def setup_cli(self, stub_universes={}):
try:
self._github_updater.update('pending', 'Setting up CLI')
self._configure_cli_sandbox() # adds to os.environ
cli_filepath = self._download_cli_to_sandbox()
self._configure_cli(self._dcos_url)
dcos_login.DCOSLogin(self._dcos_url).login()
# check for any preexisting universes and remove them -- the cluster requires no duplicate uris
if stub_universes:
logger.info('Checking for duplicate stub universes')
cur_universes = subprocess.check_output(
'dcos package repo list --json'.split()).decode('utf-8')
for repo in json.loads(cur_universes)['repositories']:
# {u'name': u'Universe', u'uri': u'https://universe.mesosphere.com/repo'}
if repo['uri'] in stub_universes.values():
logger.info(
'Removing duplicate repository: {} {}'.format(
repo['name'], repo['uri']))
subprocess.check_call(
'dcos package repo remove {}'.format(
repo['name']).split())
for name, url in stub_universes.items():
logger.info('Adding repository: {} {}'.format(name, url))
subprocess.check_call(
'dcos package repo add --index=0 {} {}'.format(
name, url).split())
except:
self._github_updater.update('error', 'CLI Setup failed')
raise
def _initialize_dcos_cli(self):
logger.info("Initializing dcos config")
subprocess.check_call(['which', 'dcos'])
subprocess.check_call(['dcos', 'config', 'set', 'core.dcos_url', self.dcos_url])
subprocess.check_call(['dcos', 'config', 'set', 'core.reporting', 'True'])
subprocess.check_call(['dcos', 'config', 'set', 'core.ssl_verify', 'False'])
subprocess.check_call(['dcos', 'config', 'set', 'core.timeout', '5'])
subprocess.check_call(['dcos', 'config', 'show'])
dcos_login.DCOSLogin(self.dcos_url).login()
def _start(self, config):
is_17_cluster = config.ccm_channel in self._DCOS_17_CHANNELS
if is_17_cluster:
hostrepo = 's3.amazonaws.com/downloads.mesosphere.io/dcos'
elif config.cf_template.startswith('ee.'):
hostrepo = 's3.amazonaws.com/downloads.mesosphere.io/dcos-enterprise'
else:
hostrepo = 's3-us-west-2.amazonaws.com/downloads.dcos.io/dcos'
template_url = 'https://{}/{}/cloudformation/{}'.format(
hostrepo, config.ccm_channel, config.cf_template)
if config.template_url:
template_url = config.template_url
cluster_name = config.name_prefix + self._rand_str(8)
payload = {
'template_url': template_url,
'name': cluster_name,
'cluster_desc': config.description,
'time': config.duration_mins,
'private_agents': str(config.private_agents),
'public_agents': str(config.public_agents),
'pre_1_8_cluster': is_17_cluster,
'adminlocation': config.admin_location,
'cloud_provider': config.cloud_provider,
'region': config.aws_region
}
logger.info('''Launching cluster:
name={}
agents={} private/{} public
duration={} minutes
mountvols={}
permissions={}
channel={}
template={}'''.format(cluster_name, config.private_agents,
config.public_agents, config.duration_mins,
config.mount_volumes, config.permissions,
config.ccm_channel, config.cf_template))
response = self._query_http('POST',
self._CCM_PATH,
request_json_payload=payload)
if not response:
raise Exception('CCM cluster creation request failed')
response_content = response.read().decode('utf-8')
response_json = json.loads(response_content)
logger.info('Launch response:\n{}'.format(
pprint.pformat(response_json)))
cluster_id = int(response_json.get('id', 0))
if not cluster_id:
raise Exception(
'No Cluster ID returned in cluster creation response: {}'.
format(response_content))
stack_id = response_json.get('stack_id', '')
if not stack_id:
raise Exception(
'No Stack ID returned in cluster creation response: {}'.format(
response_content))
cluster_info = self.wait_for_status(cluster_id, 'CREATING', 'RUNNING',
config.start_timeout_mins)
if not cluster_info:
raise Exception('CCM cluster creation failed or timed out')
dns_address = cluster_info.get('DnsAddress', '')
if not dns_address:
raise Exception(
'CCM cluster_info is missing DnsAddress: {}'.format(
cluster_info))
logger.info('Cluster is now RUNNING: {}'.format(cluster_info))
if config.mount_volumes:
logger.info(
'Enabling mount volumes for cluster {} (stack id {})'.format(
cluster_id, stack_id))
# fabric spams to stdout, which causes problems with launch_ccm_cluster.
# force total redirect to stderr:
stdout = sys.stdout
sys.stdout = sys.stderr
import enable_mount_volumes
enable_mount_volumes.main(stack_id)
sys.stdout = stdout
# we fetch the token once up-front because on Open clusters it must be reused.
# given that, we may as well use the same flow across both Open and EE.
logger.info('Fetching auth token')
dcos_url = 'https://' + dns_address
auth_token = dcos_login.DCOSLogin(dcos_url).get_acs_token()
if config.permissions:
logger.info(
'Setting up permissions for cluster {} (stack id {})'.format(
cluster_id, stack_id))
def run_script(scriptname, args=[]):
logger.info('Command: {} {}'.format(scriptname,
' '.join(args)))
script_path = os.path.join(
os.path.dirname(os.path.realpath(__file__)), scriptname)
# redirect stdout to stderr:
subprocess.check_call(['bash', script_path] + args,
stdout=sys.stderr)
run_script('create_service_account.sh',
[dcos_url, auth_token, '--strict'])
# Examples of what individual tests should run. See respective projects' "test.sh":
#run_script('setup_permissions.sh', 'nobody cassandra-role'.split())
#run_script('setup_permissions.sh', 'nobody hdfs-role'.split())
#run_script('setup_permissions.sh', 'nobody kafka-role'.split())
#run_script('setup_permissions.sh', 'nobody spark-role'.split())
return {'id': cluster_id, 'url': dcos_url, 'auth_token': auth_token}
def _start(self, config):
is_17_cluster = config.ccm_channel in self._DCOS_17_CHANNELS
if is_17_cluster:
hostrepo = 's3.amazonaws.com/downloads.mesosphere.io/dcos'
elif config.cf_template.startswith('ee.'):
hostrepo = 's3.amazonaws.com/downloads.mesosphere.io/dcos-enterprise'
else:
hostrepo = 's3-us-west-2.amazonaws.com/downloads.dcos.io/dcos'
template_url = 'https://{}/{}/cloudformation/{}'.format(
hostrepo, config.ccm_channel, config.cf_template)
if config.template_url:
template_url = config.template_url
cluster_name = config.name_prefix + self._rand_str(8)
payload = {
'template_url': template_url,
'name': cluster_name,
'cluster_desc': config.description,
'time': config.duration_mins,
'private_agents': str(config.private_agents),
'public_agents': str(config.public_agents),
'pre_1_8_cluster': is_17_cluster,
'adminlocation': config.admin_location,
'cloud_provider': config.cloud_provider,
'region': config.aws_region
}
logger.info('''Launching cluster:
name={}
agents={} private/{} public
duration={} minutes
mountvols={}
channel={}
template={}'''.format(cluster_name, config.private_agents,
config.public_agents, config.duration_mins,
config.mount_volumes, config.ccm_channel,
config.cf_template))
response = self._query_http('POST',
self._CCM_PATH,
request_json_payload=payload)
if not response:
raise Exception('CCM cluster creation request failed')
response_content = response.read().decode('utf-8')
response_json = json.loads(response_content)
logger.info('Launch response:\n{}'.format(
pprint.pformat(response_json)))
cluster_id = int(response_json.get('id', 0))
if not cluster_id:
raise Exception(
'No Cluster ID returned in cluster creation response: {}'.
format(response_content))
stack_id = response_json.get('stack_id', '')
if not stack_id:
raise Exception(
'No Stack ID returned in cluster creation response: {}'.format(
response_content))
cluster_info = self.wait_for_status(cluster_id, 'CREATING', 'RUNNING',
config.start_timeout_mins)
if not cluster_info:
raise Exception('CCM cluster creation failed or timed out')
dns_address = cluster_info.get('DnsAddress', '')
if not dns_address:
raise Exception(
'CCM cluster_info is missing DnsAddress: {}'.format(
cluster_info))
if config.mount_volumes:
logger.info(
'Enabling mount volumes for cluster {} (stack id {})'.format(
cluster_id, stack_id))
# fabric spams to stdout, which causes problems with launch_ccm_cluster.
# force total redirect to stderr:
stdout = sys.stdout
sys.stdout = sys.stderr
import enable_mount_volumes
enable_mount_volumes.main(stack_id)
sys.stdout = stdout
# we fetch the token once up-front because on Open clusters it must be reused.
# given that, we may as well use the same flow across both Open and EE.
logger.info('Fetching auth token')
dcos_url = 'https://' + dns_address
auth_token = dcos_login.DCOSLogin(dcos_url).get_acs_token()
return {'id': cluster_id, 'url': dcos_url, 'auth_token': auth_token}
def _start(self, config):
is_17_cluster = config.ccm_channel in self._DCOS_17_CHANNELS
scale_url = None
if is_17_cluster:
hostrepo = 's3.amazonaws.com/downloads.mesosphere.io/dcos'
elif config.cf_scale.startswith('ee.'):
hostrepo = 's3.amazonaws.com/downloads.mesosphere.io/dcos-enterprise-aws-advanced'
# format is different for enterprise security modes.
mode = config.security_mode
if not mode:
logger.warning("No scales known for enterprise & "
"default security (none). Cowardly bringing "
"up a permissive cluster")
mode = 'permissive'
scale_url = 'https://{}/{}/{}/cloudformation/{}'.format(
hostrepo, config.ccm_channel, mode, config.cf_scale)
else:
hostrepo = 's3-us-west-2.amazonaws.com/downloads.dcos.io/dcos'
# non-ee mode
if not scale_url:
scale_url = 'https://{}/{}/cloudformation/{}'.format(
hostrepo, config.ccm_channel, config.cf_scale)
# external override from DCOS_SCALE_URL
if config.scale_url:
scale_url = config.scale_url
logger.info("Accepting externally provided scale_url from environment.")
cluster_name = config.name_prefix + self._rand_str(8)
payload = {
'scale_url': scale_url,
'name': cluster_name,
'cluster_desc': config.description,
'time': config.duration_mins,
'private_agents': str(config.private_agents),
'public_agents': str(config.public_agents),
'pre_1_8_cluster': is_17_cluster,
'adminlocation': config.admin_location,
'cloud_provider': config.cloud_provider,
'region': config.aws_region
}
logger.info('''Launching cluster:
name={}
agents={} private/{} public
duration={} minutes
mountvols={}
permissions={}
channel={}
scale={}
scale_url={}'''.format(
cluster_name,
config.private_agents, config.public_agents,
config.duration_mins,
config.mount_volumes,
config.security_mode,
config.ccm_channel,
config.cf_scale,
scale_url))
response = self._query_http('POST', self._CCM_PATH, request_json_payload=payload)
if not response:
raise ClusterActionException('CCM cluster creation request failed')
response_content = response.read().decode('utf-8')
response_json = json.loads(response_content)
logger.info('Launch response:\n{}'.format(pprint.pformat(response_json)))
cluster_id = int(response_json.get('id', 0))
if not cluster_id:
raise ClusterActionException('No Cluster ID returned in cluster creation response: {}'.format(response_content))
stack_id = response_json.get('stack_id', '')
if not stack_id:
raise ClusterActionException('No Stack ID returned in cluster creation response: {}'.format(response_content))
cluster_info = self.wait_for_status(
cluster_id,
['CREATING', 'RUNNING_NEEDS_INFO'], # pending states
'RUNNING', # desired state
config.start_timeout_mins)
if not cluster_info:
raise ClusterActionException('CCM cluster creation failed or timed out')
dns_address = cluster_info.get('DnsAddress', '')
if not dns_address:
raise ClusterActionException('CCM cluster_info is missing DnsAddress: {}'.format(cluster_info))
logger.info('Cluster is now RUNNING: {}'.format(cluster_info))
# we fetch the token once up-front because on Open clusters it must be reused.
# given that, we may as well use the same flow across both Open and EE.
logger.info('Fetching auth token')
dcos_url = 'https://' + dns_address
auth_token = dcos_login.DCOSLogin(dcos_url).get_acs_token()
if config.postlaunch_steps != 'none':
is_enterprise = config.cf_scale.startswith('ee.')
clustinit = configure_test_cluster.ClusterInitializer(
stack_id, auth_token, dns_address, is_enterprise,
config.security_mode)
initmaster = True
if config.postlaunch_steps == 'nomaster':
initmaster = False
clustinit.apply_default_config(initmaster=initmaster)
if config.mount_volumes:
clustinit.create_mount_volumes()
return {
'id': cluster_id,
'url': dcos_url,
'auth_token': auth_token
}
