def main():
parser = argparse.ArgumentParser(description='artifactmanager')
parser.add_argument('--version', action='version', version="%(prog)s {}".format(__version__))
parser.add_argument('--port', type=int, default=ARTIFACTMANAGER_DEFAULT_PORT, help="Run server on the specified port")
args = parser.parse_args()
logging.basicConfig(format="%(asctime)s %(levelname)s %(name)s %(lineno)d %(threadName)s: %(message)s",
datefmt="%Y-%m-%dT%H:%M:%S",
stream=sys.stdout)
logging.getLogger("ax").setLevel(logging.DEBUG)
signal.signal(signal.SIGTERM, signal_handler)
signal.signal(signal.SIGINT, signal_handler)
signal.signal(signal.SIGUSR1, signal_debugger)
try:
Cloud().set_target_cloud(Cloud().own_cloud())
rest.artifact_manager = ArtifactManager()
rest.artifact_manager.init()
rest.artifact_manager.start_background_process() # start retention thread
server = pywsgi.WSGIServer(('', args.port), rest.app)
logger.info("Artifact manager %s serving on port %s", __version__, args.port)
server.serve_forever()
except SystemExit:
raise
except Exception as err:
logger.exception("Unhandled exception: %s", err)
if rest.artifact_manager:
rest.artifact_manager.stop_background_process() # stop retention thread
sys.exit(1)
def _validate_and_set_cloud_provider(self):
all_errs = []
if self.cloud_provider not in Cloud.VALID_TARGET_CLOUD_INPUT:
all_errs.append(
"Cloud provider {} not supported. Please choose from {}".
format(self.cloud_provider, Cloud.VALID_TARGET_CLOUD_INPUT))
else:
try:
# Validate placement only for AWS
c = Cloud(target_cloud=self.cloud_provider)
if c.target_cloud_aws():
ec2 = EC2(profile=self.cloud_profile,
region=self.cloud_region)
zones = ec2.get_availability_zones()
if self.cloud_placement:
if self.cloud_placement not in zones:
all_errs.append(
"Invalid cloud placement {}. Please choose from {}"
.format(self.cloud_placement, zones))
else:
self.cloud_placement = random.choice(zones)
logger.info(
"Cloud placement not provided, setting it to %s from currently available zones %s",
self.cloud_placement, zones)
except Exception as e:
all_errs.append(
"Cloud provider validation error: {}".format(e))
return all_errs
def _parse_config(self, kube_config):
"""Return config information from current kubernetes context"""
cred_info = {}
# cluster info
context = next(c['context'] for c in kube_config['contexts']
if c['name'] == kube_config['current-context'])
cred_info.update(context)
context_cluster = next(c['cluster'] for c in kube_config['clusters']
if c['name'] == context['cluster'])
cred_info['host'] = context_cluster['server'].split('/')[-1]
if 'certificate-authority-data' in context_cluster:
cred_info['certificate-authority-data'] = context_cluster[
'certificate-authority-data']
# user info
context_user = next(c['user'] for c in kube_config['users']
if c['name'] == context['user'])
if Cloud().target_cloud_aws():
if 'token' in context_user:
cred_info['token'] = "Bearer " + context_user['token']
else:
if 'username' in context_user:
cred_info['token'] = urllib3.util.make_headers(basic_auth=context_user['username'] + \
':' + context_user['password']).get('authorization')
# minikube
if 'client-certificate' in context_user:
cred_info['client-certificate'] = context_user[
'client-certificate']
if 'client-key' in context_user:
cred_info['client-key'] = context_user['client-key']
elif Cloud().target_cloud_gcp():
cred_info['token'] = GCPToken().token
return cred_info
def validate(self):
all_errs = []
all_errs += self._validate_critical_directories()
# Because we have strict validation during installation, so we can assume
# cluster has a valid name and cluster config
if not self.cluster_name:
all_errs.append("Please provide cluster name to pause the cluster")
if self.cloud_provider not in Cloud.VALID_TARGET_CLOUD_INPUT:
all_errs.append(
"Cloud provider {} not supported. Please choose from {}".
format(self.cloud_provider, Cloud.VALID_TARGET_CLOUD_INPUT))
else:
# Cloud singleton should be instantiated during validation stage so
# we can ensure customer ID
Cloud(target_cloud=self.cloud_provider)
if self.force_uninstall and self.cloud_region and self.cloud_placement:
try:
# Validate placement only for AWS
c = Cloud(target_cloud=self.cloud_provider)
if c.target_cloud_aws():
ec2 = EC2(profile=self.cloud_profile,
region=self.cloud_region)
zones = ec2.get_availability_zones()
if self.cloud_placement not in zones:
all_errs.append(
"Invalid cloud placement {}. Please choose from {}"
.format(self.cloud_placement, zones))
except Exception as e:
all_errs.append(
"Cloud provider validation error: {}".format(e))
return all_errs
def _instantiate_bucket_if_needed(self):
if not self._bucket:
logger.info("Instantiating cluster bucket ...")
self._bucket = Cloud().get_bucket(self._bucket_name,
aws_profile=self._aws_profile)
assert self._bucket.exists(), "Bucket {} not created yet".format(
self._bucket.get_bucket_name())
def _update_cluster_bucket(self):
bucket_name = AXClusterConfigPath(name_id=self._name_id).bucket()
cluster_bucket = Cloud().get_bucket(bucket_name, aws_profile=self._aws_profile, region=self._aws_region)
if not cluster_bucket.create():
raise AXPlatformException("Failed to create S3 bucket {}".format(cluster_bucket.get_bucket_name()))
logger.info("Created %s bucket ... DONE", cluster_bucket.get_bucket_name())
def create_cluster_name_id(self):
"""
User input cluster name in format of "<name>" or "<name>-<id>", and this function creates
a record in S3. If he name caller passed in does not include an ID, we generate one.
If we already have a cluster name/id record in s3, this function should not be called to avoid
existing clusters's records to get overridden
:return: <cluster-name>-<cluster-id>
"""
assert not self._cluster_name_id, "Cluster {} has it's name id already created".format(
self._cluster_name_id)
assert self._input_name, "Must provide input name to create cluster name id"
name, cid = self._format_name_id(self._input_name)
if cid is None:
logger.info("Cluster id not provided, generate one.")
if Cloud().target_cloud_gcp():
cid = str(uuid.uuid4())[:8]
elif Cloud().target_cloud_aws():
cid = str(uuid.uuid1())
else:
assert False, "Must provide valid target cloud to create cluster name id. Currently target cloud is set to {}".format(
Cloud().target_cloud())
logger.info("Created new name-id %s", name + "-" + cid)
# fill in cluster name id info
self._cluster_name = name
self._cluster_id = cid
self._cluster_name_id = self._cluster_name + "-" + self._cluster_id
return self._cluster_name_id
def _format_name_id(input_name):
if Cloud().target_cloud_aws():
return AXClusterNameIdParser.parse_cluster_name_id_aws(input_name)
elif Cloud().target_cloud_gcp():
return AXClusterNameIdParser.parse_cluster_name_id_gcp(input_name)
else:
assert False, "Invalid cloud provider: {}. Only aws and gcp are supported".format(
Cloud().target_cloud())
def __init__(self,
cluster_name_id,
kube_config=None,
key_file=None,
metadata=None,
aws_profile=None):
"""
Config file initialization
:param cluster_name_id: Cluster name_id in format of name-uuid, lcj-cluster-515d9828-7515-11e6-9b3e-a0999b1b4e15
:param kube_config: kubernetes saved config file.
:param key_file: cluster ssh key path
:param metadata: path to cluster metadata
:param aws_profile: AWS profile to access S3.
"""
assert AXEnv().is_in_pod(
) or cluster_name_id, "Must specify cluster name from outside cluster"
self._aws_profile = aws_profile
self._cluster_name_id = cluster_name_id
self._config = AXClusterConfig(cluster_name_id=cluster_name_id,
aws_profile=aws_profile)
self._kube_config = kube_config if kube_config else self.default_config_path.format(
cluster_name_id)
tmp_kube_config = kube_config if kube_config else self.default_config_path.format(
cluster_name_id)
self._kube_config = os.getenv("ARGO_KUBE_CONFIG_PATH", tmp_kube_config)
self._key_file = key_file if key_file else self.default_key_path.format(
cluster_name_id)
self._metadata_file = metadata if metadata else self.default_cluster_meta_path
config_path = AXClusterConfigPath(name_id=cluster_name_id)
self._bucket_name = config_path.bucket()
self._bucket = Cloud().get_bucket(self._bucket_name,
aws_profile=aws_profile)
self._s3_kube_config_key = config_path.kube_config()
self._s3_cluster_ssh_key = config_path.kube_ssh()
self._s3_cluster_state_before_pause = config_path.state_before_pause()
self._s3_cluster_meta = config_path.cluster_metadata()
self._s3_cluster_software_info = config_path.versions()
self._s3_platform_manifest_dir = config_path.platform_manifest_dir()
self._s3_platform_config = config_path.platform_config()
self._s3_cluster_current_state = config_path.current_state()
self._s3_portal_support_flag = config_path.portal_support()
self._s3_master_config_prefix = config_path.master_config_dir()
self._s3_master_attributes_path = config_path.master_attributes_path()
self._s3_master_user_data_path = config_path.master_user_data_path()
# For cluster staging info, stage1 and stage2 can be uploaded, downloaded, deleted with AXClusterInfo
# stage0 will can only be downloaded with AXClusterInfo. It will be uploaded during cluster information
# initialization (i.e. upload cluster id an cluster config), and deleted during cluster information
# clean up (i.e. during axinstaller uninstall)
self._staging_info = {
"stage0": config_path.cluster_install_stage0_key(),
"stage1": config_path.cluster_install_stage1_key(),
"stage2": config_path.cluster_install_stage2_key()
}
def wait_for_svc_lb_validator(status):
if bool(status["loadBalancer"] and status["loadBalancer"]["ingress"]
and len(status["loadBalancer"]["ingress"]) == 1):
if Cloud().target_cloud_aws():
return "elb.amazonaws.com" in status["loadBalancer"][
"ingress"][0]["hostname"]
elif Cloud().target_cloud_gcp():
return "ip" in status["loadBalancer"]["ingress"][0]
return False
def __init__(self, cluster_name_id=None, aws_profile=None, config=None):
self._cluster_name_id = AXClusterId(
name=cluster_name_id,
aws_profile=aws_profile).get_cluster_name_id()
self._bucket_name = AXClusterConfigPath(self._cluster_name_id).bucket()
self._bucket = Cloud().get_bucket(self._bucket_name,
aws_profile=aws_profile)
self._cluster_config_key = AXClusterConfigPath(
self._cluster_name_id).cluster_config()
self._conf = config
def __init__(self, customer_id, cluster_name_id, aws_profile):
self._customer_id = customer_id
self._cluster_name_id = cluster_name_id
self._cluster_name = AXClusterId(cluster_name_id).get_cluster_name()
self._aws_profile = aws_profile
cluster_bucket_name = AXClusterConfigPath(cluster_name_id).bucket()
self._cluster_bucket = Cloud().get_bucket(cluster_bucket_name, aws_profile=self._aws_profile)
support_bucket_name = AXSupportConfigPath(cluster_name_id).bucket()
self._support_bucket = Cloud().get_bucket(support_bucket_name, aws_profile=self._aws_profile)
def _get_bucket_region_from_aws(self):
def _do_get_region(start_region):
s3 = boto3.Session(
profile_name=self._aws_profile,
region_name=start_region).client(
"s3",
aws_access_key_id=os.environ.get("ARGO_S3_ACCESS_KEY_ID",
None),
aws_secret_access_key=os.environ.get(
"ARGO_S3_ACCESS_KEY_SECRET", None),
endpoint_url=os.environ.get("ARGO_S3_ENDPOINT", None),
config=Config(signature_version='s3v4'))
logger.debug(
"Finding region for bucket %s from with initial region %s",
self._name, start_region)
try:
response = s3.head_bucket(Bucket=self._name)
logger.debug("Head_bucket returned OK %s", response)
except ClientError as e:
response = getattr(e, "response", {})
logger.debug(
"Head_bucket returned error %s, inspecting headers",
response)
return None
headers = response.get("ResponseMetadata",
{}).get("HTTPHeaders", {})
region = headers.get("x-amz-bucket-region",
headers.get("x-amz-region", None))
logger.debug("Found region %s from head_bucket for %s, headers %s",
region, self._name, headers)
return region
if Cloud().own_cloud() == Cloud.CLOUD_AWS:
# When running on AWS instance, we query metadata server for initial region to get bucket region
return _do_get_region(Cloud().meta_data().get_region())
else:
# Assume we don't have AWS metadata server access
for r in PARTITION_DEFAULT_REGIONS:
logger.debug(
"Trying partition default region %s to get bucket region.",
r)
bucket_region = None
try:
bucket_region = _do_get_region(r)
except ClientError as e:
logger.info(
"Get region failed with: %s. Assuming region is None ...",
e)
if bucket_region:
return bucket_region
return None
def cluster(self, args):
from ax.platform.ax_cluster_info import AXClusterInfo
from ax.meta import AXClusterId
from ax.platform_client.env import AXEnv
Cloud().set_target_cloud(args.target_cloud)
assert AXEnv().is_in_pod(
) or args.cluster_name, "Must specify cluster name from outside cluster"
if args.subcommand in ['start', 'create']:
logger.error("=" * 80)
logger.error(
"axtool cluster start/create has be moved to axinstaller")
logger.error("=" * 80)
sys.exit(1)
elif args.subcommand in ['stop', 'delete']:
logger.error("=" * 80)
logger.error(
"axtool cluster stop/delete has be moved to axinstaller")
logger.error("=" * 80)
sys.exit(1)
elif args.subcommand == 'show':
import subprocess
name_id = AXClusterId(args.cluster_name,
args.aws_profile).get_cluster_name_id()
AXClusterInfo(name_id,
aws_profile=args.aws_profile).download_kube_key()
conf_file = AXClusterInfo(
name_id, aws_profile=args.aws_profile).download_kube_config()
logger.info("Kubeconfig")
with open(conf_file, "r") as f:
conf = f.read()
logger.info("%s", conf)
subprocess.call(
["kubectl", "--kubeconfig", conf_file, "cluster-info"])
subprocess.call(
["kubectl", "--kubeconfig", conf_file, "get", "no"])
subprocess.call([
"kubectl", "--kubeconfig", conf_file, "--namespace", "axsys",
"get", "po"
])
elif args.subcommand == 'download-config':
name_id = AXClusterId(args.cluster_name,
args.aws_profile).get_cluster_name_id()
if Cloud().target_cloud_aws():
AXClusterInfo(
name_id, aws_profile=args.aws_profile).download_kube_key()
AXClusterInfo(name_id,
aws_profile=args.aws_profile).download_kube_config()
def _delete_cluster_bucket(self):
logger.info("Deleting applatix-cluster bucket contents for cluster %s ...", self._name_id)
cluster_bucket = Cloud().get_bucket(AXClusterConfigPath(name_id=self._name_id).bucket(),
aws_profile=self._aws_profile, region=self._aws_region)
idobj = AXClusterId(name=self._name_id)
cluster_config_path = AXClusterConfigPath(name_id=self._name_id)
cluster_name = idobj.get_cluster_name()
prefix = cluster_name + "/"
# TODO: Not idempotent here.
# Consider the following case: if there is exception thrown when deleting S3 objects, install stage 1
# information has already been deleted but not everything are successfully deleted, the next time user
# executes "delete", this program will assume install stage 1 has been cleaned up.
exempt = [idobj.get_cluster_id_s3_key(), cluster_config_path.cluster_install_stage0_key()]
logger.info("Deleting objects for cluster %s from bucket %s. This may take some while.",
cluster_name,
cluster_bucket.get_bucket_name())
cluster_bucket.delete_all(obj_prefix=prefix, exempt=exempt)
logger.info("Deleting objects for cluster %s from bucket %s ... DONE",
cluster_name, cluster_bucket.get_bucket_name())
logger.info("Deleting stage0 information ...")
for item in exempt:
cluster_bucket.delete_object(item)
logger.info("Deleting stage0 information ... DONE")
def __init__(self, customer_image):
super(InitContainerPullImage,
self).__init__(INIT_CONTAINER_NAME_PULLIMAGE, customer_image)
nothing_hostpath = ContainerVolume("bin-nothing", "/bin/nothing")
if Cloud().in_cloud_aws():
nothing_hostpath.set_type("HOSTPATH", "/bin/nothing")
elif Cloud().in_cloud_gcp():
nothing_hostpath.set_type("HOSTPATH", "/etc/nothing")
self.add_volume(nothing_hostpath)
self.command = ["/bin/nothing"]
# AA-3175: CPU and memory are set to lowest possible so that pod requests are kept at a minimum
self.add_resource_constraints("cpu_cores", 0.001)
self.add_resource_constraints("mem_mib", 4)
def enable_docker(self, size_in_mb):
if "main" not in self.cmap:
raise AXPlatformException("Pod needs to have main container before enabling docker")
# create the dind sidecar container
dind_c = SidecarDockerDaemon(size_in_mb)
if Cloud().in_cloud_aws():
dind_c.args = ["--storage-driver=overlay2"]
elif Cloud().in_cloud_gcp():
# Current GKE defaults to overlay.
dind_c.args = ["--storage-driver=overlay"]
self.cmap["dind"] = dind_c
self.cmap["main"].add_env("DOCKER_HOST", value="tcp://localhost:2375")
return dind_c
def __init__(self, customer_image, namespace, version):
super(SidecarTask, self).__init__(SIDEKICK_WAIT_CONTAINER_NAME,
customer_image, namespace, version)
# Sidecar needs to manage logs so add the log path here
logpath = ContainerVolume("containerlogs", "/logs")
if Cloud().in_cloud_aws():
logpath.set_type("HOSTPATH", "/mnt/ephemeral/docker/containers")
elif Cloud().in_cloud_gcp():
logpath.set_type("HOSTPATH", "/var/lib/docker/containers")
self.add_volume(logpath)
self.add_env("LOGMOUNT_PATH", "/logs")
self.add_env("AX_CLUSTER_NAME_ID", os.getenv("AX_CLUSTER_NAME_ID"))
# set the arguments
self.args = ["post"]
def __init__(self, cfg):
cfg.cluster_size = AXClusterSize.CLUSTER_USER_PROVIDED
cfg.cluster_type = "standard"
cfg.vpc_id = None
cfg.vpc_cidr_base = None
cfg.subnet_mask_size = None
cfg.trusted_cidrs = ClusterInstallDefaults.TRUSTED_CIDR
cfg.user_on_demand_nodes = None
cfg.spot_instances_option = "none"
cfg.cluster_autoscaling_scan_interval = None
cfg.support_object_store_name = ""
cfg.enable_sandbox = None
cfg.software_version_info = None
self.cluster_size = cfg.cluster_size
if cfg.cloud_provider == "minikube":
self.service_manifest_root = "/ax/config/service/argo-wfe"
self.platform_bootstrap_config = "/ax/config/service/config/argo-wfe-platform-bootstrap.cfg"
Cloud(target_cloud="aws")
elif cfg.cloud_provider == "gke":
self.service_manifest_root = "/ax/config/service/argo-gke"
self.platform_bootstrap_config = "/ax/config/service/config/argo-wfe-platform-bootstrap.cfg"
Cloud(target_cloud="aws")
else:
self.service_manifest_root = "/ax/config/service/argo-all"
self.platform_bootstrap_config = "/ax/config/service/config/argo-all-platform-bootstrap.cfg"
super(PlatformOnlyInstallConfig, self).__init__(cfg)
self.install_config = ClusterInstallConfig(cfg=cfg)
self.install_config.validate()
self.cluster_bucket = cfg.cluster_bucket
self.kube_config = cfg.kubeconfig
try:
self.bucket_endpoint = cfg.endpoint
self.access_key = cfg.access_key
self.secret_key = cfg.secret_key
except Exception as ae:
self.bucket_endpoint = None
self.access_key = None
self.secret_key = None
# Overwrite the manifest_root and bootstrap_config.
self.install_config.manifest_root = self.service_manifest_root
self.install_config.bootstrap_config = self.platform_bootstrap_config
return
def __init__(self, host_ip=None):
self._host_ip = host_ip if host_ip else Cloud().meta_data(
).get_private_ip()
assert self._host_ip, "Kubelet Client is not properly initialized: Missing host ip"
logger.info("Kubelet client uses host ip %s", self._host_ip)
self._kubelet_url = "http://{}:{}".format(self._host_ip,
self.KUBELET_RO_PORT)
def _generate_raw_cluster_config_dict(self):
"""
This is a standalone method to generate cluster config dictionary based on install config. We might want to
move it to ax.platform.cluster_config package for sanity
:return:
"""
config_file_name = CLUSTER_CONFIG_TEMPLATES[self._cfg.cluster_size]
config_file_full_path = os.path.join(*[CLUSTER_CONFIG_ROOT, self._cfg.cluster_type, config_file_name])
with open(config_file_full_path, "r") as f:
config = json.load(f)
if Cloud().target_cloud_aws():
return self._generate_raw_cluster_config_dict_aws(config)
elif Cloud().target_cloud_gcp():
return self._generate_raw_cluster_config_dict_gcp(config)
else:
# Should never come here as aws/gcp is ensured at CLI validation level
return config
def _update_data_bucket(self):
data_bucket = Cloud().get_bucket(AXClusterDataPath(name_id=self._name_id).bucket(),
aws_profile=self._aws_profile, region=self._aws_region)
if not data_bucket.create():
raise AXPlatformException("Failed to create S3 bucket {}".format(data_bucket.get_bucket_name()))
# Update CORS config for data bucket too.
logger.info("Checking CORS config for %s.", data_bucket.get_bucket_name())
data_bucket.put_cors(DATA_CORS_CONFIG)
logger.info("Created %s bucket ... DONE", data_bucket.get_bucket_name())
def poll_for_elb_exists(poll_result):
if not poll_result:
return False
assert isinstance(poll_result,
list), "Poll result should be a list of objects"
for svc in poll_result:
try:
if bool(svc.status.load_balancer.ingress
and len(svc.status.load_balancer.ingress) == 1):
if Cloud().target_cloud_aws():
return "elb.amazonaws.com" in svc.status.load_balancer.ingress[
0].hostname
elif Cloud().target_cloud_gcp():
return hasattr(svc.status.load_balancer.ingress[0],
"ip")
return False
except Exception:
return False
return True
def get_region(self):
if os.environ.get("AX_AWS_REGION", None):
return os.environ.get("AX_AWS_REGION")
if Cloud().own_cloud() != Cloud.CLOUD_AWS:
return "unknown-region"
url = self._meta_url + "placement/availability-zone"
retry = AXRetry(retry_exception=(Exception, ))
r = ax_retry(requests.get, retry, url, timeout=10)
return r.text[:-1]
def _ensure_customer_id(cloud_profile):
if os.getenv("AX_CUSTOMER_ID", None):
logger.info("Using customer ID %s", os.getenv("AX_CUSTOMER_ID"))
return
# TODO (#111): set customer id to GCP
if Cloud().target_cloud_aws():
account_info = SecurityToken(aws_profile=cloud_profile).get_caller_identity()
customer_id = str(uuid.uuid5(uuid.NAMESPACE_OID, account_info["Account"]))
logger.info("Using AWS account ID hash (%s) for customer id", customer_id)
os.environ["AX_CUSTOMER_ID"] = customer_id
def _clean_up_kubernetes_cluster(self):
"""
This step cleans up Kubernetes if needed. It only touches components in cloud provider that
Kubernetes needs, including:
- Load Balancers
- Instances
- Auto scaling groups
- launch configurations
- Volumes
- Security groups
- Elastic IPs
- VPCs (If this VPC is not shared)
:return:
"""
if not check_cluster_staging(
cluster_info_obj=self._cluster_info,
stage="stage0") and not self._cfg.force_uninstall:
logger.info("Skip clean up Kubernetes cluster")
return
logger.info("Cluster uninstall step: Clean Up Kubernetes Cluster")
if self._cfg.force_uninstall:
msg = "{}\n\nIt is possible that cluster S3 bucket is accidentally deleted,\n".format(
COLOR_YELLOW)
msg += "or S3 bucket information has been altered unintentionally. In this\n"
msg += "case, we still try to delete cluster since this is force uninstall.\n"
msg += "NOTE: cluster deletion might NOT be successful and still requires\n"
msg += "user to clean up left-over resources manually.{}\n".format(
COLOR_NORM)
logger.warning(msg)
env = {
"KUBERNETES_PROVIDER": self._cfg.cloud_provider,
"KUBE_AWS_ZONE": self._cfg.cloud_placement,
"KUBE_AWS_INSTANCE_PREFIX": self._name_id
}
if self._cfg.cloud_profile:
env["AWS_DEFAULT_PROFILE"] = self._cfg.cloud_profile
else:
env["AWS_DEFAULT_PROFILE"] = AWS_DEFAULT_PROFILE
logger.info("\n\n%sCalling kube-down ...%s\n", COLOR_GREEN, COLOR_NORM)
AXKubeUpDown(cluster_name_id=self._name_id,
env=env,
aws_profile=self._cfg.cloud_profile).down()
# TODO (#111): revise volume teardown in GCP
if Cloud().target_cloud_aws():
delete_tagged_ebs(aws_profile=self._cfg.cloud_profile,
tag_key=COMMON_CLOUD_RESOURCE_TAG_KEY,
tag_value=self._name_id,
region=self._cfg.cloud_region)
def _parse_config(self, kube_config):
"""Return config information from current kubernetes context"""
cred_info = {}
# cluster info
context = next(c['context'] for c in kube_config['contexts']
if c['name'] == kube_config['current-context'])
cred_info.update(context)
context_cluster = next(c['cluster'] for c in kube_config['clusters']
if c['name'] == context['cluster'])
cred_info['host'] = context_cluster['server'].split('/')[-1]
cred_info['certificate-authority-data'] = context_cluster[
'certificate-authority-data']
# user info
context_user = next(c['user'] for c in kube_config['users']
if c['name'] == context['user'])
if Cloud().target_cloud_aws():
cred_info['token'] = context_user['token']
elif Cloud().target_cloud_gcp():
cred_info['token'] = GCPToken().token
return cred_info
def axmon_domains_list():
"""
Return a list of hosted zones (domains) that the cluster has access to
Returns:
{ 'result': [list of domains] }
"""
if Cloud().target_cloud_gcp():
return jsonify([])
else:
r53client = Route53(boto3.client("route53"))
return jsonify(result=[x.name for x in r53client.list_hosted_zones()])
def __init__(self, size_in_mb):
super(SidecarDockerDaemon, self).__init__(DIND_CONTAINER_NAME,
"argoproj/dind:1.12.6")
# Add lib modules for dind to load aufs module.
libmodule_hostpath = ContainerVolume("kernel-lib-module",
"/lib/modules")
libmodule_hostpath.set_type("HOSTPATH", "/lib/modules")
self.add_volume(libmodule_hostpath)
# Add per node dgs to sidecar
dgs_vol = ContainerVolume("docker-graph-storage", "/var/lib/docker")
if Cloud().target_cloud_aws():
dgs_vol.set_type("DOCKERGRAPHSTORAGE", size_in_mb)
elif Cloud().target_cloud_gcp():
dgs_vol.set_type("EMPTYDIR")
self.add_volume(dgs_vol)
# dind daemon needs to be privileged!
self.privileged = True
def axmon_domains_domain(domainname):
"""
Return a list of records for the domain
Returns:
{ 'result': [list of records for domain] }
"""
if Cloud().target_cloud_gcp():
return jsonify([])
else:
r53client = Route53(boto3.client("route53"))
zone = Route53HostedZone(r53client, domainname)
return jsonify(result=[x for x in zone.list_records()])
