def test__revocable_tasks_move_to_revocable_queue():
revocable_job1 = StatelessJob(job_file='test_stateless_job_revocable_spec.yaml')
revocable_job1.create()
revocable_job1.wait_for_state(goal_state='RUNNING')
revocable_job1.wait_for_all_pods_running()
# 1 task is running out of 3
def partial_tasks_running():
count = 0
for pod_id in range(0, revocable_job2.job_spec.instance_count):
pod_state = revocable_job2.get_pod(pod_id).get_pod_status().state
if pod_state == pod.POD_STATE_RUNNING:
count += 1
return count == 1
revocable_job2 = StatelessJob(
job_file='test_stateless_job_revocable_slack_limit_spec.yaml')
revocable_job2.create()
# sleep for 5 seconds to make sure job has enough time
time.sleep(5)
revocable_job2.wait_for_condition(partial_tasks_running)
non_revocable_job = StatelessJob(job_file='test_stateless_job_spec.yaml')
non_revocable_job.create()
non_revocable_job.wait_for_state('RUNNING')
non_revocable_job.wait_for_all_pods_running()
# cleanup jobs from jobmgr
revocable_job1.stop()
revocable_job2.stop()
non_revocable_job.stop()
revocable_job1.wait_for_terminated()
revocable_job2.wait_for_terminated()
non_revocable_job.wait_for_terminated()
def test__health_check_detects_healthy_tasks():
job = StatelessJob(
job_file='test_stateless_job_successful_health_check_spec.yaml',
config=IntegrationTestConfig(max_retry_attempts=100))
job.job_spec.instance_count = 1
job.create()
job.wait_for_state(goal_state='RUNNING')
def task_has_healthy_events():
for pod_event in job.get_pod(0).get_pod_events():
if pod_event.healthy == 'HEALTHY':
return True
job.wait_for_condition(task_has_healthy_events)
def test__health_check_detects_unhealthy_tasks():
job = StatelessJob(
job_file="test_stateless_job_failed_health_check_spec.yaml",
config=IntegrationTestConfig(max_retry_attempts=100),
)
job.job_spec.instance_count = 1
job.create()
job.wait_for_state(goal_state="RUNNING")
def task_has_unhealthy_events():
for pod_event in job.get_pod(0).get_pod_events():
if pod_event.healthy == "HEALTH_STATE_UNHEALTHY":
return True
job.wait_for_condition(task_has_unhealthy_events)
def test__stop_nonrevocable_job_to_free_resources_for_revocable_job():
non_revocable_job1 = StatelessJob(
job_file="test_stateless_job_memory_large_spec.yaml"
)
non_revocable_job1.create()
non_revocable_job1.wait_for_state("RUNNING")
non_revocable_job2 = StatelessJob(
job_file="test_stateless_preemptible_job_memory_large_spec.yaml"
)
non_revocable_job2.create()
non_revocable_job2.wait_for_state("RUNNING")
non_revocable_job1.wait_for_all_pods_running()
non_revocable_job2.wait_for_all_pods_running()
revocable_job = StatelessJob(
job_file="test_stateless_job_revocable_spec.yaml"
)
revocable_job.create()
# no tasks should be running
def no_task_running():
count = 0
for pod_id in range(0, revocable_job.job_spec.instance_count):
pod_state = revocable_job.get_pod(pod_id).get_pod_status().state
if pod_state == pod.POD_STATE_RUNNING:
count += 1
return count == 0
# give job 5 seconds to run, even after that no tasks should be running
time.sleep(5)
revocable_job.wait_for_condition(no_task_running)
# stop non_revocable job to free up resources for revocable job
non_revocable_job2.stop()
non_revocable_job2.wait_for_terminated()
# After non_revocable job is killed, all revocable tasks should be running
revocable_job.wait_for_all_pods_running()
# cleanup jobs from jobmgr
non_revocable_job1.stop()
revocable_job.stop()
non_revocable_job1.wait_for_terminated()
revocable_job.wait_for_terminated()
def test__failed_task_automatically_restart():
job = StatelessJob(job_file='test_stateless_job_exit_with_err_spec.yaml',
config=IntegrationTestConfig(max_retry_attempts=100))
job.create()
job.wait_for_state(goal_state='RUNNING')
old_pod_id = job.get_pod(0).get_pod_status().pod_id.value
def job_not_running():
return job.get_status().state != 'JOB_STATE_RUNNING'
job.wait_for_condition(job_not_running)
def pod_id_changed():
new_pod_id = job.get_pod(0).get_pod_status().pod_id.value
return old_pod_id != new_pod_id
job.wait_for_condition(pod_id_changed)
def test__revocable_job_slack_limit():
revocable_job = StatelessJob(
job_file="test_stateless_job_revocable_slack_limit_spec.yaml")
revocable_job.create()
revocable_job.wait_for_state(goal_state="RUNNING")
# 2 tasks are running out of 3
def partial_tasks_running():
count = 0
for pod_id in range(0, revocable_job.job_spec.instance_count):
pod_state = revocable_job.get_pod(pod_id).get_pod_status().state
if pod_state == pod.POD_STATE_RUNNING:
count += 1
return count == 2
revocable_job.wait_for_condition(partial_tasks_running)
# cleanup job from jobmgr
revocable_job.stop()
revocable_job.wait_for_terminated()
def test__preempt_revocable_job_to_run_non_revocable_job():
non_revocable_job1 = StatelessJob(
job_file="test_stateless_preemptible_job_memory_large_spec.yaml"
)
non_revocable_job1.create()
non_revocable_job1.wait_for_state(goal_state="RUNNING")
non_revocable_job1.wait_for_all_pods_running()
revocable_job = StatelessJob(
job_file="test_stateless_job_revocable_spec.yaml"
)
revocable_job.create()
revocable_job.wait_for_state(goal_state="RUNNING")
revocable_job.wait_for_all_pods_running()
# launch second non-revocable job which will pre-empt revocable job
non_revocable_job2 = StatelessJob(
job_file="test_stateless_job_memory_large_spec.yaml"
)
non_revocable_job2.create()
non_revocable_job2.wait_for_state(goal_state="RUNNING")
non_revocable_job2.wait_for_all_pods_running()
# no revocable job tasks should be running
def zero_tasks_running():
count = 0
for pod_id in range(0, revocable_job.job_spec.instance_count):
pod_state = revocable_job.get_pod(pod_id).get_pod_status().state
if pod_state == pod.POD_STATE_RUNNING:
count += 1
return count == 0
revocable_job.wait_for_condition(zero_tasks_running)
revocable_job.stop()
non_revocable_job1.stop()
non_revocable_job2.stop()
revocable_job.wait_for_terminated()
non_revocable_job1.wait_for_terminated()
non_revocable_job2.wait_for_terminated()