def ssh_into_node(base_client,
pool_id,
node_id,
username,
ssh_key=None,
password=None,
port_forward_list=None,
internal=False):
if internal:
result = base_client.batch_client.compute_node.get(pool_id=pool_id,
node_id=node_id)
rls = models.RemoteLogin(ip_address=result.ip_address, port="22")
else:
result = base_client.batch_client.compute_node.get_remote_login_settings(
pool_id, node_id)
rls = models.RemoteLogin(ip_address=result.remote_login_ip_address,
port=str(result.remote_login_port))
ssh_lib.node_ssh(
username=username,
hostname=rls.ip_address,
port=rls.port,
ssh_key=ssh_key,
password=password,
port_forward_list=port_forward_list,
)
def node_run(base_client,
cluster_id,
node_id,
command,
internal,
container_name=None,
timeout=None):
cluster = base_client.get(cluster_id)
pool, nodes = cluster.pool, list(cluster.nodes)
try:
node = next(node for node in nodes if node.id == node_id)
except StopIteration:
raise error.AztkError("Node with id {} not found".format(node_id))
if internal:
node_rls = models.RemoteLogin(ip_address=node.ip_address, port="22")
else:
node_rls = base_client.get_remote_login_settings(pool.id, node.id)
try:
generated_username, ssh_key = base_client.generate_user_on_node(
pool.id, node.id)
output = ssh_lib.node_exec_command(
node.id,
command,
generated_username,
node_rls.ip_address,
node_rls.port,
ssh_key=ssh_key.exportKey().decode('utf-8'),
container_name=container_name,
timeout=timeout)
return output
finally:
base_client.delete_user_on_node(cluster_id, node.id,
generated_username)
def __cluster_copy(self,
cluster_id,
source_path,
destination_path,
container_name=None,
internal=False,
get=False,
timeout=None):
pool, nodes = self.__get_pool_details(cluster_id)
nodes = [node for node in nodes]
if internal:
cluster_nodes = [(node,
models.RemoteLogin(ip_address=node.ip_address,
port="22")) for node in nodes]
else:
cluster_nodes = [
(node, self.__get_remote_login_settings(pool.id, node.id))
for node in nodes
]
try:
ssh_key = self.__create_user_on_pool('aztk', pool.id, nodes)
output = asyncio.get_event_loop().run_until_complete(
ssh_lib.clus_copy(container_name=container_name,
username='******',
nodes=cluster_nodes,
source_path=source_path,
destination_path=destination_path,
ssh_key=ssh_key.exportKey().decode('utf-8'),
get=get,
timeout=timeout))
return output
except (OSError, batch_error.BatchErrorException) as exc:
raise exc
finally:
self.__delete_user_on_pool('aztk', pool.id, nodes)
def __cluster_run(self,
cluster_id,
command,
internal,
container_name=None,
timeout=None):
pool, nodes = self.__get_pool_details(cluster_id)
nodes = [node for node in nodes]
if internal:
cluster_nodes = [(node,
models.RemoteLogin(ip_address=node.ip_address,
port="22")) for node in nodes]
else:
cluster_nodes = [
(node, self.__get_remote_login_settings(pool.id, node.id))
for node in nodes
]
try:
ssh_key = self.__create_user_on_pool('aztk', pool.id, nodes)
output = asyncio.get_event_loop().run_until_complete(
ssh_lib.clus_exec_command(
command,
'aztk',
cluster_nodes,
ssh_key=ssh_key.exportKey().decode('utf-8'),
container_name=container_name,
timeout=timeout))
return output
except OSError as exc:
raise exc
finally:
self.__delete_user_on_pool('aztk', pool.id, nodes)
def __node_run(self,
cluster_id,
node_id,
command,
internal,
container_name=None,
timeout=None):
pool, nodes = self.__get_pool_details(cluster_id)
try:
node = next(node for node in nodes if node.id == node_id)
except StopIteration:
raise error.AztkError("Node with id {} not found".format(node_id))
if internal:
node_rls = models.RemoteLogin(ip_address=node.ip_address,
port="22")
else:
node_rls = self.__get_remote_login_settings(pool.id, node.id)
try:
generated_username, ssh_key = self.__generate_user_on_node(
pool.id, node.id)
output = ssh_lib.node_exec_command(
node.id,
command,
generated_username,
node_rls.ip_address,
node_rls.port,
ssh_key=ssh_key.exportKey().decode("utf-8"),
container_name=container_name,
timeout=timeout,
)
return output
finally:
self.__delete_user(cluster_id, node.id, generated_username)
def _get_remote_login_settings(base_client, pool_id: str, node_id: str):
"""
Get the remote_login_settings for node
:param pool_id
:param node_id
:returns aztk.models.RemoteLogin
"""
result = base_client.batch_client.compute_node.get_remote_login_settings(pool_id, node_id)
return models.RemoteLogin(ip_address=result.remote_login_ip_address, port=str(result.remote_login_port))
def cluster_copy(
cluster_operations,
cluster_id,
source_path,
destination_path=None,
container_name=None,
internal=False,
get=False,
timeout=None,
):
cluster = cluster_operations.get(cluster_id)
pool, nodes = cluster.pool, list(cluster.nodes)
if internal:
cluster_nodes = [(node,
models.RemoteLogin(ip_address=node.ip_address,
port="22")) for node in nodes]
else:
cluster_nodes = [
(node,
cluster_operations.get_remote_login_settings(pool.id, node.id))
for node in nodes
]
try:
generated_username, ssh_key = cluster_operations.generate_user_on_cluster(
pool.id, nodes)
except BatchErrorException as e:
raise error.AztkError(helpers.format_batch_exception(e))
try:
output = asyncio.get_event_loop().run_until_complete(
ssh_lib.clus_copy(
container_name=container_name,
username=generated_username,
nodes=cluster_nodes,
source_path=source_path,
destination_path=destination_path,
ssh_key=ssh_key.exportKey().decode("utf-8"),
get=get,
timeout=timeout,
))
return output
except (OSError, BatchErrorException) as exc:
raise exc
finally:
cluster_operations.delete_user_on_cluster(pool.id, nodes,
generated_username)
def cluster_run(base_operations,
cluster_id,
command,
internal,
container_name=None,
timeout=None):
cluster = base_operations.get(cluster_id)
pool, nodes = cluster.pool, list(cluster.nodes)
if internal:
cluster_nodes = [(node,
models.RemoteLogin(ip_address=node.ip_address,
port="22")) for node in nodes]
else:
cluster_nodes = [
(node, base_operations.get_remote_login_settings(pool.id, node.id))
for node in nodes
]
try:
generated_username, ssh_key = base_operations.generate_user_on_cluster(
pool.id, nodes)
except batch_error.BatchErrorException as e:
raise error.AztkError(helpers.format_batch_exception(e))
try:
output = asyncio.get_event_loop().run_until_complete(
ssh_lib.clus_exec_command(
command,
generated_username,
cluster_nodes,
ssh_key=ssh_key.exportKey().decode("utf-8"),
container_name=container_name,
timeout=timeout,
))
return output
except OSError as exc:
raise exc
finally:
base_operations.delete_user_on_cluster(pool.id, nodes,
generated_username)
