def test_healthy():
with instance_for_test(
overrides={
"run_coordinator": {
"module":
"dagster.core.run_coordinator.queued_run_coordinator",
"class": "QueuedRunCoordinator",
},
}) as instance:
init_time = pendulum.now("UTC")
beyond_tolerated_time = init_time.float_timestamp + 100
controller = DagsterDaemonController(instance)
assert not all_daemons_healthy(
instance, curr_time_seconds=init_time.float_timestamp)
assert not all_daemons_live(
instance, curr_time_seconds=init_time.float_timestamp)
controller.run_iteration(init_time)
assert all_daemons_healthy(instance,
curr_time_seconds=init_time.float_timestamp)
assert all_daemons_live(instance,
curr_time_seconds=init_time.float_timestamp)
assert not all_daemons_healthy(instance,
curr_time_seconds=beyond_tolerated_time)
assert not all_daemons_live(instance,
curr_time_seconds=beyond_tolerated_time)
def run_command():
with DagsterInstance.get() as instance:
controller = DagsterDaemonController(instance)
while True:
curr_time = pendulum.now("UTC")
controller.run_iteration(curr_time)
time.sleep(0.5)
def run_command():
with DagsterInstance.get() as instance:
if instance.is_ephemeral:
raise Exception(
"dagster-daemon can't run using an in-memory instance. Make sure "
"the DAGSTER_HOME environment variable has been set correctly and that "
"you have created a dagster.yaml file there.")
controller = DagsterDaemonController(instance)
while True:
curr_time = pendulum.now("UTC")
controller.run_iteration(curr_time)
time.sleep(0.5)
def test_healthy_with_different_daemons():
with instance_for_test() as instance:
init_time = pendulum.now("UTC")
controller = DagsterDaemonController(instance)
controller.run_iteration(init_time)
with instance_for_test(
overrides={
"run_coordinator": {
"module": "dagster.core.run_coordinator.queued_run_coordinator",
"class": "QueuedRunCoordinator",
},
}
) as instance:
assert not all_daemons_healthy(instance, curr_time=init_time)
def test_error_daemon(monkeypatch):
with instance_for_test(overrides={}) as instance:
from dagster.daemon.daemon import SensorDaemon
def run_iteration_error(_):
raise DagsterInvariantViolationError("foobar")
monkeypatch.setattr(SensorDaemon, "run_iteration", run_iteration_error)
controller = DagsterDaemonController(instance)
init_time = pendulum.now("UTC")
controller.run_iteration(init_time)
status = get_daemon_status(instance, SensorDaemon.daemon_type(),
init_time.float_timestamp)
assert status.healthy == False
assert (status.last_heartbeat.error.message.strip() ==
"dagster.core.errors.DagsterInvariantViolationError: foobar")
def test_multiple_error_daemon(monkeypatch):
with instance_for_test(overrides={}) as instance:
from dagster.daemon.daemon import SensorDaemon
def run_iteration_error(_):
# ?message stack cls_name cause"
yield SerializableErrorInfo("foobar", None, None, None)
yield SerializableErrorInfo("bizbuz", None, None, None)
monkeypatch.setattr(SensorDaemon, "run_iteration", run_iteration_error)
controller = DagsterDaemonController(instance)
init_time = pendulum.now("UTC")
controller.run_iteration(init_time)
status = get_daemon_status(instance, SensorDaemon.daemon_type(),
init_time.float_timestamp)
assert status.healthy == False
assert len(status.last_heartbeat.errors) == 2
assert status.last_heartbeat.errors[0].message.strip() == "foobar"
assert status.last_heartbeat.errors[1].message.strip() == "bizbuz"
def test_healthy():
with instance_for_test(
overrides={
"run_coordinator": {
"module":
"dagster.core.run_coordinator.queued_run_coordinator",
"class": "QueuedRunCoordinator",
},
}) as instance:
init_time = pendulum.now("UTC")
beyond_tolerated_time = init_time + datetime.timedelta(seconds=60)
controller = DagsterDaemonController(instance)
assert not controller.daemon_healthy(instance, curr_time=init_time)
controller.run_iteration(init_time)
assert controller.daemon_healthy(instance, curr_time=init_time)
assert not controller.daemon_healthy(instance,
curr_time=beyond_tolerated_time)
def test_different_intervals(caplog):
with instance_for_test(
overrides={
"scheduler": {
"module": "dagster.core.scheduler",
"class": "DagsterDaemonScheduler",
},
"run_coordinator": {
"module":
"dagster.core.run_coordinator.queued_run_coordinator",
"class": "QueuedRunCoordinator",
"config": {
"dequeue_interval_seconds": 5
},
},
}) as instance:
init_time = pendulum.now("UTC")
controller = DagsterDaemonController(instance)
assert caplog.record_tuples == [(
"dagster-daemon",
logging.INFO,
"instance is configured with the following daemons: ['QueuedRunCoordinatorDaemon', 'SchedulerDaemon', 'SensorDaemon']",
)]
controller.run_iteration(init_time)
scheduler_daemon = controller.get_daemon(SchedulerDaemon.__name__)
run_daemon = controller.get_daemon(QueuedRunCoordinatorDaemon.__name__)
assert scheduler_daemon
assert scheduler_daemon.last_iteration_time == init_time
assert _scheduler_ran(caplog)
assert run_daemon
assert run_daemon.last_iteration_time == init_time
assert _run_coordinator_ran(caplog)
caplog.clear()
next_time = init_time + datetime.timedelta(seconds=5)
controller.run_iteration(next_time)
# Run coordinator does another iteration, scheduler does not
assert scheduler_daemon.last_iteration_time == init_time
assert not _scheduler_ran(caplog)
assert run_daemon.last_iteration_time == next_time
assert _run_coordinator_ran(caplog)
caplog.clear()
next_time = init_time + datetime.timedelta(seconds=30)
controller.run_iteration(next_time)
# 30 seconds later both daemons do another iteration
assert scheduler_daemon.last_iteration_time == next_time
assert _scheduler_ran(caplog)
assert run_daemon.last_iteration_time == next_time
assert _run_coordinator_ran(caplog)
def test_error_daemon(monkeypatch):
with instance_for_test(overrides={}) as instance:
from dagster.daemon.daemon import SensorDaemon
def run_iteration_error(_):
raise DagsterInvariantViolationError("foobar")
yield # pylint: disable=unreachable
monkeypatch.setattr(SensorDaemon, "run_iteration", run_iteration_error)
controller = DagsterDaemonController(instance)
init_time = pendulum.now("UTC")
controller.run_iteration(init_time)
status = get_daemon_status(instance, SensorDaemon.daemon_type(),
init_time.float_timestamp)
assert status.healthy == False
assert len(status.last_heartbeat.errors) == 1
assert (status.last_heartbeat.errors[0].message.strip() ==
"dagster.core.errors.DagsterInvariantViolationError: foobar")
assert not all_daemons_healthy(
instance, curr_time_seconds=init_time.float_timestamp)
assert all_daemons_live(instance,
curr_time_seconds=init_time.float_timestamp)
def test_warn_multiple_daemons(capsys):
with instance_for_test() as instance:
init_time = pendulum.now("UTC")
next_time = init_time.add(seconds=100)
controller1 = DagsterDaemonController(instance)
controller1.run_iteration(init_time)
captured = capsys.readouterr()
assert "Taking over from another SENSOR daemon process" not in captured.out
controller2 = DagsterDaemonController(instance)
controller2.run_iteration(init_time)
captured = capsys.readouterr()
assert "Taking over from another SENSOR daemon process" not in captured.out
controller1.run_iteration(next_time)
captured = capsys.readouterr()
assert "Taking over from another SENSOR daemon process" in captured.out