def stop_task(self, job_id, instance_id):
"""
param job_id: id of the job
param instance_id: instance id of the task to stop
type job_id: str
type instance_id: int
rtype: task.StopResponse
"""
rng = task.InstanceRange(to=instance_id + 1)
setattr(rng, "from", instance_id)
request = task.StopRequest(
jobId=peloton.JobID(value=job_id), ranges=[rng]
)
try:
print_okblue("Stopping task %d of Job %s" % (instance_id, job_id))
resp = self.client.task_svc.Stop(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
return resp
except Exception as e:
print_fail("Exception calling Stop Tasks :%s" % str(e))
raise
def update_stateless_job(self, job_id, new_job_config):
"""
param job_id: id of the job
param new_job_config: new config of the job
type job_id: str
type new_job_config: job.JobConfig
rtype: job.UpdateResponse
"""
request = update_svc.CreateUpdateRequest(
jobId=peloton.JobID(value=job_id),
jobConfig=new_job_config,
updateConfig=update_pb2.UpdateConfig(),
)
try:
print_okblue("Updating Job %s" % job_id)
resp = self.client.update_svc.CreateUpdate(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
return resp
except Exception as e:
print_fail("Exception calling Update Stateless Job: %s" % str(e))
raise
def stop_job(self):
request = task.StopRequest(jobId=peloton.JobID(value=self.job_id))
self.client.task_svc.Stop(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
def update_job(
self,
instance_inc,
batch_size,
use_instance_config,
sleep_time,
host_limit_1=False,
):
job_info = self.get_job_info()
job_config = job_info.config
if use_instance_config:
instance_config = {}
for i in range(0, instance_inc):
count = job_config.instanceCount + i
instance_config[count] = self.create_pod_config(
sleep_time, "instance %s" % i)
job_config.instanceConfig.MergeFrom(instance_config)
job_config.instanceCount = job_config.instanceCount + instance_inc
request = job.UpdateRequest(id=peloton.JobID(value=self.job_id),
config=job_config)
self.client.job_svc.Update(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
def rolling_restart(self,
ranges=None,
resource_version=None,
batch_size=None):
"""
Restart a job or certain tasks in a rolling fashion based on the ranges
and batch size
:param ranges: the instance ranges to stop
:param resource_version: the resource_version to use,
if not set. the API would fetch if from job runtime
:param batch_size: the batch size of rolling stop
:return: WorkflowResp
"""
if resource_version is None:
job_info = self.get_info()
resource_version = job_info.runtime.configurationVersion
request = job.RestartRequest(
id=peloton.JobID(value=self.job_id),
ranges=ranges,
resourceVersion=resource_version,
restartConfig=job.RestartConfig(batchSize=batch_size),
)
resp = self.client.job_svc.Restart(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
return Job.WorkflowResp(
resp.updateID.value,
resp.resourceVersion,
client=self.client,
config=self.config,
)
def test__delete_active_job(jobs_by_state):
job = jobs_by_state[1]['RUNNING'][0]
job.create()
job.wait_for_state(goal_state='RUNNING')
client = Client()
request = job_pb2.DeleteRequest(
id=peloton.JobID(value=job.job_id),
)
failed = True
try:
client.job_svc.Delete(
request,
metadata=client.jobmgr_metadata,
timeout=10,
)
failed = False
except grpc.RpcError as e:
log.info(e)
errmsg = "Job is not in a terminal state"
assert errmsg in e.details()
assert e.code() is grpc.StatusCode.INTERNAL
job.stop()
job.wait_for_state(goal_state='KILLED')
assert failed is True
def get_job_info(self):
request = job.GetRequest(id=peloton.JobID(value=self.job_id))
resp = self.client.job_svc.Get(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
return resp.jobInfo
def get_pod_events(self, instance_id):
"""
:return: A list of all of the pod events of a given task
"""
request = task.GetPodEventsRequest(
jobId=peloton.JobID(value=self.job_id), instanceId=instance_id)
resp = self.client.task_svc.GetPodEvents(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
return resp.result
def list_tasks(self):
"""
:return: The map of instance ID to task info for all matching tasks
"""
request = task.ListRequest(jobId=peloton.JobID(value=self.job_id))
resp = self.client.task_svc.List(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
assert not resp.HasField("notFound")
return resp.result
def get_info(self):
"""
:return: The job info
"""
request = job.GetRequest(id=peloton.JobID(value=self.job_id))
resp = self.client.job_svc.Get(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
assert not resp.HasField("error")
return resp.jobInfo
def query_request(
job_id, task_state=None, pagination=None, names=None, hosts=None
):
""" Constructs a task.QueryRequest object for task query api. """
job_id_spec = peloton.JobID(value=job_id)
task_states = [task_state] if task_state else []
pagination = pagination if pagination else default_pagination
task_query_spec = task.QuerySpec(
taskStates=task_states, pagination=pagination, names=names, hosts=hosts
)
request = task.QueryRequest(jobId=job_id_spec, spec=task_query_spec)
return request
def create(self, config_version=None):
"""
creates an update based on the job and config.
if config_version is provided, create will use the provided value,
and raise an exception if version is wrong.
if config_version is provided, create will query job runtime to
get config version and retry until version is correct.
:return: the update ID
"""
respool_id = self.pool.ensure_exists()
self.updated_job_config.respoolID.value = respool_id
while True:
job_config_version = self.job.get_runtime().configurationVersion
self.updated_job_config.changeLog.version = (config_version
or job_config_version)
request = update_svc.CreateUpdateRequest(
jobId=peloton.JobID(value=self.job.job_id),
jobConfig=self.updated_job_config,
updateConfig=update.UpdateConfig(
batchSize=self.batch_size,
rollbackOnFailure=self.roll_back_on_failure,
maxInstanceAttempts=self.max_instance_attempts,
maxFailureInstances=self.max_failure_instances,
startPaused=self.start_paused,
),
)
try:
resp = self.client.update_svc.CreateUpdate(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
except grpc.RpcError as e:
# if config version is incorrect and caller does not specify a
# config version, get config version from job runtime
# and try again.
if (e.code() == grpc.StatusCode.ABORTED
and e.details() == INVALID_VERSION_ERR_MESSAGE
and config_version is None):
continue
raise
break
assert resp.updateID.value
self.workflow = Workflow(resp.updateID.value,
client=self.client,
config=self.config)
log.info("created update %s", self.workflow.workflow_id)
def get_task_runs(self, instance_id):
"""
:param instance_id: The instance id of the task
:return: Returns all active and completed tasks of the given instance
"""
request = task.GetRequest(jobId=peloton.JobID(value=self.job_id),
instanceId=instance_id)
resp = self.client.task_svc.Get(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
assert not resp.HasField("notFound")
assert not resp.HasField("outOfRange")
return resp.results
def get_task_info(self, instance_id):
"""
:param instance_id: The instance id of the task
:return: The task info for the instance id
"""
request = task.GetRequest(jobId=peloton.JobID(value=self.job_id),
instanceId=instance_id)
resp = self.client.task_svc.Get(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
assert not resp.HasField("notFound")
assert not resp.HasField("outOfRange")
return resp.result
def test__delete_completed_job(jobs_by_state):
job = jobs_by_state[1]["SUCCEEDED"][0]
job.create()
job.wait_for_state()
client = Client()
request = job_pb2.DeleteRequest(id=peloton.JobID(value=job.job_id))
try:
client.job_svc.Delete(request,
metadata=client.jobmgr_metadata,
timeout=10)
except grpc.RpcError as e:
log.info(e)
assert e is None
def delete(self):
"""
Deletes a job
:return: delete job response from the API
"""
request = job.DeleteRequest(id=peloton.JobID(value=self.job_id))
response = self.client.job_svc.Delete(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
assert not response.HasField("error")
log.info("deleting job {0}".format(self.job_id))
return response
def test__delete_non_existing_job(peloton_client):
client = peloton_client
request = job_pb2.DeleteRequest(id=peloton.JobID(
value="00010203-0405-0607-0809-0a0b0c0d0e0f"))
failed = True
try:
client.job_svc.Delete(request,
metadata=client.jobmgr_metadata,
timeout=10)
failed = False
except grpc.RpcError as e:
log.info(e)
assert e.details() == "job not found"
assert e.code() is grpc.StatusCode.NOT_FOUND
assert failed is True
def stop(self, ranges=None):
"""
Stops a job or certain tasks based on the ranges
:param ranges: the instance ranges to stop
:return: task stop response from the API
"""
request = task.StopRequest(jobId=peloton.JobID(value=self.job_id),
ranges=ranges)
response = self.client.task_svc.Stop(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
log.info("stopping tasks in job {0} with ranges {1}".format(
self.job_id, ranges))
return response
def browse_task_sandbox(self, instance_id, task_id):
"""
:param instance_id: The instance id of the task
:param task_id: The mesos task id of the task
:return: The BrowseSandboxResponse
"""
request = task.BrowseSandboxRequest(
jobId=peloton.JobID(value=self.job_id),
instanceId=instance_id,
taskId=task_id)
resp = self.client.task_svc.BrowseSandbox(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
assert not resp.HasField('error')
return resp
def get_tasks(self, job_id):
"""
param job_id: id of the job
type job_id: str
rtype: job.ListResponse
"""
request = task.ListRequest(jobId=peloton.JobID(value=job_id))
try:
resp = self.client.task_svc.List(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
).result.value
return resp
except Exception as e:
print_fail("Exception calling List Tasks :%s" % str(e))
raise
def get_job(self, job_id):
"""
:param job_id: the id of the job
:type job_id: str
:rtype: Response
"""
request = job.GetRequest(id=peloton.JobID(value=job_id))
try:
resp = self.client.job_svc.Get(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
return resp
except Exception as e:
print_fail("Exception calling Get job :%s" % str(e))
raise
def delete_job(self, job_id):
"""
param job_id: id of the job
type job_id: str
rtype: job.DeleteResponse
"""
request = job.DeleteRequest(id=peloton.JobID(value=job_id))
try:
print_okblue("Deleting job %s" % job_id)
resp = self.client.job_svc.Delete(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
return resp
except Exception as e:
print_fail("Exception calling delete job :%s" % str(e))
raise
def stop_job(self, job_id):
"""
param job_id: id of the job
type job_id: str
rtype: job.StopResponse
"""
request = task.StopRequest(jobId=peloton.JobID(value=job_id))
try:
print_okblue("Killing all tasks of Job %s" % job_id)
resp = self.client.task_svc.Stop(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
return resp
except Exception as e:
print_fail("Exception calling List Tasks :%s" % str(e))
raise
def wait_for_terminated(self):
"""
Waits for the job to be terminated
"""
state = ''
attempts = 0
log.info('%s waiting for terminal state', self.job_id)
terminated = False
while attempts < self.config.max_retry_attempts:
try:
request = job.GetRequest(
id=peloton.JobID(value=self.job_id),
)
resp = self.client.job_svc.Get(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
runtime = resp.jobInfo.runtime
new_state = job.JobState.Name(runtime.state)
if state != new_state:
log.info('%s transitioned to state %s', self.job_id,
new_state)
state = new_state
if state in ['SUCCEEDED', 'FAILED', 'KILLED']:
terminated = True
break
log.debug(format_stats(runtime.taskStats))
except Exception as e:
log.warn(e)
finally:
time.sleep(self.config.sleep_time_sec)
attempts += 1
if terminated:
log.info('%s job terminated', self.job_id)
assert True
if attempts == self.config.max_retry_attempts:
log.info('% max attempts reached to wait for goal state',
self.job_id)
log.info('current_state:%s', state)
assert False
def update(self, new_job_file):
"""
updates a job
:param new_job_file: The job config file used for updating
"""
job_config_dump = load_test_config(new_job_file)
new_job_config = job.JobConfig()
json_format.ParseDict(job_config_dump, new_job_config)
request = job.UpdateRequest(id=peloton.JobID(value=self.job_id),
config=new_job_config)
resp = self.client.job_svc.Update(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
assert not resp.HasField("error")
# update the config
self.job_config = new_job_config
log.info("updated job %s", self.job_id)
def wait_for_leader(self):
"""
utility method to wait for job manger leader to come up.
good practice to check before all write apis
"""
attempts = 0
while attempts < self.config.max_retry_attempts:
try:
request = job.DeleteRequest(id=peloton.JobID(
value="dummy_job_id"))
self.client.job_svc.Delete(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
except grpc.RpcError as e:
if e.code() == grpc.StatusCode.UNAVAILABLE:
time.sleep(self.config.sleep_time_sec)
attempts += 1
continue
break
def start(self, ranges=None):
"""
Starts a job or certain tasks based on the ranges
:param ranges: the instance ranges to start
:return: task start response from the API
"""
# wait for job manager leader
self.wait_for_jobmgr_available()
request = task.StartRequest(
jobId=peloton.JobID(value=self.job_id), ranges=ranges
)
response = self.client.task_svc.Start(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
log.info(
"starting tasks in job {0} with ranges {1}".format(
self.job_id, ranges
)
)
return response
def update_job(self, job_id, new_job_config):
"""
param job_id: id of the job
param new_job_config: new config of the job
type job_id: str
type new_job_config: job.JobConfig
rtype: job.UpdateResponse
"""
request = job.UpdateRequest(
id=peloton.JobID(value=job_id),
config=new_job_config,
)
try:
print_okblue("Updating Job %s" % job_id)
resp = self.client.job_svc.Update(
request,
metadata=self.client.jobmgr_metadata,
timeout=default_timeout,
)
return resp
except Exception, e:
print_fail('Exception calling Update Job: %s' % str(e))
raise
def wait_for_state(self, goal_state="SUCCEEDED", failed_state="FAILED"):
"""
Waits for the job to reach a particular state
:param goal_state: The state to reach
:param failed_state: The failed state of the job
"""
state = ""
attempts = 0
start = time.time()
log.info("%s waiting for state %s", self.job_id, goal_state)
state_transition_failure = False
while attempts < self.config.max_retry_attempts:
try:
request = job.GetRequest(id=peloton.JobID(value=self.job_id))
resp = self.client.job_svc.Get(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
runtime = resp.jobInfo.runtime
new_state = job.JobState.Name(runtime.state)
if state != new_state:
log.info("%s transitioned to state %s", self.job_id,
new_state)
state = new_state
if state == goal_state:
break
log.debug(format_stats(runtime.taskStats))
# If we assert here, we will log the exception,
# and continue with the finally block. Set a flag
# here to indicate failure and then break the loop
# in the finally block
if state == failed_state:
state_transition_failure = True
except Exception as e:
log.warn(e)
finally:
if state_transition_failure:
break
time.sleep(self.config.sleep_time_sec)
attempts += 1
if state_transition_failure:
task_failure = self._get_task_failure()
log.info(
"%s goal_state:%s current_state:%s attempts: %s",
self.job_id,
goal_state,
state,
str(attempts),
)
for t, failure in task_failure.iteritems():
log.info(
"%s task id:%s failed for reason:%s message:%s",
self.job_id,
t,
failure["reason"],
failure["message"],
)
assert False
if attempts == self.config.max_retry_attempts:
log.info("%s max attempts reached to wait for goal state",
self.job_id)
log.info(
"%s goal_state:%s current_state:%s",
self.job_id,
goal_state,
state,
)
assert False
end = time.time()
elapsed = end - start
log.info("%s state transition took %s seconds", self.job_id, elapsed)
assert state == goal_state
