def assert_correct_status_calls(cls, api):
# getTasksStatus gets called a lot of times. The exact number isn't fixed; it loops
# over the health checks until all of them pass for a configured period of time.
# The minumum number of calls is 5: once before the tasks are restarted, and then
# once for each batch of restarts (Since the batch size is set to 5, and the
# total number of jobs is 20, that's 4 batches.)
assert api.getTasksStatus.call_count >= 5
# In the first getStatus call, it uses an expansive query; in the rest, it only queries for
# status RUNNING.
status_calls = api.getTasksStatus.call_args_list
assert status_calls[0][0][0] == TaskQuery(taskIds=None,
jobName='hello',
environment='test',
owner=Identity(
role=u'mchucarroll',
user=None),
statuses=cls.QUERY_STATUSES)
for status_call in status_calls[1:]:
status_call[0][0] == TaskQuery(taskIds=None,
jobName='hello',
environment='test',
owner=Identity(role='mchucarroll',
user=None),
statuses=set(
[ScheduleStatus.RUNNING]))
def get_expected_task_query(cls, shards=None):
instance_ids = frozenset(shards) if shards is not None else None
# Helper to create the query that will be a parameter to job kill.
return TaskQuery(taskIds=None,
jobName=cls.TEST_JOB,
environment=cls.TEST_ENV,
instanceIds=instance_ids,
owner=Identity(role=cls.TEST_ROLE, user=None))
def __init__(self, client, role, env, jobname):
self._client = client
self._query = TaskQuery(owner=Identity(role=role),
environment=env,
jobName=jobname)
self._initial_tasks = set()
self._initial_tasks = set(task.assignedTask.taskId
for task in self.iter_query())
def assert_correct_killtask_calls(cls, api):
assert api.killTasks.call_count == 4
# Check the last call's parameters.
api.killTasks.assert_called_with(
TaskQuery(taskIds=None,
jobName='hello',
environment='test',
instanceIds=frozenset([16, 17, 18, 19, 15]),
owner=Identity(role=u'mchucarroll', user=None),
statuses=cls.QUERY_STATUSES), 'foo')
def scheduler_delete_recovery_tasks(cluster, task_ids):
"""usage: scheduler_delete_recovery_tasks cluster task_ids
Deletes a comma-separated list of task IDs from a staged recovery.
"""
ids = set(task_ids.split(','))
options = app.get_options()
check_and_log_response(
AuroraClientAPI(CLUSTERS[cluster],
options.verbosity).delete_recovery_tasks(
TaskQuery(taskIds=ids)))
def expect_kill(self, instance_ids, response_code=None):
response_code = ResponseCode.OK if response_code is None else response_code
response = Response(responseCode=response_code, message='test')
query = TaskQuery(owner=Identity(role=self._job_key.role),
environment=self._job_key.environment,
jobName=self._job_key.name,
statuses=ACTIVE_STATES,
instanceIds=frozenset([int(s)
for s in instance_ids]))
self._scheduler.killTasks(query, self._lock,
self._session_key).AndReturn(response)
def test_successful_run(self):
"""Test the run command."""
# Calls api.check_status, which calls scheduler.getJobs
mock_options = self.setup_mock_options()
(mock_api, mock_scheduler) = self.create_mock_api()
mock_scheduler.getTasksStatus.return_value = self.create_status_response(
)
sandbox_args = {
'slave_root': '/slaveroot',
'slave_run_directory': 'slaverun'
}
with contextlib.nested(
patch('twitter.aurora.client.api.SchedulerProxy',
return_value=mock_scheduler),
patch('twitter.aurora.client.factory.CLUSTERS',
new=self.TEST_CLUSTERS),
patch('twitter.aurora.client.commands.run.CLUSTERS',
new=self.TEST_CLUSTERS),
patch('twitter.common.app.get_options',
return_value=mock_options),
patch(
'twitter.aurora.client.api.command_runner.DistributedCommandRunner.sandbox_args',
return_value=sandbox_args),
patch('subprocess.Popen',
return_value=self.create_mock_process())) as (
mock_scheduler_proxy_class, mock_clusters,
mock_clusters_runpatch, options,
mock_runner_args_patch, mock_subprocess):
run(['west/mchucarroll/test/hello', 'ls'], mock_options)
# The status command sends a getTasksStatus query to the scheduler,
# and then prints the result.
mock_scheduler.getTasksStatus.assert_called_with(
TaskQuery(jobName='hello',
environment='test',
owner=Identity(role='mchucarroll'),
statuses=set([
ScheduleStatus.RUNNING, ScheduleStatus.KILLING,
ScheduleStatus.RESTARTING,
ScheduleStatus.PREEMPTING
])))
# The mock status call returns 3 three ScheduledTasks, so three commands should have been run
assert mock_subprocess.call_count == 3
mock_subprocess.assert_called_with([
'ssh', '-n', '-q', 'mchucarroll@slavehost',
'cd /slaveroot/slaves/*/frameworks/*/executors/thermos-1287391823/runs/'
'slaverun/sandbox;ls'
],
stderr=-2,
stdout=-1)
def test_init(self):
result = Result(scheduleStatusResult=ScheduleStatusResult(tasks=[]))
response = Response(responseCode=ResponseCode.OK,
message="test",
result=result)
query = TaskQuery(owner=Identity(role=ROLE),
environment=ENV,
jobName=JOB_NAME)
self.mock_scheduler.getTasksStatus(query).AndReturn(response)
self.mox.ReplayAll()
JobMonitor(self.mock_api, ROLE, ENV, JOB_NAME)
def expect_get_tasks(self, tasks, ignore_ids=None, response_code=None):
response_code = ResponseCode.OK if response_code is None else response_code
response = Response(responseCode=response_code, message='test')
scheduled = []
for index, task in enumerate(tasks):
if not ignore_ids or index not in ignore_ids:
scheduled.append(
ScheduledTask(assignedTask=AssignedTask(task=task,
instanceId=index)))
response.result = Result(scheduleStatusResult=ScheduleStatusResult(
tasks=scheduled))
query = TaskQuery(owner=Identity(role=self._job_key.role),
environment=self._job_key.environment,
jobName=self._job_key.name,
statuses=ACTIVE_STATES)
self._scheduler.getTasksStatus(query).AndReturn(response)
def test_unsuccessful_status(self):
"""Test the status command when the user asks the status of a job that doesn't exist."""
# Calls api.check_status, which calls scheduler.getJobs
mock_options = self.setup_mock_options()
(mock_api, mock_scheduler) = self.create_mock_api()
mock_scheduler.getTasksStatus.return_value = self.create_failed_status_response()
with contextlib.nested(
patch('twitter.aurora.client.api.SchedulerProxy', return_value=mock_scheduler),
patch('twitter.aurora.client.factory.CLUSTERS', new=self.TEST_CLUSTERS),
patch('twitter.common.app.get_options', return_value=mock_options)) as (
mock_scheduler_proxy_class,
mock_clusters,
options):
self.assertRaises(SystemExit, status, ['west/mchucarroll/test/hello'], mock_options)
mock_scheduler.getTasksStatus.assert_called_with(TaskQuery(jobName='hello',
environment='test', owner=Identity(role='mchucarroll')))
def scheduler_print_recovery_tasks(cluster):
"""usage: scheduler_print_recovery_tasks cluster
Prints all active tasks in a staged recovery.
"""
options = app.get_options()
resp = AuroraClientAPI(CLUSTERS[cluster],
options.verbosity).query_recovery(
TaskQuery(statuses=ACTIVE_STATES))
check_and_log_response(resp)
log.info('Role\tJob\tShard\tStatus\tTask ID')
for task in resp.tasks:
assigned = task.assignedTask
conf = assigned.task
log.info('\t'.join(
(conf.owner.role, conf.jobName, str(assigned.instanceId),
ScheduleStatus._VALUES_TO_NAMES[task.status], assigned.taskId)))
def test_successful_status(self):
"""Test the status command."""
# Calls api.check_status, which calls scheduler.getJobs
mock_options = self.setup_mock_options()
(mock_api, mock_scheduler) = self.create_mock_api()
mock_scheduler.getTasksStatus.return_value = self.create_status_response()
with contextlib.nested(
patch('twitter.aurora.client.api.SchedulerProxy', return_value=mock_scheduler),
patch('twitter.aurora.client.factory.CLUSTERS', new=self.TEST_CLUSTERS),
patch('twitter.common.app.get_options', return_value=mock_options)) as (
mock_scheduler_proxy_class,
mock_clusters,
options):
status(['west/mchucarroll/test/hello'], mock_options)
# The status command sends a getTasksStatus query to the scheduler,
# and then prints the result.
mock_scheduler.getTasksStatus.assert_called_with(TaskQuery(jobName='hello',
environment='test', owner=Identity(role='mchucarroll')))
def _get_tasks_by_instance_id(self, instance_ids):
log.debug('Querying instance statuses.')
query = TaskQuery()
query.owner = Identity(role=self._job_key.role)
query.environment = self._job_key.environment
query.jobName = self._job_key.name
query.statuses = set([ScheduleStatus.RUNNING])
query.instanceIds = instance_ids
try:
resp = self._scheduler.getTasksStatus(query)
except IOError as e:
log.error('IO Exception during scheduler call: %s' % e)
return []
tasks = []
if resp.responseCode == ResponseCode.OK:
tasks = resp.result.scheduleStatusResult.tasks
log.debug(
'Response from scheduler: %s (message: %s)' %
(ResponseCode._VALUES_TO_NAMES[resp.responseCode], resp.message))
return tasks
def _get_tasks_by_instance_id(self, instance_ids):
log.debug('Querying instance statuses.')
query = TaskQuery()
query.owner = Identity(role=self._job_key.role)
query.environment = self._job_key.environment
query.jobName = self._job_key.name
query.statuses = set([ScheduleStatus.RUNNING])
query.instanceIds = instance_ids
try:
resp = self._scheduler.getTasksStatus(query)
except IOError as e:
log.error('IO Exception during scheduler call: %s' % e)
return []
tasks = []
if resp.responseCode == ResponseCode.OK:
tasks = resp.result.scheduleStatusResult.tasks
log.debug('Response from scheduler: %s (message: %s)'
% (ResponseCode._VALUES_TO_NAMES[resp.responseCode], resp.message))
return tasks
def create_mock_query(cls):
return TaskQuery(owner=Identity(role=cls.TEST_ROLE),
environment=cls.TEST_ENV,
jobName=cls.TEST_JOB)
def _create_task_query(self, instanceIds=None):
return TaskQuery(owner=Identity(role=self._job_key.role),
environment=self._job_key.environment,
jobName=self._job_key.name,
statuses=ACTIVE_STATES,
instanceIds=instanceIds)
def to_thrift_query(self): return TaskQuery(owner=Identity(role=self.role), environment=self.env, jobName=self.name)
def build_query(cls, role, job, instances=None, statuses=LIVE_STATES, env=None):
return TaskQuery(owner=Identity(role=role),
jobName=job,
statuses=statuses,
instanceIds=instances,
environment=env)
def query_from(cls, role, env, job):
return TaskQuery(statuses=LIVE_STATES,
owner=Identity(role),
jobName=job,
environment=env)
