def create_cluster(cmd, client, # pylint: disable=too-many-locals
resource_group, cluster_name, json_file=None, location=None, user_name=None,
ssh_key=None, password=None, image='UbuntuLTS', vm_size=None, min_nodes=0, max_nodes=None,
nfs_name=None, nfs_resource_group=None, nfs_mount_path='nfs', azure_file_share=None,
afs_mount_path='afs', container_name=None, container_mount_path='bfs', account_name=None,
account_key=None, raw=False):
if json_file:
with open(json_file) as f:
json_obj = json.load(f)
params = _get_deserializer()('ClusterCreateParameters', json_obj)
else:
params = models.ClusterCreateParameters(None, None, None)
params = _update_cluster_create_parameters_with_env_variables(cmd.cli_ctx, params, account_name, account_key)
params = _update_user_account_settings(params, user_name, ssh_key, password)
if location:
params.location = location
if not params.location:
raise CLIError('Please provide location for cluster creation.')
params = _update_nodes_information(params, image, vm_size, min_nodes, max_nodes)
if nfs_name:
file_server = client.file_servers.get(nfs_resource_group if nfs_resource_group else resource_group, nfs_name)
params = _add_nfs_to_cluster_create_parameters(params, file_server.id, nfs_mount_path)
if azure_file_share:
params = _add_azure_file_share_to_cluster_create_parameters(cmd.cli_ctx, params, azure_file_share,
afs_mount_path, account_name, account_key)
if container_name:
params = _add_azure_container_to_cluster_create_parameters(cmd.cli_ctx, params, container_name,
container_mount_path, account_name, account_key)
return client.clusters.create(resource_group, cluster_name, params, raw=raw)
def prepare_batch_ai_workspace(client, service, config):
# Create Batch AI workspace
client.workspaces.create(config.workspace_resource_group,
config.workspace,
config.location)
# Create GPU cluster
parameters = models.ClusterCreateParameters(
# VM size. Use N-series for GPU
vm_size=config.workspace_vm_size,
# Configure the ssh users
user_account_settings=models.UserAccountSettings(
admin_user_name=config.admin,
admin_user_password=config.admin_password),
# Number of VMs in the cluster
scale_settings=models.ScaleSettings(
manual=models.ManualScaleSettings(target_node_count=config.workspace_node_count)
),
# Configure each node in the cluster
node_setup=models.NodeSetup(
# Mount shared volumes to the host
mount_volumes=models.MountVolumes(
azure_file_shares=[
models.AzureFileShareReference(
account_name=config.storage_account_name,
credentials=models.AzureStorageCredentialsInfo(
account_key=config.storage_account_key),
azure_file_url='https://{0}/{1}'.format(
service.primary_endpoint, config.workspace_file_share),
relative_mount_path=config.workspace_relative_mount_path)],
),
),
)
client.clusters.create(config.workspace_resource_group, config.workspace, config.workspace_cluster, parameters).result()
def create_cluster(cmd, client, # pylint: disable=too-many-locals
resource_group, cluster_name, json_file=None, location=None, user_name=None,
ssh_key=None, password=None, generate_ssh_keys=None, image=None, custom_image=None,
use_auto_storage=False, vm_size=None, vm_priority='dedicated', target=None, min_nodes=None,
max_nodes=None, subnet=None, nfs_name=None, nfs_resource_group=None, nfs_mount_path='nfs',
azure_file_share=None, afs_mount_path='afs', container_name=None, container_mount_path='bfs',
account_name=None, account_key=None, setup_task=None, setup_task_output=None):
if generate_ssh_keys:
_generate_ssh_keys()
if ssh_key is None:
ssh_key = _get_default_ssh_public_key_location()
_ensure_resource_not_exist(client.clusters, resource_group, cluster_name)
_verify_subnet(client, subnet, nfs_name, nfs_resource_group or resource_group)
if json_file:
with open(json_file) as f:
json_obj = json.load(f)
params = _get_deserializer()('ClusterCreateParameters', json_obj)
else:
# noinspection PyTypeChecker
params = models.ClusterCreateParameters()
if params.node_setup:
params.node_setup.mount_volumes = _patch_mount_volumes(
cmd.cli_ctx, params.node_setup.mount_volumes, account_name, account_key)
params = _update_user_account_settings(params, user_name, ssh_key, password)
params.location = location or _get_resource_group_location(cmd.cli_ctx, resource_group)
params = _update_nodes_information(params, image, custom_image, vm_size, vm_priority, target, min_nodes, max_nodes)
if nfs_name or azure_file_share or container_name:
params.node_setup = params.node_setup or models.NodeSetup()
mount_volumes = params.node_setup.mount_volumes if params.node_setup else None
if nfs_name:
file_server = client.file_servers.get(nfs_resource_group or resource_group, nfs_name)
mount_volumes = _add_nfs_to_mount_volumes(mount_volumes, file_server.id, nfs_mount_path)
if azure_file_share:
mount_volumes = _add_azure_file_share_to_mount_volumes(cmd.cli_ctx, mount_volumes, azure_file_share,
afs_mount_path, account_name, account_key)
if container_name:
mount_volumes = _add_azure_container_to_mount_volumes(cmd.cli_ctx, mount_volumes, container_name,
container_mount_path, account_name, account_key)
if use_auto_storage:
auto_storage_account, auto_storage_key = _configure_auto_storage(cmd.cli_ctx, params.location)
mount_volumes = _add_azure_file_share_to_mount_volumes(
cmd.cli_ctx, mount_volumes, AUTO_STORAGE_SHARE_NAME, AUTO_STORAGE_SHARE_PATH,
auto_storage_account, auto_storage_key)
mount_volumes = _add_azure_container_to_mount_volumes(
cmd.cli_ctx, mount_volumes, AUTO_STORAGE_CONTAINER_NAME, AUTO_STORAGE_CONTAINER_PATH,
auto_storage_account, auto_storage_key)
if mount_volumes:
if params.node_setup is None:
params.node_setup = models.NodeSetup()
params.node_setup.mount_volumes = mount_volumes
if subnet:
params.subnet = models.ResourceId(id=subnet)
if setup_task:
params = _add_setup_task(setup_task, setup_task_output, params)
return client.clusters.create(resource_group, cluster_name, params)
def test_experiments_isolation(self, resource_group, location):
self.client.workspaces.create(resource_group.name, 'first',
location).result()
self.client.workspaces.create(resource_group.name, 'second',
location).result()
# Create a cluster, two experiments and a job in each experiment
for workspace in ['first', 'second']:
cluster = self.client.clusters.create(
resource_group.name,
workspace,
'cluster',
parameters=models.ClusterCreateParameters(
vm_size='STANDARD_D1',
scale_settings=models.ScaleSettings(
manual=models.ManualScaleSettings(
target_node_count=0)),
user_account_settings=models.UserAccountSettings(
admin_user_name=helpers.ADMIN_USER_NAME,
admin_user_password=helpers.ADMIN_USER_PASSWORD),
vm_priority='lowpriority')).result()
for experiment in ['exp1', 'exp2']:
self.client.experiments.create(resource_group.name, workspace,
experiment).result()
self.client.jobs.create(
resource_group.name,
workspace,
experiment,
'job',
parameters=models.JobCreateParameters(
cluster=models.ResourceId(id=cluster.id),
node_count=1,
std_out_err_path_prefix='$AZ_BATCHAI_MOUNT_ROOT',
custom_toolkit_settings=models.CustomToolkitSettings(
command_line='true'))).result()
# Delete exp1 in the first workspace
self.client.experiments.delete(resource_group.name, 'first',
'exp1').result()
# Ensure the experiment was actually deleted
self.assertRaises(
CloudError, lambda: self.client.experiments.get(
resource_group.name, 'first', 'exp1'))
for workspace in ['first', 'second']:
# Ensure the clusters are not affected
self.client.clusters.get(resource_group.name, workspace, 'cluster')
# Ensure the other experiments are not affected
for experiment in ['exp1', 'exp2']:
if workspace == 'first' and experiment == 'exp1':
continue
self.client.experiments.get(resource_group.name, workspace,
experiment)
job = self.client.jobs.get(resource_group.name, workspace,
experiment, 'job')
# And check the job are not terminated
self.assertEqual(job.execution_state,
models.ExecutionState.queued)
def cluster_parameters_for(config, container_settings, volumes):
return models.ClusterCreateParameters(
virtual_machine_configuration=models.VirtualMachineConfiguration(
image_reference=models.ImageReference(offer='UbuntuServer',
publisher='Canonical',
sku='16.04-LTS',
version='16.04.201708151')),
location=config.location,
vm_size=config.vm_type,
user_account_settings=models.UserAccountSettings(
admin_user_name=config.admin_user['name'],
admin_user_password=config.admin_user['password']),
scale_settings=models.ScaleSettings(manual=models.ManualScaleSettings(
target_node_count=config.node_count)),
node_setup=models.NodeSetup(mount_volumes=volumes))
def create_cluster(client,
location,
resource_group,
cluster_name,
vm_size,
target_nodes,
storage_account,
storage_account_key,
file_servers=None,
file_systems=None,
subnet_id=None,
setup_task_cmd=None,
setup_task_env=None,
setup_task_secrets=None):
"""Creates a cluster with given parameters and mounted Azure Files
:param BatchAIManagementClient client: client instance.
:param str location: location.
:param str resource_group: resource group name.
:param str cluster_name: name of the cluster.
:param str vm_size: vm size.
:param int target_nodes: number of nodes.
:param str storage_account: name of the storage account.
:param str storage_account_key: storage account key.
:param list(models.FileServerReference) file_servers: file servers.
:param list(models.UnmanagedFileServerReference) file_systems: file systems.
:param str setup_task_cmd: start task cmd line.
:param dict[str, str] setup_task_env: environment variables for start task.
:param dict[str, str] setup_task_secrets: environment variables with secret values for start task, server doesn't
return values for these environment variables in get cluster responses.
:param str subnet_id: virtual network subnet id.
:return models.Cluster: the created cluster
"""
Helpers._create_file_share(storage_account, storage_account_key)
setup_task = None
if setup_task_cmd:
setup_task = models.SetupTask(
command_line=setup_task_cmd,
environment_variables=[
models.EnvironmentVariable(name=k, value=v)
for k, v in setup_task_env.items()
],
secrets=[
models.EnvironmentVariableWithSecretValue(name=k, value=v)
for k, v in setup_task_secrets.items()
],
std_out_err_path_prefix='$AZ_BATCHAI_MOUNT_ROOT/{0}'.format(
Helpers.AZURE_FILES_MOUNTING_PATH))
client.workspaces.create(resource_group,
Helpers.DEFAULT_WORKSPACE_NAME,
location).result()
return client.clusters.create(
resource_group,
Helpers.DEFAULT_WORKSPACE_NAME,
cluster_name,
parameters=models.ClusterCreateParameters(
vm_size=vm_size,
scale_settings=models.ScaleSettings(
manual=models.ManualScaleSettings(
target_node_count=target_nodes)),
node_setup=models.NodeSetup(
mount_volumes=models.MountVolumes(
azure_file_shares=[
models.AzureFileShareReference(
azure_file_url=
'https://{0}.file.core.windows.net/{1}'.format(
storage_account, Helpers.AZURE_FILES_NAME),
relative_mount_path=Helpers.
AZURE_FILES_MOUNTING_PATH,
account_name=storage_account,
credentials=models.AzureStorageCredentialsInfo(
account_key=storage_account_key),
)
],
file_servers=file_servers,
unmanaged_file_systems=file_systems),
setup_task=setup_task),
subnet=subnet_id,
user_account_settings=models.UserAccountSettings(
admin_user_name=Helpers.ADMIN_USER_NAME,
admin_user_password=Helpers.ADMIN_USER_PASSWORD),
vm_priority='lowpriority')).result()
cluster_name = 'shwarscluster'
relative_mount_point = 'azurefileshare'
parameters = models.ClusterCreateParameters(
location='northeurope',
vm_size='STANDARD_NC6',
user_account_settings=models.UserAccountSettings(
admin_user_name="shwars",
admin_user_password="******"),
scale_settings=models.ScaleSettings(
manual=models.ManualScaleSettings(target_node_count=1)
),
node_setup=models.NodeSetup(
# Mount shared volumes to the host
mount_volumes=models.MountVolumes(
azure_file_shares=[
models.AzureFileShareReference(
account_name=storage_account_name,
credentials=models.AzureStorageCredentialsInfo(
account_key=storage_account_key),
azure_file_url='https://{0}.file.core.windows.net/{1}'.format(
storage_account_name, fileshare),
relative_mount_path = relative_mount_point)],
),
),
)
client.clusters.create(resource_group_name, cluster_name, parameters).result()
cluster = client.clusters.get(resource_group_name, cluster_name)