def launch_node(node_id, creds, params, init, ssh_username, ssh_private_key):
# "pip install azure"
sms = ServiceManagementService(
subscription_id='69581868-8a08-4d98-a5b0-1d111c616fc3',
cert_file='/Users/dgriffin/certs/iOSWAToolkit.pem')
for i in sms.list_os_images():
print 'I is ', i.name, ' -- ', i.label, ' -- ', i.location, ' -- ', i.media_link
media_link = \
'http://opdemandstorage.blob.core.windows.net/communityimages/' + \
'b39f27a8b8c64d52b05eac6a62ebad85__Ubuntu_DAILY_BUILD-' + \
'precise-12_04_2-LTS-amd64-server-20130702-en-us-30GB.vhd'
config = LinuxConfigurationSet(user_name="ubuntu",
user_password="******")
hard_disk = OSVirtualHardDisk(
'b39f27a8b8c64d52b05eac6a62ebad85__Ubuntu_DAILY_BUILD-' +
'precise-12_04_2-LTS-amd64-server-20130702-en-us-30GB',
media_link,
disk_label='opdemandservice')
ret = sms.create_virtual_machine_deployment('opdemandservice', 'deploy1',
'production',
'opdemandservice2',
'opdemandservice3', config,
hard_disk)
# service_name, deployment_name, deployment_slot, label, role_name
# system_config, os_virtual_hard_disk
print 'Ret ', ret
return sms
def launch_node(node_id, creds, params, init, ssh_username, ssh_private_key):
# "pip install azure"
sms = ServiceManagementService(
subscription_id='69581868-8a08-4d98-a5b0-1d111c616fc3',
cert_file='/Users/dgriffin/certs/iOSWAToolkit.pem')
for i in sms.list_os_images():
print 'I is ', i.name, ' -- ', i.label, ' -- ', i.location, ' -- ', i.media_link
media_link = \
'http://opdemandstorage.blob.core.windows.net/communityimages/' + \
'b39f27a8b8c64d52b05eac6a62ebad85__Ubuntu_DAILY_BUILD-' + \
'precise-12_04_2-LTS-amd64-server-20130702-en-us-30GB.vhd'
config = LinuxConfigurationSet(user_name="ubuntu", user_password="******")
hard_disk = OSVirtualHardDisk(
'b39f27a8b8c64d52b05eac6a62ebad85__Ubuntu_DAILY_BUILD-' +
'precise-12_04_2-LTS-amd64-server-20130702-en-us-30GB',
media_link, disk_label='opdemandservice')
ret = sms.create_virtual_machine_deployment(
'opdemandservice', 'deploy1', 'production', 'opdemandservice2',
'opdemandservice3', config, hard_disk)
# service_name, deployment_name, deployment_slot, label, role_name
# system_config, os_virtual_hard_disk
print 'Ret ', ret
return sms
def create_machine(self, name, region='West US',
image=None, role_size='Small',
min_count=1, max_count=1,
media='storage_url_blob_cloudrunner',
username='', password='', ssh_pub_key='',
server=CR_SERVER,
cleanup=None, **kwargs):
self.log.info("Registering Azure machine [%s::%s] for [%s]" %
(name, image, CR_SERVER))
try:
sms = ServiceManagementService(self.profile.username,
self._cert_path)
server_config = LinuxConfigurationSet('myhostname', 'myuser',
'mypassword', True)
media_link = "%s__%s" % (media, name)
os_hd = OSVirtualHardDisk(image, media_link)
res = sms.create_virtual_machine_deployment(
service_name=name,
deployment_name=name,
deployment_slot='production',
label=name,
role_name=name,
system_config=server_config,
os_virtual_hard_disk=os_hd,
role_size='Small')
instance_ids = []
meta = {}
if not res:
return self.FAIL, [], {}
meta['deployment_name'] = name
meta['cleanup_service'] = cleanup in ['1', 'True', 'true']
return self.OK, instance_ids, meta
except Exception, ex:
self.log.exception(ex)
return self.FAIL, [], {}
class AzureServicesManager:
# Storage
container = 'vhds'
windows_blob_url = 'blob.core.windows.net'
# Linux
linux_user = '******'
linux_pass = '******'
location = 'West US'
# SSH Keys
def __init__(self, subscription_id, cert_file):
self.subscription_id = subscription_id
self.cert_file = cert_file
self.sms = ServiceManagementService(self.subscription_id, self.cert_file)
@property
def sms(self):
return self.sms
def list_locations(self):
locations = self.sms.list_locations()
for location in locations:
print location
def list_images(self):
return self.sms.list_os_images()
@utils.resource_not_found_handler
def get_hosted_service(self, service_name):
resp = self.sms.get_hosted_service_properties(service_name)
properties = resp.hosted_service_properties
return properties.__dict__
def delete_hosted_service(self, service_name):
res = self.sms.check_hosted_service_name_availability(service_name)
if not res.result:
return
self.sms.delete_hosted_service(service_name)
def create_hosted_service(self, os_user, service_name=None, random=False):
if not service_name:
service_name = self.generate_cloud_service_name(os_user, random)
available = False
while not available:
res = self.sms.check_hosted_service_name_availability(service_name)
if not res.result:
service_name = self.generate_cloud_service_name(os_user,
random)
else:
available = True
self.sms.create_hosted_service(service_name=service_name,
label=service_name,
location='West US')
return service_name
def create_virtual_machine(self, service_name, vm_name, image_name, role_size):
media_link = self._get_media_link(vm_name)
# Linux VM configuration
hostname = '-'.join((vm_name, 'host'))
linux_config = LinuxConfigurationSet(hostname,
self.linux_user,
self.linux_pass,
True)
# Hard disk for the OS
os_hd = OSVirtualHardDisk(image_name, media_link)
# Create vm
result = self.sms.create_virtual_machine_deployment(
service_name=service_name, deployment_name=vm_name,
deployment_slot='production', label=vm_name,
role_name=vm_name, system_config=linux_config,
os_virtual_hard_disk=os_hd,
role_size=role_size
)
request_id = result.request_id
return {
'request_id': request_id,
'media_link': media_link
}
def delete_virtual_machine(self, service_name, vm_name):
resp = self.sms.delete_deployment(service_name, vm_name, True)
self.sms.wait_for_operation_status(resp.request_id)
result = self.sms.delete_hosted_service(service_name)
return result
def generate_cloud_service_name(self, os_user=None, random=False):
if random:
return utils.generate_random_name(10)
return '-'.join((os_user, utils.generate_random_name(6)))
@utils.resource_not_found_handler
def get_virtual_machine_info(self, service_name, vm_name):
vm_info = {}
deploy_info = self.sms.get_deployment_by_name(service_name, vm_name)
if deploy_info and deploy_info.role_instance_list:
vm_info = deploy_info.role_instance_list[0].__dict__
return vm_info
def list_virtual_machines(self):
vm_list = []
services = self.sms.list_hosted_services()
for service in services:
deploys = service.deployments
if deploys and deploys.role_instance_list:
vm_name = deploys.role_instance_list[0].instance_name
vm_list.append(vm_name)
return vm_list
def power_on(self, service_name, vm_name):
resp = self.sms.start_role(service_name, vm_name, vm_name)
return resp.request_id
def power_off(self, service_name, vm_name):
resp = self.sms.shutdown_role(service_name, vm_name, vm_name)
return resp.request_id
def soft_reboot(self, service_name, vm_name):
resp = self.sms.restart_role(service_name, vm_name, vm_name)
return resp.request_id
def hard_reboot(self, service_name, vm_name):
resp = self.sms.reboot_role_instance(service_name, vm_name, vm_name)
return resp.request_id
def attach_volume(self, service_name, vm_name, size, lun):
disk_name = utils.generate_random_name(5, vm_name)
media_link = self._get_media_link(vm_name, disk_name)
self.sms.add_data_disk(service_name,
vm_name,
vm_name,
lun,
host_caching='ReadWrite',
media_link=media_link,
disk_name=disk_name,
logical_disk_size_in_gb=size)
def detach_volume(self, service_name, vm_name, lun):
self.sms.delete_data_disk(service_name, vm_name, vm_name, lun, True)
def get_available_lun(self, service_name, vm_name):
try:
role = self.sms.get_role(service_name, vm_name, vm_name)
except Exception:
return 0
disks = role.data_virtual_hard_disks
luns = [disk.lun for disk in disks].sort()
for i in range(1, 16):
if i not in luns:
return i
return None
def snapshot(self, service_name, vm_name, image_id, snanshot_name):
image_desc = 'Snapshot for image %s' % vm_name
image = CaptureRoleAsVMImage('Specialized', snanshot_name,
image_id, image_desc, 'english')
resp = self.sms.capture_vm_image(service_name, vm_name, vm_name, image)
self.sms.wait_for_operation_status(resp.request_id)
def _get_media_link(self, vm_name, filename=None, storage_account=None):
""" The MediaLink should be constructed as:
https://<storageAccount>.<blobLink>/<blobContainer>/<filename>.vhd
"""
if not storage_account:
storage_account = self._get_or_create_storage_account()
container = self.container
filename = vm_name if filename is None else filename
blob = vm_name + '-' + filename + '.vhd'
media_link = "http://%s.%s/%s/%s" % (storage_account,
self.windows_blob_url,
container, blob)
return media_link
def _get_or_create_storage_account(self):
account_list = self.sms.list_storage_accounts()
if account_list:
return account_list[-1].service_name
storage_account = utils.generate_random_name(10)
description = "Storage account %s description" % storage_account
label = storage_account + 'label'
self.sms.create_storage_account(storage_account,
description,
label,
location=self.location)
return storage_account
def _wait_for_operation(self, request_id, timeout=3000,
failure_callback=None,
failure_callback_kwargs=None):
try:
self.sms.wait_for_operation_status(request_id, timeout=timeout)
except Exception as ex:
if failure_callback and failure_callback_kwargs:
failure_callback(**failure_callback_kwargs)
raise ex
class Deployment(object):
"""
Helper class to handle deployment of the web site.
"""
def __init__(self, config):
self.config = config
self.sms = ServiceManagementService(config.getAzureSubscriptionId(),
config.getAzureCertificatePath())
self.sbms = ServiceBusManagementService(
config.getAzureSubscriptionId(), config.getAzureCertificatePath())
@staticmethod
def _resource_exists(get_resource):
"""
Helper to check for the existence of a resource in Azure.
get_resource: Parameter-less function to invoke in order to get the resource. The resource
is assumed to exist when the call to get_resource() returns a value that is not None.
If the call to get_resource() returns None or throws a WindowsAzureMissingResourceError
exception, then it is assumed that the resource does not exist.
Returns: A boolean value which is True if the resource exists.
"""
resource = None
try:
resource = get_resource()
except WindowsAzureMissingResourceError:
pass
return resource is not None
def _wait_for_operation_success(self, request_id, timeout=600, wait=5):
"""
Waits for an asynchronous Azure operation to finish.
request_id: The ID of the request to track.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to fetch the latest operation status.
"""
result = self.sms.get_operation_status(request_id)
start_time = time.time()
max_time = start_time + timeout
now = start_time
while result.status == 'InProgress':
if now >= max_time:
raise Exception(
"Operation did not finish within the expected timeout")
logger.info(
'Waiting for operation to finish (last_status=%s wait_so_far=%s)',
result.status, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
result = self.sms.get_operation_status(request_id)
now = time.time()
if result.status != 'Succeeded':
raise Exception("Operation terminated but it did not succeed.")
def _wait_for_role_instance_status(self,
role_instance_name,
service_name,
expected_status,
timeout=600,
wait=5):
"""
Waits for a role instance within the web site's cloud service to reach the status specified.
role_instance_name: Name of the role instance.
service_name: Name of service in which to find the role instance.
expected_status: Expected instance status.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to fetch the latest role status.
"""
start_time = time.time()
max_time = start_time + timeout
now = start_time
while True:
status = None
deployment = self.sms.get_deployment_by_name(
service_name, service_name)
for role_instance in deployment.role_instance_list:
if role_instance.instance_name == role_instance_name:
status = role_instance.instance_status
if status == expected_status:
break
if now >= max_time:
raise Exception(
"Operation did not finish within the expected timeout")
logger.info(
'Waiting for deployment status: expecting %s but got %s (wait_so_far=%s)',
expected_status, status, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
now = time.time()
def _wait_for_disk_deletion(self, disk_name, timeout=600, wait=5):
"""
Waits for a VM disk to disappear when it is being deleted.
disk_name: Name of the VHD.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to check for the existence of the disk.
"""
start_time = time.time()
max_time = start_time + timeout
now = start_time
logger.info("Checking that disk %s has been deleted.", disk_name)
while self._resource_exists(lambda: self.sms.get_disk(disk_name)):
if now >= max_time:
raise Exception(
"Disk %s was not deleted within the expected timeout.".
format(disk_name))
logger.info("Waiting for disk %s to disappear (wait_so_far=%s).",
disk_name, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
now = time.time()
logger.info("Disk %s has been deleted.", disk_name)
def _wait_for_namespace_active(self, name, timeout=600, wait=5):
"""
Waits for a service bus namespace to become Active.
name: Namespace name.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to check for the existence of the disk.
"""
start_time = time.time()
max_time = start_time + timeout
now = start_time
while True:
status = None
props = self.sbms.get_namespace(name)
status = props.status
if status == 'Active':
break
if now >= max_time:
raise Exception(
"Operation did not finish within the expected timeout")
logger.info(
'Waiting for namepsace status: expecting Active but got %s (wait_so_far=%s)',
status, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
now = time.time()
def _getRoleInstances(self, service_name):
"""
Returns the role instances in the given cloud service deployment. The results are provided as
a dictionary where keys are role instance names and values are RoleInstance objects.
"""
role_instances = {}
if self._resource_exists(lambda: self.sms.get_deployment_by_name(
service_name, service_name)):
deployment = self.sms.get_deployment_by_name(
service_name, service_name)
for role_instance in deployment.role_instance_list:
role_instances[role_instance.instance_name] = role_instance
return role_instances
def _ensureAffinityGroupExists(self):
"""
Creates the affinity group if it does not exist.
"""
name = self.config.getAffinityGroupName()
location = self.config.getServiceLocation()
logger.info(
"Checking for existence of affinity group (name=%s; location=%s).",
name, location)
if self._resource_exists(
lambda: self.sms.get_affinity_group_properties(name)):
logger.warn("An affinity group named %s already exists.", name)
else:
self.sms.create_affinity_group(name, name, location)
logger.info("Created affinity group %s.", name)
def _ensureStorageAccountExists(self, name):
"""
Creates the storage account if it does not exist.
"""
logger.info("Checking for existence of storage account (name=%s).",
name)
if self._resource_exists(
lambda: self.sms.get_storage_account_properties(name)):
logger.warn("A storage account named %s already exists.", name)
else:
result = self.sms.create_storage_account(
name,
"",
name,
affinity_group=self.config.getAffinityGroupName())
self._wait_for_operation_success(
result.request_id,
timeout=self.config.getAzureOperationTimeout())
logger.info("Created storage account %s.", name)
def _getStorageAccountKey(self, account_name):
"""
Gets the storage account key (primary key) for the given storage account.
"""
storage_props = self.sms.get_storage_account_keys(account_name)
return storage_props.storage_service_keys.primary
def _ensureStorageContainersExist(self):
"""
Creates Blob storage containers required by the service.
"""
logger.info("Checking for existence of Blob containers.")
account_name = self.config.getServiceStorageAccountName()
account_key = self._getStorageAccountKey(account_name)
blob_service = BlobService(account_name, account_key)
name_and_access_list = [
(self.config.getServicePublicStorageContainer(), 'blob'),
(self.config.getServiceBundleStorageContainer(), None)
]
for name, access in name_and_access_list:
logger.info("Checking for existence of Blob container %s.", name)
blob_service.create_container(name,
x_ms_blob_public_access=access,
fail_on_exist=False)
access_info = 'private' if access is None else 'public {0}'.format(
access)
logger.info("Blob container %s is ready (access: %s).", name,
access_info)
def ensureStorageHasCorsConfiguration(self):
"""
Ensures Blob storage container for bundles is configured to allow cross-origin resource sharing.
"""
logger.info("Setting CORS rules.")
account_name = self.config.getServiceStorageAccountName()
account_key = self._getStorageAccountKey(account_name)
cors_rule = CorsRule()
cors_rule.allowed_origins = self.config.getServiceStorageCorsAllowedOrigins(
)
cors_rule.allowed_methods = 'PUT'
cors_rule.exposed_headers = '*'
cors_rule.allowed_headers = '*'
cors_rule.max_age_in_seconds = 1800
cors_rules = Cors()
cors_rules.cors_rule.append(cors_rule)
set_storage_service_cors_properties(account_name, account_key,
cors_rules)
def _ensureServiceExists(self, service_name, affinity_group_name):
"""
Creates the specified cloud service host if it does not exist.
service_name: Name of the cloud service.
affinity_group_name: Name of the affinity group (which should exists).
"""
logger.info("Checking for existence of cloud service (name=%s).",
service_name)
if self._resource_exists(
lambda: self.sms.get_hosted_service_properties(service_name)):
logger.warn("A cloud service named %s already exists.",
service_name)
else:
self.sms.create_hosted_service(service_name,
service_name,
affinity_group=affinity_group_name)
logger.info("Created cloud service %s.", service_name)
def _ensureServiceCertificateExists(self, service_name):
"""
Adds certificate to the specified cloud service.
service_name: Name of the target cloud service (which should exist).
"""
cert_format = self.config.getServiceCertificateFormat()
cert_algorithm = self.config.getServiceCertificateAlgorithm()
cert_thumbprint = self.config.getServiceCertificateThumbprint()
cert_path = self.config.getServiceCertificateFilename()
cert_password = self.config.getServiceCertificatePassword()
logger.info(
"Checking for existence of cloud service certificate for service %s.",
service_name)
get_cert = lambda: self.sms.get_service_certificate(
service_name, cert_algorithm, cert_thumbprint)
if self._resource_exists(get_cert):
logger.info("Found expected cloud service certificate.")
else:
with open(cert_path, 'rb') as f:
cert_data = base64.b64encode(f.read())
if len(cert_data) <= 0:
raise Exception("Detected invalid certificate data.")
result = self.sms.add_service_certificate(service_name, cert_data,
cert_format,
cert_password)
self._wait_for_operation_success(
result.request_id,
timeout=self.config.getAzureOperationTimeout())
logger.info("Added service certificate.")
def _assertOsImageExists(self, os_image_name):
"""
Asserts that the named OS image exists.
"""
logger.info("Checking for availability of OS image (name=%s).",
os_image_name)
if self.sms.get_os_image(os_image_name) is None:
raise Exception(
"Unable to find OS Image '{0}'.".format(os_image_name))
def _ensureVirtualMachinesExist(self):
"""
Creates the VMs for the web site.
"""
service_name = self.config.getServiceName()
cert_thumbprint = self.config.getServiceCertificateThumbprint()
vm_username = self.config.getVirtualMachineLogonUsername()
vm_password = self.config.getVirtualMachineLogonPassword()
vm_role_size = self.config.getServiceInstanceRoleSize()
vm_numbers = self.config.getServiceInstanceCount()
if vm_numbers < 1:
raise Exception(
"Detected an invalid number of instances: {0}.".format(
vm_numbers))
self._assertOsImageExists(self.config.getServiceOSImageName())
role_instances = self._getRoleInstances(service_name)
for vm_number in range(1, vm_numbers + 1):
vm_hostname = '{0}-{1}'.format(service_name, vm_number)
if vm_hostname in role_instances:
logger.warn(
"Role instance %s already exists: skipping creation.",
vm_hostname)
continue
logger.info("Role instance %s provisioning begins.", vm_hostname)
vm_diskname = '{0}.vhd'.format(vm_hostname)
vm_disk_media_link = 'http://{0}.blob.core.windows.net/vhds/{1}'.format(
self.config.getServiceStorageAccountName(), vm_diskname)
ssh_port = str(self.config.getServiceInstanceSshPort() + vm_number)
os_hd = OSVirtualHardDisk(self.config.getServiceOSImageName(),
vm_disk_media_link,
disk_name=vm_diskname,
disk_label=vm_diskname)
linux_config = LinuxConfigurationSet(vm_hostname, vm_username,
vm_password, True)
linux_config.ssh.public_keys.public_keys.append(
PublicKey(
cert_thumbprint,
u'/home/{0}/.ssh/authorized_keys'.format(vm_username)))
linux_config.ssh.key_pairs.key_pairs.append(
KeyPair(cert_thumbprint,
u'/home/{0}/.ssh/id_rsa'.format(vm_username)))
network_config = ConfigurationSet()
network_config.configuration_set_type = 'NetworkConfiguration'
ssh_endpoint = ConfigurationSetInputEndpoint(name='SSH',
protocol='TCP',
port=ssh_port,
local_port=u'22')
network_config.input_endpoints.input_endpoints.append(ssh_endpoint)
http_endpoint = ConfigurationSetInputEndpoint(
name='HTTP',
protocol='TCP',
port=u'80',
local_port=u'80',
load_balanced_endpoint_set_name=service_name)
http_endpoint.load_balancer_probe.port = '80'
http_endpoint.load_balancer_probe.protocol = 'TCP'
network_config.input_endpoints.input_endpoints.append(
http_endpoint)
if vm_number == 1:
result = self.sms.create_virtual_machine_deployment(
service_name=service_name,
deployment_name=service_name,
deployment_slot='Production',
label=vm_hostname,
role_name=vm_hostname,
system_config=linux_config,
os_virtual_hard_disk=os_hd,
network_config=network_config,
availability_set_name=service_name,
data_virtual_hard_disks=None,
role_size=vm_role_size)
self._wait_for_operation_success(
result.request_id,
timeout=self.config.getAzureOperationTimeout())
self._wait_for_role_instance_status(
vm_hostname, service_name, 'ReadyRole',
self.config.getAzureOperationTimeout())
else:
result = self.sms.add_role(service_name=service_name,
deployment_name=service_name,
role_name=vm_hostname,
system_config=linux_config,
os_virtual_hard_disk=os_hd,
network_config=network_config,
availability_set_name=service_name,
role_size=vm_role_size)
self._wait_for_operation_success(
result.request_id,
timeout=self.config.getAzureOperationTimeout())
self._wait_for_role_instance_status(
vm_hostname, service_name, 'ReadyRole',
self.config.getAzureOperationTimeout())
logger.info("Role instance %s has been created.", vm_hostname)
def _deleteVirtualMachines(self, service_name):
"""
Deletes the VMs in the given cloud service.
"""
if self._resource_exists(lambda: self.sms.get_deployment_by_name(
service_name, service_name)) == False:
logger.warn("Deployment %s not found: no VMs to delete.",
service_name)
else:
logger.info("Attempting to delete deployment %s.", service_name)
# Get set of role instances before we remove them
role_instances = self._getRoleInstances(service_name)
def update_request(request):
"""
A filter to intercept the HTTP request sent by the ServiceManagementService
so we can take advantage of a newer feature ('comp=media') in the delete deployment API
(see http://msdn.microsoft.com/en-us/library/windowsazure/ee460812.aspx)
"""
hdrs = []
for name, value in request.headers:
if 'x-ms-version' == name:
value = '2013-08-01'
hdrs.append((name, value))
request.headers = hdrs
request.path = request.path + '?comp=media'
#pylint: disable=W0212
response = self.sms._filter(request)
return response
svc = ServiceManagementService(self.sms.subscription_id,
self.sms.cert_file)
#pylint: disable=W0212
svc._filter = update_request
result = svc.delete_deployment(service_name, service_name)
logger.info(
"Deployment %s deletion in progress: waiting for delete_deployment operation.",
service_name)
self._wait_for_operation_success(result.request_id)
logger.info(
"Deployment %s deletion in progress: waiting for VM disks to be removed.",
service_name)
# Now wait for the disks to disappear
for role_instance_name in role_instances.keys():
disk_name = "{0}.vhd".format(role_instance_name)
self._wait_for_disk_deletion(disk_name)
logger.info("Deployment %s deleted.", service_name)
def _ensureBuildMachineExists(self):
"""
Creates the VM for the build server.
"""
service_name = self.config.getBuildServiceName()
service_storage_name = self.config.getStorageAccountName()
cert_thumbprint = self.config.getServiceCertificateThumbprint()
vm_username = self.config.getVirtualMachineLogonUsername()
vm_password = self.config.getVirtualMachineLogonPassword()
vm_hostname = service_name
role_instances = self._getRoleInstances(service_name)
if vm_hostname in role_instances:
logger.warn("Role instance %s already exists: skipping creation.",
vm_hostname)
else:
logger.info("Role instance %s provisioning begins.", vm_hostname)
self._assertOsImageExists(self.config.getBuildOSImageName())
vm_diskname = '{0}.vhd'.format(vm_hostname)
vm_disk_media_link = 'http://{0}.blob.core.windows.net/vhds/{1}'.format(
service_storage_name, vm_diskname)
os_hd = OSVirtualHardDisk(self.config.getBuildOSImageName(),
vm_disk_media_link,
disk_name=vm_diskname,
disk_label=vm_diskname)
linux_config = LinuxConfigurationSet(vm_hostname, vm_username,
vm_password, True)
linux_config.ssh.public_keys.public_keys.append(
PublicKey(
cert_thumbprint,
u'/home/{0}/.ssh/authorized_keys'.format(vm_username)))
linux_config.ssh.key_pairs.key_pairs.append(
KeyPair(cert_thumbprint,
u'/home/{0}/.ssh/id_rsa'.format(vm_username)))
network_config = ConfigurationSet()
network_config.configuration_set_type = 'NetworkConfiguration'
ssh_endpoint = ConfigurationSetInputEndpoint(name='SSH',
protocol='TCP',
port=u'22',
local_port=u'22')
network_config.input_endpoints.input_endpoints.append(ssh_endpoint)
result = self.sms.create_virtual_machine_deployment(
service_name=service_name,
deployment_name=service_name,
deployment_slot='Production',
label=vm_hostname,
role_name=vm_hostname,
system_config=linux_config,
os_virtual_hard_disk=os_hd,
network_config=network_config,
availability_set_name=None,
data_virtual_hard_disks=None,
role_size=self.config.getBuildInstanceRoleSize())
self._wait_for_operation_success(
result.request_id,
timeout=self.config.getAzureOperationTimeout())
self._wait_for_role_instance_status(
vm_hostname, service_name, 'ReadyRole',
self.config.getAzureOperationTimeout())
logger.info("Role instance %s has been created.", vm_hostname)
def _deleteStorageAccount(self, name):
"""
Deletes the storage account for the web site.
"""
logger.info("Attempting to delete storage account %s.", name)
if self._resource_exists(
lambda: self.sms.get_storage_account_properties(name
)) == False:
logger.warn("Storage account %s not found: nothing to delete.",
name)
else:
self.sms.delete_storage_account(name)
logger.info("Storage account %s deleted.", name)
def _deleteService(self, name):
"""
Deletes the specified cloud service.
"""
logger.info("Attempting to delete cloud service %s.", name)
if self._resource_exists(
lambda: self.sms.get_hosted_service_properties(name)) == False:
logger.warn("Cloud service %s not found: nothing to delete.", name)
else:
self.sms.delete_hosted_service(name)
logger.info("Cloud service %s deleted.", name)
def _deleteAffinityGroup(self):
"""
Deletes the affinity group for the web site.
"""
name = self.config.getAffinityGroupName()
logger.info("Attempting to delete affinity group %s.", name)
if self._resource_exists(
lambda: self.sms.get_affinity_group_properties(name)) == False:
logger.warn("Affinity group %s not found: nothing to delete.",
name)
else:
self.sms.delete_affinity_group(name)
logger.info("Affinity group %s deleted.", name)
def _ensureServiceBusNamespaceExists(self):
"""
Creates the Azure Service Bus Namespace if it does not exist.
"""
name = self.config.getServiceBusNamespace()
logger.info(
"Checking for existence of service bus namespace (name=%s).", name)
if self._resource_exists(lambda: self.sbms.get_namespace(name)):
logger.warn("A namespace named %s already exists.", name)
else:
self.sbms.create_namespace(name, self.config.getServiceLocation())
self._wait_for_namespace_active(name)
logger.info("Created namespace %s.", name)
def _ensureServiceBusQueuesExist(self):
"""
Creates Azure service bus queues required by the service.
"""
logger.info("Checking for existence of Service Bus Queues.")
namespace = self.sbms.get_namespace(
self.config.getServiceBusNamespace())
sbs = ServiceBusService(namespace.name,
namespace.default_key,
issuer='owner')
queue_names = [
'jobresponsequeue', 'windowscomputequeue', 'linuxcomputequeue'
]
for name in queue_names:
logger.info("Checking for existence of Queue %s.", name)
sbs.create_queue(name, fail_on_exist=False)
logger.info("Queue %s is ready.", name)
def _deleteServiceBusNamespace(self):
"""
Deletes the Azure Service Bus Namespace.
"""
name = self.config.getServiceBusNamespace()
logger.info("Attempting to delete service bus namespace %s.", name)
if self._resource_exists(
lambda: self.sbms.get_namespace(name)) == False:
logger.warn("Namespace %s not found: nothing to delete.", name)
else:
self.sbms.delete_namespace(name)
logger.info("Namespace %s deleted.", name)
def Deploy(self, assets):
"""
Creates a deployment.
assets: The set of assets to create. The full set is: {'build', 'web'}.
"""
if len(assets) == 0:
raise ValueError("Set of assets to deploy is not specified.")
logger.info("Starting deployment operation.")
self._ensureAffinityGroupExists()
self._ensureStorageAccountExists(self.config.getStorageAccountName())
## Build instance
if 'build' in assets:
self._ensureServiceExists(self.config.getBuildServiceName(),
self.config.getAffinityGroupName())
self._ensureServiceCertificateExists(
self.config.getBuildServiceName())
self._ensureBuildMachineExists()
# Web instances
if 'web' in assets:
self._ensureStorageAccountExists(
self.config.getServiceStorageAccountName())
self._ensureStorageContainersExist()
self.ensureStorageHasCorsConfiguration()
self._ensureServiceBusNamespaceExists()
self._ensureServiceBusQueuesExist()
self._ensureServiceExists(self.config.getServiceName(),
self.config.getAffinityGroupName())
self._ensureServiceCertificateExists(self.config.getServiceName())
self._ensureVirtualMachinesExist()
#queues
logger.info("Deployment operation is complete.")
def Teardown(self, assets):
"""
Deletes a deployment.
assets: The set of assets to delete. The full set is: {'web', 'build'}.
"""
if len(assets) == 0:
raise ValueError("Set of assets to teardown is not specified.")
logger.info("Starting teardown operation.")
if 'web' in assets:
self._deleteVirtualMachines(self.config.getServiceName())
self._deleteService(self.config.getServiceName())
self._deleteStorageAccount(
self.config.getServiceStorageAccountName())
if 'build' in assets:
self._deleteVirtualMachines(self.config.getBuildServiceName())
self._deleteService(self.config.getBuildServiceName())
self._deleteStorageAccount(self.config.getStorageAccountName())
if ('web' in assets) and ('build' in assets):
self._deleteServiceBusNamespace()
self._deleteAffinityGroup()
logger.info("Teardown operation is complete.")
def getSettingsFileContent(self):
"""
Generates the content of the local Django settings file.
"""
allowed_hosts = [
'{0}.cloudapp.net'.format(self.config.getServiceName())
]
allowed_hosts.extend(self.config.getWebHostnames())
allowed_hosts.extend(['www.codalab.org', 'codalab.org'])
ssl_allowed_hosts = self.config.getSslRewriteHosts()
if len(ssl_allowed_hosts) == 0:
ssl_allowed_hosts = allowed_hosts
storage_key = self._getStorageAccountKey(
self.config.getServiceStorageAccountName())
namespace = self.sbms.get_namespace(
self.config.getServiceBusNamespace())
if len(self.config.getSslCertificateInstalledPath()) > 0:
bundle_auth_scheme = "https"
else:
bundle_auth_scheme = "http"
if len(ssl_allowed_hosts) == 0:
bundle_auth_host = '{0}.cloudapp.net'.format(
self.config.getServiceName())
else:
bundle_auth_host = ssl_allowed_hosts[0]
bundle_auth_url = "{0}://{1}".format(bundle_auth_scheme,
bundle_auth_host)
lines = [
"from base import Base",
"from default import *",
"from configurations import Settings",
"",
"import sys",
"from os.path import dirname, abspath, join",
"from pkgutil import extend_path",
"import codalab",
"",
"class {0}(Base):".format(self.config.getDjangoConfiguration()),
"",
" DEBUG=False",
"",
" ALLOWED_HOSTS = {0}".format(allowed_hosts),
"",
" SSL_PORT = '443'",
" SSL_CERTIFICATE = '{0}'".format(
self.config.getSslCertificateInstalledPath()),
" SSL_CERTIFICATE_KEY = '{0}'".format(
self.config.getSslCertificateKeyInstalledPath()),
" SSL_ALLOWED_HOSTS = {0}".format(ssl_allowed_hosts),
"",
" DEFAULT_FILE_STORAGE = 'codalab.azure_storage.AzureStorage'",
" AZURE_ACCOUNT_NAME = '{0}'".format(
self.config.getServiceStorageAccountName()),
" AZURE_ACCOUNT_KEY = '{0}'".format(storage_key),
" AZURE_CONTAINER = '{0}'".format(
self.config.getServicePublicStorageContainer()),
" BUNDLE_AZURE_ACCOUNT_NAME = AZURE_ACCOUNT_NAME",
" BUNDLE_AZURE_ACCOUNT_KEY = AZURE_ACCOUNT_KEY",
" BUNDLE_AZURE_CONTAINER = '{0}'".format(
self.config.getServiceBundleStorageContainer()),
"",
" SBS_NAMESPACE = '{0}'".format(
self.config.getServiceBusNamespace()),
" SBS_ISSUER = 'owner'",
" SBS_ACCOUNT_KEY = '{0}'".format(namespace.default_key),
" SBS_RESPONSE_QUEUE = 'jobresponsequeue'",
" SBS_COMPUTE_QUEUE = 'windowscomputequeue'",
"",
" EMAIL_BACKEND = 'django.core.mail.backends.smtp.EmailBackend'",
" EMAIL_HOST = '{0}'".format(self.config.getEmailHost()),
" EMAIL_HOST_USER = '******'".format(self.config.getEmailUser()),
" EMAIL_HOST_PASSWORD = '******'".format(
self.config.getEmailPassword()),
" EMAIL_PORT = 587",
" EMAIL_USE_TLS = True",
" DEFAULT_FROM_EMAIL = '*****@*****.**'",
" SERVER_EMAIL = '*****@*****.**'",
"",
" # Django secret",
" SECRET_KEY = '{0}'".format(self.config.getDjangoSecretKey()),
"",
" ADMINS = (('CodaLab', '*****@*****.**'),)",
" MANAGERS = ADMINS",
"",
" DATABASES = {",
" 'default': {",
" 'ENGINE': '{0}',".format(
self.config.getDatabaseEngine()),
" 'NAME': '{0}',".format(self.config.getDatabaseName()),
" 'USER': '******',".format(self.config.getDatabaseUser()),
" 'PASSWORD': '******',".format(
self.config.getDatabasePassword()),
" 'HOST': '{0}',".format(self.config.getDatabaseHost()),
" 'PORT': '{0}', ".format(
self.config.getDatabasePort()),
" 'OPTIONS' : {",
" 'init_command': 'SET SESSION TRANSACTION ISOLATION LEVEL READ COMMITTED',",
" 'read_timeout': 5",
" }",
" }",
" }",
"",
" BUNDLE_DB_NAME = '{0}'".format(
self.config.getBundleServiceDatabaseName()),
" BUNDLE_DB_USER = '******'".format(
self.config.getBundleServiceDatabaseUser()),
" BUNDLE_DB_PASSWORD = '******'".format(
self.config.getBundleServiceDatabasePassword()),
" BUNDLE_APP_ID = '{0}'".format(
self.config.getBundleServiceAppId()),
" BUNDLE_APP_KEY = '{0}'".format(
self.config.getBundleServiceAppKey()),
" BUNDLE_AUTH_URL = '{0}'".format(bundle_auth_url),
"",
" BUNDLE_SERVICE_URL = '{0}'".format(
self.config.getBundleServiceUrl()),
" BUNDLE_SERVICE_CODE_PATH = '/home/{0}/deploy/bundles'".format(
self.config.getVirtualMachineLogonUsername()),
" sys.path.append(BUNDLE_SERVICE_CODE_PATH)",
" codalab.__path__ = extend_path(codalab.__path__, codalab.__name__)",
"",
]
preview = self.config.getShowPreviewFeatures()
if preview >= 1:
if preview == 1:
lines.append(" PREVIEW_WORKSHEETS = True")
if preview > 1:
lines.append(" SHOW_BETA_FEATURES = True")
lines.append("")
return '\n'.join(lines)
enable_direct_server_return=False)
http_endpoint = ConfigurationSetInputEndpoint(
name='http',
protocol='tcp',
port='80',
local_port='80',
load_balanced_endpoint_set_name=None,
enable_direct_server_return=False)
endpoint_config.input_endpoints.input_endpoints.append(ssh_endpoint)
endpoint_config.input_endpoints.input_endpoints.append(http_endpoint)
result = sms.create_virtual_machine_deployment(
service_name=hosted_service_name,
deployment_name=hosted_service_name,
deployment_slot='production',
label=hosted_service_name,
role_name=hosted_service_name,
system_config=linux_config,
network_config=endpoint_config,
os_virtual_hard_disk=os_hd,
role_size='Small')
operation_result = sms.get_operation_status(result.request_id)
print "Created VM with name:" + hosted_service_name
host_entry = hosted_service_name + ".cloudapp.net"
host_entry += " ansible_ssh_user=rohan"
host_entry += " ansible_ssh_private_key_file="
host_entry += config_params["vm_pvt_key_path"] + "\n"
with open("inventory", "a") as inventory_file:
class Deployment(object):
"""
Helper class to handle deployment of the web site.
"""
def __init__(self, config):
self.config = config
self.sms = ServiceManagementService(config.getAzureSubscriptionId(), config.getAzureCertificatePath())
self.sbms = ServiceBusManagementService(config.getAzureSubscriptionId(), config.getAzureCertificatePath())
@staticmethod
def _resource_exists(get_resource):
"""
Helper to check for the existence of a resource in Azure.
get_resource: Parameter-less function to invoke in order to get the resource. The resource
is assumed to exist when the call to get_resource() returns a value that is not None.
If the call to get_resource() returns None or throws a WindowsAzureMissingResourceError
exception, then it is assumed that the resource does not exist.
Returns: A boolean value which is True if the resource exists.
"""
resource = None
try:
resource = get_resource()
except WindowsAzureMissingResourceError:
pass
return resource is not None
def _wait_for_operation_success(self, request_id, timeout=600, wait=5):
"""
Waits for an asynchronous Azure operation to finish.
request_id: The ID of the request to track.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to fetch the latest operation status.
"""
result = self.sms.get_operation_status(request_id)
start_time = time.time()
max_time = start_time + timeout
now = start_time
while result.status == 'InProgress':
if now >= max_time:
raise Exception("Operation did not finish within the expected timeout")
logger.info('Waiting for operation to finish (last_status=%s wait_so_far=%s)',
result.status, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
result = self.sms.get_operation_status(request_id)
now = time.time()
if result.status != 'Succeeded':
raise Exception("Operation terminated but it did not succeed.")
def _wait_for_role_instance_status(self, role_instance_name, service_name, expected_status, timeout=600, wait=5):
"""
Waits for a role instance within the web site's cloud service to reach the status specified.
role_instance_name: Name of the role instance.
service_name: Name of service in which to find the role instance.
expected_status: Expected instance status.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to fetch the latest role status.
"""
start_time = time.time()
max_time = start_time + timeout
now = start_time
while True:
status = None
deployment = self.sms.get_deployment_by_name(service_name, service_name)
for role_instance in deployment.role_instance_list:
if role_instance.instance_name == role_instance_name:
status = role_instance.instance_status
if status == expected_status:
break
if now >= max_time:
raise Exception("Operation did not finish within the expected timeout")
logger.info('Waiting for deployment status: expecting %s but got %s (wait_so_far=%s)',
expected_status, status, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
now = time.time()
def _wait_for_disk_deletion(self, disk_name, timeout=600, wait=5):
"""
Waits for a VM disk to disappear when it is being deleted.
disk_name: Name of the VHD.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to check for the existence of the disk.
"""
start_time = time.time()
max_time = start_time + timeout
now = start_time
logger.info("Checking that disk %s has been deleted.", disk_name)
while self._resource_exists(lambda: self.sms.get_disk(disk_name)):
if now >= max_time:
raise Exception("Disk %s was not deleted within the expected timeout.".format(disk_name))
logger.info("Waiting for disk %s to disappear (wait_so_far=%s).", disk_name, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
now = time.time()
logger.info("Disk %s has been deleted.", disk_name)
def _wait_for_namespace_active(self, name, timeout=600, wait=5):
"""
Waits for a service bus namespace to become Active.
name: Namespace name.
timeout: Maximum duration (in seconds) allowed for the operation to complete.
wait: Wait time (in seconds) between consecutive calls to check for the existence of the disk.
"""
start_time = time.time()
max_time = start_time + timeout
now = start_time
while True:
status = None
props = self.sbms.get_namespace(name)
status = props.status
if status == 'Active':
break
if now >= max_time:
raise Exception("Operation did not finish within the expected timeout")
logger.info('Waiting for namespace status: expecting Active but got %s (wait_so_far=%s)',
status, round(now - start_time, 1))
time_to_wait = max(0.0, min(max_time - now, wait))
time.sleep(time_to_wait)
now = time.time()
def _getRoleInstances(self, service_name):
"""
Returns the role instances in the given cloud service deployment. The results are provided as
a dictionary where keys are role instance names and values are RoleInstance objects.
"""
role_instances = {}
if self._resource_exists(lambda: self.sms.get_deployment_by_name(service_name, service_name)):
deployment = self.sms.get_deployment_by_name(service_name, service_name)
for role_instance in deployment.role_instance_list:
role_instances[role_instance.instance_name] = role_instance
return role_instances
def _ensureAffinityGroupExists(self):
"""
Creates the affinity group if it does not exist.
"""
name = self.config.getAffinityGroupName()
location = self.config.getServiceLocation()
logger.info("Checking for existence of affinity group (name=%s; location=%s).", name, location)
if self._resource_exists(lambda: self.sms.get_affinity_group_properties(name)):
logger.warn("An affinity group named %s already exists.", name)
else:
self.sms.create_affinity_group(name, name, location)
logger.info("Created affinity group %s.", name)
def _ensureStorageAccountExists(self, name):
"""
Creates the storage account if it does not exist.
"""
logger.info("Checking for existence of storage account (name=%s).", name)
if self._resource_exists(lambda: self.sms.get_storage_account_properties(name)):
logger.warn("A storage account named %s already exists.", name)
else:
result = self.sms.create_storage_account(name, "", name, affinity_group=self.config.getAffinityGroupName())
self._wait_for_operation_success(result.request_id, timeout=self.config.getAzureOperationTimeout())
logger.info("Created storage account %s.", name)
def _getStorageAccountKey(self, account_name):
"""
Gets the storage account key (primary key) for the given storage account.
"""
storage_props = self.sms.get_storage_account_keys(account_name)
return storage_props.storage_service_keys.primary
def _ensureStorageContainersExist(self):
"""
Creates Blob storage containers required by the service.
"""
logger.info("Checking for existence of Blob containers.")
account_name = self.config.getServiceStorageAccountName()
account_key = self._getStorageAccountKey(account_name)
blob_service = BlobService(account_name, account_key)
name_and_access_list = [(self.config.getServicePublicStorageContainer(), 'blob'),
(self.config.getServiceBundleStorageContainer(), None)]
for name, access in name_and_access_list:
logger.info("Checking for existence of Blob container %s.", name)
blob_service.create_container(name, x_ms_blob_public_access=access, fail_on_exist=False)
access_info = 'private' if access is None else 'public {0}'.format(access)
logger.info("Blob container %s is ready (access: %s).", name, access_info)
def ensureStorageHasCorsConfiguration(self):
"""
Ensures Blob storage container for bundles is configured to allow cross-origin resource sharing.
"""
logger.info("Setting CORS rules.")
account_name = self.config.getServiceStorageAccountName()
account_key = self._getStorageAccountKey(account_name)
cors_rule = CorsRule()
cors_rule.allowed_origins = self.config.getServiceStorageCorsAllowedOrigins()
cors_rule.allowed_methods = 'PUT'
cors_rule.exposed_headers = '*'
cors_rule.allowed_headers = '*'
cors_rule.max_age_in_seconds = 1800
cors_rules = Cors()
cors_rules.cors_rule.append(cors_rule)
set_storage_service_cors_properties(account_name, account_key, cors_rules)
def _ensureServiceExists(self, service_name, affinity_group_name):
"""
Creates the specified cloud service host if it does not exist.
service_name: Name of the cloud service.
affinity_group_name: Name of the affinity group (which should exists).
"""
logger.info("Checking for existence of cloud service (name=%s).", service_name)
if self._resource_exists(lambda: self.sms.get_hosted_service_properties(service_name)):
logger.warn("A cloud service named %s already exists.", service_name)
else:
self.sms.create_hosted_service(service_name, service_name, affinity_group=affinity_group_name)
logger.info("Created cloud service %s.", service_name)
def _ensureServiceCertificateExists(self, service_name):
"""
Adds certificate to the specified cloud service.
service_name: Name of the target cloud service (which should exist).
"""
cert_format = self.config.getServiceCertificateFormat()
cert_algorithm = self.config.getServiceCertificateAlgorithm()
cert_thumbprint = self.config.getServiceCertificateThumbprint()
cert_path = self.config.getServiceCertificateFilename()
cert_password = self.config.getServiceCertificatePassword()
logger.info("Checking for existence of cloud service certificate for service %s.", service_name)
get_cert = lambda: self.sms.get_service_certificate(service_name, cert_algorithm, cert_thumbprint)
if self._resource_exists(get_cert):
logger.info("Found expected cloud service certificate.")
else:
with open(cert_path, 'rb') as f:
cert_data = base64.b64encode(f.read())
if len(cert_data) <= 0:
raise Exception("Detected invalid certificate data.")
result = self.sms.add_service_certificate(service_name, cert_data, cert_format, cert_password)
self._wait_for_operation_success(result.request_id, timeout=self.config.getAzureOperationTimeout())
logger.info("Added service certificate.")
def _assertOsImageExists(self, os_image_name):
"""
Asserts that the named OS image exists.
"""
logger.info("Checking for availability of OS image (name=%s).", os_image_name)
if self.sms.get_os_image(os_image_name) is None:
raise Exception("Unable to find OS Image '{0}'.".format(os_image_name))
def _ensureVirtualMachinesExist(self):
"""
Creates the VMs for the web site.
"""
service_name = self.config.getServiceName()
cert_thumbprint = self.config.getServiceCertificateThumbprint()
vm_username = self.config.getVirtualMachineLogonUsername()
vm_password = self.config.getVirtualMachineLogonPassword()
vm_role_size = self.config.getServiceInstanceRoleSize()
vm_numbers = self.config.getServiceInstanceCount()
if vm_numbers < 1:
raise Exception("Detected an invalid number of instances: {0}.".format(vm_numbers))
self._assertOsImageExists(self.config.getServiceOSImageName())
role_instances = self._getRoleInstances(service_name)
for vm_number in range(1, vm_numbers+1):
vm_hostname = '{0}-{1}'.format(service_name, vm_number)
if vm_hostname in role_instances:
logger.warn("Role instance %s already exists: skipping creation.", vm_hostname)
continue
logger.info("Role instance %s provisioning begins.", vm_hostname)
vm_diskname = '{0}.vhd'.format(vm_hostname)
vm_disk_media_link = 'http://{0}.blob.core.windows.net/vhds/{1}'.format(
self.config.getServiceStorageAccountName(), vm_diskname
)
ssh_port = str(self.config.getServiceInstanceSshPort() + vm_number)
os_hd = OSVirtualHardDisk(self.config.getServiceOSImageName(),
vm_disk_media_link,
disk_name=vm_diskname,
disk_label=vm_diskname)
linux_config = LinuxConfigurationSet(vm_hostname, vm_username, vm_password, True)
linux_config.ssh.public_keys.public_keys.append(
PublicKey(cert_thumbprint, u'/home/{0}/.ssh/authorized_keys'.format(vm_username))
)
linux_config.ssh.key_pairs.key_pairs.append(
KeyPair(cert_thumbprint, u'/home/{0}/.ssh/id_rsa'.format(vm_username))
)
network_config = ConfigurationSet()
network_config.configuration_set_type = 'NetworkConfiguration'
ssh_endpoint = ConfigurationSetInputEndpoint(name='SSH',
protocol='TCP',
port=ssh_port,
local_port=u'22')
network_config.input_endpoints.input_endpoints.append(ssh_endpoint)
http_endpoint = ConfigurationSetInputEndpoint(name='HTTP',
protocol='TCP',
port=u'80',
local_port=u'80',
load_balanced_endpoint_set_name=service_name)
http_endpoint.load_balancer_probe.port = '80'
http_endpoint.load_balancer_probe.protocol = 'TCP'
network_config.input_endpoints.input_endpoints.append(http_endpoint)
if vm_number == 1:
result = self.sms.create_virtual_machine_deployment(service_name=service_name,
deployment_name=service_name,
deployment_slot='Production',
label=vm_hostname,
role_name=vm_hostname,
system_config=linux_config,
os_virtual_hard_disk=os_hd,
network_config=network_config,
availability_set_name=service_name,
data_virtual_hard_disks=None,
role_size=vm_role_size)
self._wait_for_operation_success(result.request_id,
timeout=self.config.getAzureOperationTimeout())
self._wait_for_role_instance_status(vm_hostname, service_name, 'ReadyRole',
self.config.getAzureOperationTimeout())
else:
result = self.sms.add_role(service_name=service_name,
deployment_name=service_name,
role_name=vm_hostname,
system_config=linux_config,
os_virtual_hard_disk=os_hd,
network_config=network_config,
availability_set_name=service_name,
role_size=vm_role_size)
self._wait_for_operation_success(result.request_id,
timeout=self.config.getAzureOperationTimeout())
self._wait_for_role_instance_status(vm_hostname, service_name, 'ReadyRole',
self.config.getAzureOperationTimeout())
logger.info("Role instance %s has been created.", vm_hostname)
def _deleteVirtualMachines(self, service_name):
"""
Deletes the VMs in the given cloud service.
"""
if self._resource_exists(lambda: self.sms.get_deployment_by_name(service_name, service_name)) == False:
logger.warn("Deployment %s not found: no VMs to delete.", service_name)
else:
logger.info("Attempting to delete deployment %s.", service_name)
# Get set of role instances before we remove them
role_instances = self._getRoleInstances(service_name)
def update_request(request):
"""
A filter to intercept the HTTP request sent by the ServiceManagementService
so we can take advantage of a newer feature ('comp=media') in the delete deployment API
(see http://msdn.microsoft.com/en-us/library/windowsazure/ee460812.aspx)
"""
hdrs = []
for name, value in request.headers:
if 'x-ms-version' == name:
value = '2013-08-01'
hdrs.append((name, value))
request.headers = hdrs
request.path = request.path + '?comp=media'
#pylint: disable=W0212
response = self.sms._filter(request)
return response
svc = ServiceManagementService(self.sms.subscription_id, self.sms.cert_file)
#pylint: disable=W0212
svc._filter = update_request
result = svc.delete_deployment(service_name, service_name)
logger.info("Deployment %s deletion in progress: waiting for delete_deployment operation.", service_name)
self._wait_for_operation_success(result.request_id)
logger.info("Deployment %s deletion in progress: waiting for VM disks to be removed.", service_name)
# Now wait for the disks to disappear
for role_instance_name in role_instances.keys():
disk_name = "{0}.vhd".format(role_instance_name)
self._wait_for_disk_deletion(disk_name)
logger.info("Deployment %s deleted.", service_name)
def _ensureBuildMachineExists(self):
"""
Creates the VM for the build server.
"""
service_name = self.config.getBuildServiceName()
service_storage_name = self.config.getStorageAccountName()
cert_thumbprint = self.config.getServiceCertificateThumbprint()
vm_username = self.config.getVirtualMachineLogonUsername()
vm_password = self.config.getVirtualMachineLogonPassword()
vm_hostname = service_name
role_instances = self._getRoleInstances(service_name)
if vm_hostname in role_instances:
logger.warn("Role instance %s already exists: skipping creation.", vm_hostname)
else:
logger.info("Role instance %s provisioning begins.", vm_hostname)
self._assertOsImageExists(self.config.getBuildOSImageName())
vm_diskname = '{0}.vhd'.format(vm_hostname)
vm_disk_media_link = 'http://{0}.blob.core.windows.net/vhds/{1}'.format(service_storage_name, vm_diskname)
os_hd = OSVirtualHardDisk(self.config.getBuildOSImageName(),
vm_disk_media_link,
disk_name=vm_diskname,
disk_label=vm_diskname)
linux_config = LinuxConfigurationSet(vm_hostname, vm_username, vm_password, True)
linux_config.ssh.public_keys.public_keys.append(
PublicKey(cert_thumbprint, u'/home/{0}/.ssh/authorized_keys'.format(vm_username))
)
linux_config.ssh.key_pairs.key_pairs.append(
KeyPair(cert_thumbprint, u'/home/{0}/.ssh/id_rsa'.format(vm_username))
)
network_config = ConfigurationSet()
network_config.configuration_set_type = 'NetworkConfiguration'
ssh_endpoint = ConfigurationSetInputEndpoint(name='SSH',
protocol='TCP',
port=u'22',
local_port=u'22')
network_config.input_endpoints.input_endpoints.append(ssh_endpoint)
result = self.sms.create_virtual_machine_deployment(service_name=service_name,
deployment_name=service_name,
deployment_slot='Production',
label=vm_hostname,
role_name=vm_hostname,
system_config=linux_config,
os_virtual_hard_disk=os_hd,
network_config=network_config,
availability_set_name=None,
data_virtual_hard_disks=None,
role_size=self.config.getBuildInstanceRoleSize())
self._wait_for_operation_success(result.request_id, timeout=self.config.getAzureOperationTimeout())
self._wait_for_role_instance_status(vm_hostname, service_name, 'ReadyRole',
self.config.getAzureOperationTimeout())
logger.info("Role instance %s has been created.", vm_hostname)
def _deleteStorageAccount(self, name):
"""
Deletes the storage account for the web site.
"""
logger.info("Attempting to delete storage account %s.", name)
if self._resource_exists(lambda: self.sms.get_storage_account_properties(name)) == False:
logger.warn("Storage account %s not found: nothing to delete.", name)
else:
self.sms.delete_storage_account(name)
logger.info("Storage account %s deleted.", name)
def _deleteService(self, name):
"""
Deletes the specified cloud service.
"""
logger.info("Attempting to delete cloud service %s.", name)
if self._resource_exists(lambda: self.sms.get_hosted_service_properties(name)) == False:
logger.warn("Cloud service %s not found: nothing to delete.", name)
else:
self.sms.delete_hosted_service(name)
logger.info("Cloud service %s deleted.", name)
def _deleteAffinityGroup(self):
"""
Deletes the affinity group for the web site.
"""
name = self.config.getAffinityGroupName()
logger.info("Attempting to delete affinity group %s.", name)
if self._resource_exists(lambda: self.sms.get_affinity_group_properties(name)) == False:
logger.warn("Affinity group %s not found: nothing to delete.", name)
else:
self.sms.delete_affinity_group(name)
logger.info("Affinity group %s deleted.", name)
def _ensureServiceBusNamespaceExists(self):
"""
Creates the Azure Service Bus Namespace if it does not exist.
"""
name = self.config.getServiceBusNamespace()
logger.info("Checking for existence of service bus namespace (name=%s).", name)
if self._resource_exists(lambda: self.sbms.get_namespace(name)):
logger.warn("A namespace named %s already exists.", name)
else:
self.sbms.create_namespace(name, self.config.getServiceLocation())
self._wait_for_namespace_active(name)
logger.info("Created namespace %s.", name)
def _ensureServiceBusQueuesExist(self):
"""
Creates Azure service bus queues required by the service.
"""
logger.info("Checking for existence of Service Bus Queues.")
namespace = self.sbms.get_namespace(self.config.getServiceBusNamespace())
sbs = ServiceBusService(namespace.name, namespace.default_key, issuer='owner')
queue_names = ['jobresponsequeue', 'windowscomputequeue', 'linuxcomputequeue']
for name in queue_names:
logger.info("Checking for existence of Queue %s.", name)
sbs.create_queue(name, fail_on_exist=False)
logger.info("Queue %s is ready.", name)
def _deleteServiceBusNamespace(self):
"""
Deletes the Azure Service Bus Namespace.
"""
name = self.config.getServiceBusNamespace()
logger.info("Attempting to delete service bus namespace %s.", name)
if self._resource_exists(lambda: self.sbms.get_namespace(name)) == False:
logger.warn("Namespace %s not found: nothing to delete.", name)
else:
self.sbms.delete_namespace(name)
logger.info("Namespace %s deleted.", name)
def Deploy(self, assets):
"""
Creates a deployment.
assets: The set of assets to create. The full set is: {'build', 'web'}.
"""
if len(assets) == 0:
raise ValueError("Set of assets to deploy is not specified.")
logger.info("Starting deployment operation.")
self._ensureAffinityGroupExists()
self._ensureStorageAccountExists(self.config.getStorageAccountName())
## Build instance
if 'build' in assets:
self._ensureServiceExists(self.config.getBuildServiceName(), self.config.getAffinityGroupName())
self._ensureServiceCertificateExists(self.config.getBuildServiceName())
self._ensureBuildMachineExists()
# Web instances
if 'web' in assets:
self._ensureStorageAccountExists(self.config.getServiceStorageAccountName())
self._ensureStorageContainersExist()
self.ensureStorageHasCorsConfiguration()
self._ensureServiceBusNamespaceExists()
self._ensureServiceBusQueuesExist()
self._ensureServiceExists(self.config.getServiceName(), self.config.getAffinityGroupName())
self._ensureServiceCertificateExists(self.config.getServiceName())
self._ensureVirtualMachinesExist()
#queues
logger.info("Deployment operation is complete.")
def Teardown(self, assets):
"""
Deletes a deployment.
assets: The set of assets to delete. The full set is: {'web', 'build'}.
"""
if len(assets) == 0:
raise ValueError("Set of assets to teardown is not specified.")
logger.info("Starting teardown operation.")
if 'web' in assets:
self._deleteVirtualMachines(self.config.getServiceName())
self._deleteService(self.config.getServiceName())
self._deleteStorageAccount(self.config.getServiceStorageAccountName())
if 'build' in assets:
self._deleteVirtualMachines(self.config.getBuildServiceName())
self._deleteService(self.config.getBuildServiceName())
self._deleteStorageAccount(self.config.getStorageAccountName())
if ('web' in assets) and ('build' in assets):
self._deleteServiceBusNamespace()
self._deleteAffinityGroup()
logger.info("Teardown operation is complete.")
def getSettingsFileContent(self):
"""
Generates the content of the local Django settings file.
"""
allowed_hosts = ['{0}.cloudapp.net'.format(self.config.getServiceName())]
allowed_hosts.extend(self.config.getWebHostnames())
allowed_hosts.extend(['www.codalab.org', 'codalab.org'])
ssl_allowed_hosts = self.config.getSslRewriteHosts();
if len(ssl_allowed_hosts) == 0:
ssl_allowed_hosts = allowed_hosts
storage_key = self._getStorageAccountKey(self.config.getServiceStorageAccountName())
namespace = self.sbms.get_namespace(self.config.getServiceBusNamespace())
if len(self.config.getSslCertificateInstalledPath()) > 0:
bundle_auth_scheme = "https"
else:
bundle_auth_scheme = "http"
if len(ssl_allowed_hosts) == 0:
bundle_auth_host = '{0}.cloudapp.net'.format(self.config.getServiceName())
else:
bundle_auth_host = ssl_allowed_hosts[0]
bundle_auth_url = "{0}://{1}".format(bundle_auth_scheme, bundle_auth_host)
lines = [
"from base import Base",
"from default import *",
"from configurations import Settings",
"",
"import sys",
"from os.path import dirname, abspath, join",
"from pkgutil import extend_path",
"import codalab",
"",
"class {0}(Base):".format(self.config.getDjangoConfiguration()),
"",
" DEBUG=False",
"",
" ALLOWED_HOSTS = {0}".format(allowed_hosts),
"",
" SSL_PORT = '443'",
" SSL_CERTIFICATE = '{0}'".format(self.config.getSslCertificateInstalledPath()),
" SSL_CERTIFICATE_KEY = '{0}'".format(self.config.getSslCertificateKeyInstalledPath()),
" SSL_ALLOWED_HOSTS = {0}".format(ssl_allowed_hosts),
"",
" DEFAULT_FILE_STORAGE = 'codalab.azure_storage.AzureStorage'",
" AZURE_ACCOUNT_NAME = '{0}'".format(self.config.getServiceStorageAccountName()),
" AZURE_ACCOUNT_KEY = '{0}'".format(storage_key),
" AZURE_CONTAINER = '{0}'".format(self.config.getServicePublicStorageContainer()),
" BUNDLE_AZURE_ACCOUNT_NAME = AZURE_ACCOUNT_NAME",
" BUNDLE_AZURE_ACCOUNT_KEY = AZURE_ACCOUNT_KEY",
" BUNDLE_AZURE_CONTAINER = '{0}'".format(self.config.getServiceBundleStorageContainer()),
"",
" SBS_NAMESPACE = '{0}'".format(self.config.getServiceBusNamespace()),
" SBS_ISSUER = 'owner'",
" SBS_ACCOUNT_KEY = '{0}'".format(namespace.default_key),
" SBS_RESPONSE_QUEUE = 'jobresponsequeue'",
" SBS_COMPUTE_QUEUE = 'windowscomputequeue'",
"",
" EMAIL_BACKEND = 'django.core.mail.backends.smtp.EmailBackend'",
" EMAIL_HOST = '{0}'".format(self.config.getEmailHost()),
" EMAIL_HOST_USER = '******'".format(self.config.getEmailUser()),
" EMAIL_HOST_PASSWORD = '******'".format(self.config.getEmailPassword()),
" EMAIL_PORT = 587",
" EMAIL_USE_TLS = True",
" DEFAULT_FROM_EMAIL = 'CodaLab <*****@*****.**>'",
" SERVER_EMAIL = '*****@*****.**'",
"",
" # Django secret",
" SECRET_KEY = '{0}'".format(self.config.getDjangoSecretKey()),
"",
" ADMINS = (('CodaLab', '*****@*****.**'),)",
" MANAGERS = ADMINS",
"",
" DATABASES = {",
" 'default': {",
" 'ENGINE': '{0}',".format(self.config.getDatabaseEngine()),
" 'NAME': '{0}',".format(self.config.getDatabaseName()),
" 'USER': '******',".format(self.config.getDatabaseUser()),
" 'PASSWORD': '******',".format(self.config.getDatabasePassword()),
" 'HOST': '{0}',".format(self.config.getDatabaseHost()),
" 'PORT': '{0}', ".format(self.config.getDatabasePort()),
" 'OPTIONS' : {",
" 'init_command': 'SET SESSION TRANSACTION ISOLATION LEVEL READ COMMITTED',",
" 'read_timeout': 5",
" }",
" }",
" }",
"",
" BUNDLE_DB_NAME = '{0}'".format(self.config.getBundleServiceDatabaseName()),
" BUNDLE_DB_USER = '******'".format(self.config.getBundleServiceDatabaseUser()),
" BUNDLE_DB_PASSWORD = '******'".format(self.config.getBundleServiceDatabasePassword()),
" BUNDLE_APP_ID = '{0}'".format(self.config.getBundleServiceAppId()),
" BUNDLE_APP_KEY = '{0}'".format(self.config.getBundleServiceAppKey()),
" BUNDLE_AUTH_URL = '{0}'".format(bundle_auth_url),
"",
" BUNDLE_SERVICE_URL = '{0}'".format(self.config.getBundleServiceUrl()),
" BUNDLE_SERVICE_CODE_PATH = '/home/{0}/deploy/bundles'".format(self.config.getVirtualMachineLogonUsername()),
" sys.path.append(BUNDLE_SERVICE_CODE_PATH)",
" codalab.__path__ = extend_path(codalab.__path__, codalab.__name__)",
"",
]
preview = self.config.getShowPreviewFeatures()
if preview >= 1:
if preview == 1:
lines.append(" PREVIEW_WORKSHEETS = True")
if preview > 1:
lines.append(" SHOW_BETA_FEATURES = True")
lines.append("")
return '\n'.join(lines)
class AzureServicesManager:
# Storage
container = 'vhds'
windows_blob_url = 'blob.core.windows.net'
# Linux
linux_user = '******'
linux_pass = '******'
location = 'West US'
# SSH Keys
def __init__(self, subscription_id, cert_file):
self.subscription_id = subscription_id
self.cert_file = cert_file
self.sms = ServiceManagementService(self.subscription_id,
self.cert_file)
@property
def sms(self):
return self.sms
def list_locations(self):
locations = self.sms.list_locations()
for location in locations:
print location
def list_images(self):
return self.sms.list_os_images()
@utils.resource_not_found_handler
def get_hosted_service(self, service_name):
resp = self.sms.get_hosted_service_properties(service_name)
properties = resp.hosted_service_properties
return properties.__dict__
def delete_hosted_service(self, service_name):
res = self.sms.check_hosted_service_name_availability(service_name)
if not res.result:
return
self.sms.delete_hosted_service(service_name)
def create_hosted_service(self, os_user, service_name=None, random=False):
if not service_name:
service_name = self.generate_cloud_service_name(os_user, random)
available = False
while not available:
res = self.sms.check_hosted_service_name_availability(service_name)
if not res.result:
service_name = self.generate_cloud_service_name(
os_user, random)
else:
available = True
self.sms.create_hosted_service(service_name=service_name,
label=service_name,
location='West US')
return service_name
def create_virtual_machine(self, service_name, vm_name, image_name,
role_size):
media_link = self._get_media_link(vm_name)
# Linux VM configuration
hostname = '-'.join((vm_name, 'host'))
linux_config = LinuxConfigurationSet(hostname, self.linux_user,
self.linux_pass, True)
# Hard disk for the OS
os_hd = OSVirtualHardDisk(image_name, media_link)
# Create vm
result = self.sms.create_virtual_machine_deployment(
service_name=service_name,
deployment_name=vm_name,
deployment_slot='production',
label=vm_name,
role_name=vm_name,
system_config=linux_config,
os_virtual_hard_disk=os_hd,
role_size=role_size)
request_id = result.request_id
return {'request_id': request_id, 'media_link': media_link}
def delete_virtual_machine(self, service_name, vm_name):
resp = self.sms.delete_deployment(service_name, vm_name, True)
self.sms.wait_for_operation_status(resp.request_id)
result = self.sms.delete_hosted_service(service_name)
return result
def generate_cloud_service_name(self, os_user=None, random=False):
if random:
return utils.generate_random_name(10)
return '-'.join((os_user, utils.generate_random_name(6)))
@utils.resource_not_found_handler
def get_virtual_machine_info(self, service_name, vm_name):
vm_info = {}
deploy_info = self.sms.get_deployment_by_name(service_name, vm_name)
if deploy_info and deploy_info.role_instance_list:
vm_info = deploy_info.role_instance_list[0].__dict__
return vm_info
def list_virtual_machines(self):
vm_list = []
services = self.sms.list_hosted_services()
for service in services:
deploys = service.deployments
if deploys and deploys.role_instance_list:
vm_name = deploys.role_instance_list[0].instance_name
vm_list.append(vm_name)
return vm_list
def power_on(self, service_name, vm_name):
resp = self.sms.start_role(service_name, vm_name, vm_name)
return resp.request_id
def power_off(self, service_name, vm_name):
resp = self.sms.shutdown_role(service_name, vm_name, vm_name)
return resp.request_id
def soft_reboot(self, service_name, vm_name):
resp = self.sms.restart_role(service_name, vm_name, vm_name)
return resp.request_id
def hard_reboot(self, service_name, vm_name):
resp = self.sms.reboot_role_instance(service_name, vm_name, vm_name)
return resp.request_id
def attach_volume(self, service_name, vm_name, size, lun):
disk_name = utils.generate_random_name(5, vm_name)
media_link = self._get_media_link(vm_name, disk_name)
self.sms.add_data_disk(service_name,
vm_name,
vm_name,
lun,
host_caching='ReadWrite',
media_link=media_link,
disk_name=disk_name,
logical_disk_size_in_gb=size)
def detach_volume(self, service_name, vm_name, lun):
self.sms.delete_data_disk(service_name, vm_name, vm_name, lun, True)
def get_available_lun(self, service_name, vm_name):
try:
role = self.sms.get_role(service_name, vm_name, vm_name)
except Exception:
return 0
disks = role.data_virtual_hard_disks
luns = [disk.lun for disk in disks].sort()
for i in range(1, 16):
if i not in luns:
return i
return None
def snapshot(self, service_name, vm_name, image_id, snanshot_name):
image_desc = 'Snapshot for image %s' % vm_name
image = CaptureRoleAsVMImage('Specialized', snanshot_name, image_id,
image_desc, 'english')
resp = self.sms.capture_vm_image(service_name, vm_name, vm_name, image)
self.sms.wait_for_operation_status(resp.request_id)
def _get_media_link(self, vm_name, filename=None, storage_account=None):
""" The MediaLink should be constructed as:
https://<storageAccount>.<blobLink>/<blobContainer>/<filename>.vhd
"""
if not storage_account:
storage_account = self._get_or_create_storage_account()
container = self.container
filename = vm_name if filename is None else filename
blob = vm_name + '-' + filename + '.vhd'
media_link = "http://%s.%s/%s/%s" % (
storage_account, self.windows_blob_url, container, blob)
return media_link
def _get_or_create_storage_account(self):
account_list = self.sms.list_storage_accounts()
if account_list:
return account_list[-1].service_name
storage_account = utils.generate_random_name(10)
description = "Storage account %s description" % storage_account
label = storage_account + 'label'
self.sms.create_storage_account(storage_account,
description,
label,
location=self.location)
return storage_account
def _wait_for_operation(self,
request_id,
timeout=3000,
failure_callback=None,
failure_callback_kwargs=None):
try:
self.sms.wait_for_operation_status(request_id, timeout=timeout)
except Exception as ex:
if failure_callback and failure_callback_kwargs:
failure_callback(**failure_callback_kwargs)
raise ex
ssh_endpoint = ConfigurationSetInputEndpoint(name='ssh', protocol='tcp',
port='22', local_port='22',
load_balanced_endpoint_set_name=None,
enable_direct_server_return=False)
http_endpoint = ConfigurationSetInputEndpoint(name='http', protocol='tcp',
port='80', local_port='80',
load_balanced_endpoint_set_name=None,
enable_direct_server_return=False)
endpoint_config.input_endpoints.input_endpoints.append(ssh_endpoint)
endpoint_config.input_endpoints.input_endpoints.append(http_endpoint)
result = sms.create_virtual_machine_deployment(service_name=hosted_service_name,
deployment_name=hosted_service_name,
deployment_slot='production',
label=hosted_service_name,
role_name=hosted_service_name,
system_config=linux_config,
network_config=endpoint_config,
os_virtual_hard_disk=os_hd,
role_size='Small')
operation_result = sms.get_operation_status(result.request_id)
print "Created VM with name:" + hosted_service_name
host_entry = hosted_service_name + ".cloudapp.net"
host_entry += " ansible_ssh_user=rohan"
host_entry += " ansible_ssh_private_key_file="
host_entry += config_params["vm_pvt_key_path"] + "\n"
with open("inventory", "a") as inventory_file:
