def test_modify_app_config():
check_healthy()
app_config_field = 'TASKCFG_ALL_CLIENT_READ_SHORTCIRCUIT_STREAMS_CACHE_SIZE_EXPIRY_MS'
journal_ids = tasks.get_task_ids(PACKAGE_NAME, 'journal')
name_ids = tasks.get_task_ids(PACKAGE_NAME, 'name')
zkfc_ids = tasks.get_task_ids(PACKAGE_NAME, 'zkfc')
data_ids = tasks.get_task_ids(PACKAGE_NAME, 'data')
print('journal ids: ' + str(journal_ids))
print('name ids: ' + str(name_ids))
print('zkfc ids: ' + str(zkfc_ids))
print('data ids: ' + str(data_ids))
config = marathon.get_config(PACKAGE_NAME)
print('marathon config: ')
print(config)
expiry_ms = int(config['env'][app_config_field])
config['env'][app_config_field] = str(expiry_ms + 1)
r = cmd.request('put',
marathon.api_url('apps/' + PACKAGE_NAME),
json=config)
# All tasks should be updated because hdfs-site.xml has changed
tasks.check_tasks_updated(PACKAGE_NAME, 'journal', journal_ids)
tasks.check_tasks_updated(PACKAGE_NAME, 'name', name_ids)
tasks.check_tasks_updated(PACKAGE_NAME, 'zkfc', zkfc_ids)
tasks.check_tasks_updated(PACKAGE_NAME, 'data', journal_ids)
check_healthy()
def test_deploy():
wait_time = 30
# taskcfg.yml will initially fail to deploy because several options are missing in the default
# marathon.json.mustache. verify that tasks are failing for 30s before continuing.
print('Checking that tasks are failing to launch for at least {}s'.format(wait_time))
# we can get brief blips of TASK_RUNNING but they shouldnt last more than 2-3s:
consecutive_task_running = 0
def fn():
nonlocal consecutive_task_running
svc_tasks = shakedown.get_service_tasks(PACKAGE_NAME)
states = [t['state'] for t in svc_tasks]
print('Task states: {}'.format(states))
if 'TASK_RUNNING' in states:
consecutive_task_running += 1
assert consecutive_task_running <= 3
else:
consecutive_task_running = 0
return False
try:
spin.time_wait_noisy(lambda: fn(), timeout_seconds=wait_time)
except shakedown.TimeoutExpired:
print('Timeout reached as expected')
# add the needed envvars in marathon and confirm that the deployment succeeds:
config = marathon.get_config(PACKAGE_NAME)
env = config['env']
del env['SLEEP_DURATION']
env['TASKCFG_ALL_OUTPUT_FILENAME'] = 'output'
env['TASKCFG_ALL_SLEEP_DURATION'] = '1000'
cmd.request('put', marathon.api_url('apps/' + PACKAGE_NAME), json=config)
check_running()
def test_modify_app_config_rollback():
check_healthy()
app_config_field = 'TASKCFG_ALL_CLIENT_READ_SHORTCIRCUIT_STREAMS_CACHE_SIZE_EXPIRY_MS'
journal_ids = tasks.get_task_ids(PACKAGE_NAME, 'journal')
name_ids = tasks.get_task_ids(PACKAGE_NAME, 'name')
zkfc_ids = tasks.get_task_ids(PACKAGE_NAME, 'zkfc')
data_ids = tasks.get_task_ids(PACKAGE_NAME, 'data')
print('journal ids: ' + str(journal_ids))
print('name ids: ' + str(name_ids))
print('zkfc ids: ' + str(zkfc_ids))
print('data ids: ' + str(data_ids))
old_config = marathon.get_config(PACKAGE_NAME)
config = marathon.get_config(PACKAGE_NAME)
print('marathon config: ')
print(config)
expiry_ms = int(config['env'][app_config_field])
print('expiry ms: ' + str(expiry_ms))
config['env'][app_config_field] = str(expiry_ms + 1)
r = cmd.request('put',
marathon.api_url('apps/' + PACKAGE_NAME),
json=config)
# Wait for journal nodes to be affected by the change
tasks.check_tasks_updated(PACKAGE_NAME, 'journal', journal_ids)
journal_ids = tasks.get_task_ids(PACKAGE_NAME, 'journal')
print('old config: ')
print(old_config)
# Put the old config back (rollback)
r = cmd.request('put',
marathon.api_url('apps/' + PACKAGE_NAME),
json=old_config)
# Wait for the journal nodes to return to their old configuration
tasks.check_tasks_updated(PACKAGE_NAME, 'journal', journal_ids)
check_healthy()
config = marathon.get_config(PACKAGE_NAME)
assert int(config['env'][app_config_field]) == expiry_ms
# ZKFC and Data tasks should not have been affected
tasks.check_tasks_not_updated(PACKAGE_NAME, 'zkfc', zkfc_ids)
tasks.check_tasks_not_updated(PACKAGE_NAME, 'data', data_ids)
def test_bump_data_nodes():
check_healthy()
data_ids = tasks.get_task_ids(PACKAGE_NAME, 'data')
print('data ids: ' + str(data_ids))
config = marathon.get_config(PACKAGE_NAME)
node_count = int(config['env']['DATA_COUNT']) + 1
config['env']['DATA_COUNT'] = str(node_count)
cmd.request('put', marathon.api_url('apps/' + PACKAGE_NAME), json=config)
check_healthy(DEFAULT_TASK_COUNT + 1)
tasks.check_tasks_not_updated(PACKAGE_NAME, 'data', data_ids)
def test_bump_journal_cpus():
check_healthy()
journal_ids = tasks.get_task_ids(PACKAGE_NAME, 'journal')
print('journal ids: ' + str(journal_ids))
config = marathon.get_config(PACKAGE_NAME)
print('marathon config: ')
print(config)
cpus = float(config['env']['JOURNAL_CPUS'])
config['env']['JOURNAL_CPUS'] = str(cpus + 0.1)
cmd.request('put', marathon.api_url('apps/' + PACKAGE_NAME), json=config)
tasks.check_tasks_updated(PACKAGE_NAME, 'journal', journal_ids)
check_healthy()
def test_deploy():
wait_time = 30
# taskcfg.yml will initially fail to deploy because several options are missing in the default
# marathon.json.mustache. verify that tasks are failing for 30s before continuing.
print('Checking that tasks are failing to launch for at least {}s'.format(
wait_time))
# we can get brief blips of TASK_RUNNING but they shouldnt last more than 2-3s:
consecutive_task_running = 0
def fn():
nonlocal consecutive_task_running
svc_tasks = shakedown.get_service_tasks(PACKAGE_NAME)
states = [t['state'] for t in svc_tasks]
print('Task states: {}'.format(states))
if 'TASK_RUNNING' in states:
consecutive_task_running += 1
assert consecutive_task_running <= 3
else:
consecutive_task_running = 0
return False
try:
spin.time_wait_noisy(lambda: fn(), timeout_seconds=wait_time)
except shakedown.TimeoutExpired:
print('Timeout reached as expected')
# add the needed envvars in marathon and confirm that the deployment succeeds:
config = marathon.get_config(PACKAGE_NAME)
env = config['env']
del env['SLEEP_DURATION']
env['TASKCFG_ALL_OUTPUT_FILENAME'] = 'output'
env['TASKCFG_ALL_SLEEP_DURATION'] = '1000'
cmd.request('put', marathon.api_url('apps/' + PACKAGE_NAME), json=config)
check_running()