data["error_file"]
)
)
return
if self.job_data.get("target_file"):
target_file_error = self._check_target_file_is_written(target_file_mtime)
if target_file_error:
yield exited_message.with_error(target_file_error)
return
yield exited_message