def vm_create(self,
vm_name,
vm_type="CernVM",
vm_user="******",
vm_networkassoc="",
vm_image="",
vm_mem=1,
vm_cores=1,
vm_storage=30,
vm_keepalive=0,
myproxy_creds_name=None,
myproxy_server=None,
myproxy_server_port=None,
job_per_core=False,
vm_loc=''):
self.log.debug(
"Running new instance with Marketplace id %s in StratusLab",
str(vm_loc))
runner = None
if vm_loc not in StratusLabCluster.__idMap:
#vm_loc: URL of VM or key? Does not seem to matter in Runner (l.506)
runner = Runner(vm_loc, StratusLabCluster._v_configHolder)
StratusLabCluster.__idMap[vm_loc] = runner
else:
runner = StratusLabCluster.__idMap[vm_loc]
try:
ids = runner.runInstance()
self.log.debug("Created instances: %s", str(ids))
#for new_id in ids:
new_id = ids[len(ids) - 1]
#if job didnt't set a keep_alive use the clouds default
if not vm_keepalive and self.keep_alive:
vm_keepalive = self.keep_alive
new_vm = cluster_tools.VM(name=vm_name,
id=str(new_id),
vmtype=vm_type,
user=vm_user,
network=vm_networkassoc,
image=vm_image,
memory=vm_mem,
cpucores=vm_cores,
storage=vm_storage,
keep_alive=vm_keepalive,
myproxy_creds_name=myproxy_creds_name,
myproxy_server=myproxy_server,
myproxy_server_port=myproxy_server_port,
job_per_core=job_per_core)
StratusLabCluster.__vmMap[str(new_id)] = vm_loc
if vm_loc not in self.__runnerIds:
self.__runnerIds[vm_loc] = [
str(new_id),
]
else:
self.__runnerIds[vm_loc].append(str(new_id))
self.vms.append(new_vm)
try:
self.resource_checkout(new_vm)
except:
self.log.exception(
"Unexpected error checking out resources when creating a VM."
" Programming error?")
return self.ERROR
#endfor
return 0
except Exception, e:
self.log.exception(
"Exception running new instance in StratusLab: %s", str(e))
return -1
and e.errors[0][0] == "ImageNotFound":
self.failed_image_set.add(vm_ami)
except:
return self.ERROR
return self.ERROR
#if job didn't set a keep_alive use the clouds default
if not vm_keepalive and self.keep_alive:
vm_keepalive = self.keep_alive
new_vm = cluster_tools.VM(name=vm_name,
id=instance_id,
vmtype=vm_type,
user=vm_user,
clusteraddr=self.network_address,
cloudtype=self.cloud_type,
network=vm_networkassoc,
image=vm_ami,
flavor=instance_type,
memory=vm_mem,
cpucores=vm_cores,
storage=vm_storage,
keep_alive=vm_keepalive,
job_per_core=job_per_core)
try:
new_vm.spot_id = spot_id
except:
log.verbose("No spot ID to add to VM %s" % instance_id)
try:
self.resource_checkout(new_vm)
except:
class GoogleComputeEngineCluster(cluster_tools.ICluster):
"""
Connector class for Google Compute Engine.
"""
VM_STATES = {
"RUNNING" : "Running",
"Starting" : "Starting",
"shutting-down" : "Shutdown",
"terminated" : "Shutdown",
"PROVISIONING" : "Provisioning",
"error" : "Error",
}
GCE_SCOPE = 'https://www.googleapis.com/auth/compute'
API_VERSION = 'v1'
GCE_URL = 'https://www.googleapis.com/compute/%s/projects/' % (API_VERSION)
DEFAULT_ZONE = 'us-central1-b' # will need to be option in job
DEFAULT_MACHINE_TYPE = 'n1-standard-1' # option specified in job config
DEFAULT_INSTANCE_TYPE_LIST = _attr_list_to_dict(config.config_options.get('job', 'default_VMInstanceTypeList'))
DEFAULT_IMAGE = 'cloudscheduler-centos-9'
DEFAULT_ROOT_PD_NAME = 'hepgc-uvic-root-pd'
DEFAULT_NETWORK = 'default' # job option setup
DEFAULT_SERVICE_EMAIL = 'default'
DEFAULT_SCOPES = ['https://www.googleapis.com/auth/devstorage.full_control',
'https://www.googleapis.com/auth/compute']
def __init__(self, name="Dummy Cluster", host="localhost",
cloud_type="Dummy", memory=[], max_vm_mem=-1, networks=[],
vm_slots=0, cpu_cores=0, storage=0, boot_timeout=None,
auth_dat_file=None, secret_file=None, security_group=None,
project_id=None, enabled=True, priority=0,
total_cpu_cores=-1, keep_alive=0,):
"""
Google Compute constructor
:param name:
:param host:
:param cloud_type:
:param memory:
:param max_vm_mem:
:param networks:
:param vm_slots:
:param cpu_cores:
:param storage:
:param boot_timeout:
:param auth_dat_file:
:param secret_file:
:param security_group:
:param project_id:
:param enabled:
:param priority:
:param total_cpu_cores:
:param keep_alive:
"""
log.debug("Init GCE cores %s, storage %s", cpu_cores, storage)
self.gce_hostname_prefix = 'gce-cs-vm'
self.security_group = security_group
self.auth_dat_file_path = auth_dat_file
self.secret_file_path = secret_file
self.project_id = project_id
self.total_cpu_cores = total_cpu_cores
self.keep_alive = keep_alive
if not project_id:
return None
# Perform OAuth 2.0 authorization.
flow = flow_from_clientsecrets(self.secret_file_path, scope=self.GCE_SCOPE)
auth_storage = Storage(self.auth_dat_file_path)
credentials = auth_storage.get()
if credentials is None or credentials.invalid:
credentials = run_flow(flow, auth_storage)
http = httplib2.Http()
self.auth_http = credentials.authorize(http)
#if not security_group:
# security_group = ["default"]
#self.security_groups = security_group
# Build service object
self.gce_service = build('compute', self.API_VERSION)
self.project_url = '%s%s' % (self.GCE_URL, self.project_id)
# Call super class's init
cluster_tools.ICluster.__init__(self, name=name, host=host, cloud_type=cloud_type,
memory=memory, max_vm_mem=max_vm_mem, networks=networks,
vm_slots=vm_slots, cpu_cores=cpu_cores,
storage=storage, boot_timeout=boot_timeout, enabled=enabled,
priority=priority, keep_alive=0,)
def vm_create(self, vm_type, vm_user, vm_networkassoc,
vm_image, vm_mem, vm_cores, vm_storage, customization=None,
vm_keepalive=0, instance_type="",
job_per_core=False, pre_customization=None,
use_cloud_init=False, extra_userdata=[]):
"""
Create a VM on GCE.
:param vm_type:
:param vm_user:
:param vm_networkassoc:
:param vm_image:
:param vm_mem:
:param vm_cores:
:param vm_storage:
:param customization:
:param vm_keepalive:
:param instance_type:
:param job_per_core:
:param pre_customization:
:param use_cloud_init:
:param extra_userdata:
:return:
"""
try:
if self.network_address in vm_image.keys():
vm_ami = vm_image[self.network_address]
elif self.name in vm_image.keys():
vm_ami = vm_image[self.name]
except:
log.debug("No AMI for %s, trying default", self.network_address)
try:
vm_ami = vm_image["default"]
except:
log.exception("Can't find a suitable AMI")
return
# Construct URLs
try:
if self.name in instance_type.keys():
i_type = instance_type[self.name]
else:
i_type = instance_type[self.network_address]
except:
log.debug("No instance type for %s, trying default", self.network_address)
try:
if self.name in self.DEFAULT_INSTANCE_TYPE_LIST.keys():
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[self.name]
else:
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[self.network_address]
except:
log.debug("No default instance type found for %s, trying single default",
self.network_address)
i_type = self.DEFAULT_MACHINE_TYPE
vm_instance_type = i_type
if vm_image:
vm_image_name = vm_ami
else:
vm_image_name = self.DEFAULT_IMAGE
#Ensures the VM's Root Disks are Unique
self.DEFAULT_ROOT_PD_NAME = '%s-%s'%('hepgc-uvic-root-pd', self.generate_next_instance_name())
#temporary variable for disk_url
disk_url = '%s%s/zones/%s/disks/%s'%(self.GCE_URL, self.project_id,
self.DEFAULT_ZONE, self.DEFAULT_ROOT_PD_NAME)
machine_type_url = '%s/zones/%s/machineTypes/%s' % (
self.project_url, self.DEFAULT_ZONE, vm_instance_type)
#zone_url = '%s/zones/%s' % (self.project_url, self.DEFAULT_ZONE)
network_url = '%s/global/networks/%s' % (self.project_url, self.DEFAULT_NETWORK)
# Construct the request body
disk = {
'name': self.DEFAULT_ROOT_PD_NAME,
'sourceSnapshot':
'https://www.googleapis.com/compute/v1/projects/atlasgce/global/snapshots/%s'%vm_image_name,
'sizeGb':vm_storage
}
# Create the root pd
try:
request = self.gce_service.disks().insert(project=self.project_id,
body=disk, zone=self.DEFAULT_ZONE)
response = request.execute(http=self.auth_http)
response = self._blocking_call(self.gce_service, self.auth_http, response)
except:
log.exception('Error Trying to create disk, one already exists ... returning ')
return
use_cloud_init = use_cloud_init or config.config_options.getboolean('global', 'use_cloud_init')
if customization:
user_data = cloud_init_util.build_write_files_cloud_init(customization)
else:
user_data = ""
if pre_customization:
user_data = cloud_init_util.inject_customizations(pre_customization, user_data)
elif use_cloud_init:
user_data = cloud_init_util.inject_customizations([], user_data)[0]
if len(extra_userdata) > 0:
# need to use the multi-mime type functions
user_data = cloud_init_util.build_multi_mime_message([(user_data,
'cloud-config')],
extra_userdata)
if not user_data:
log.error("Problem building cloud-config user data.")
return 1
# Compress the user data to try and get under the limit
user_data = utilities.gzip_userdata(user_data)
next_instance_name = self.generate_next_instance_name()
instance = {
'name': next_instance_name,
'machineType': machine_type_url,
#'disks': [{
# 'autoDelete': 'true',
# 'boot': 'true',
# 'type': 'PERSISTENT',
# 'initializeParams' : {
# 'diskname': self.DEFAULT_ROOT_PD_NAME,
# 'sourceImage': image_url
# }
# }],
'disks': [{
'source':disk_url,
'boot': 'true',
'autoDelete':'true',
'type': 'PERSISTENT',
}],
#'image': image_url,
'networkInterfaces': [{
'accessConfigs': [{
'type': 'ONE_TO_ONE_NAT',
'name': 'External NAT'
}],
'network': network_url
}],
'serviceAccounts': [{
'email': self.DEFAULT_SERVICE_EMAIL,
'scopes': self.DEFAULT_SCOPES
}],
'metadata': {
'items': [
{
'key': 'user-data',
'value': user_data,
},
]
}
}
# Create the instance
response = None
request = self.gce_service.instances().insert(
project=self.project_id, body=instance, zone=self.DEFAULT_ZONE)
try:
response = request.execute(http=self.auth_http)
response = self._blocking_call(self.gce_service, self.auth_http, response)
#log.info('Created VM ')
except Exception, e:
log.error("Error creating VM on gce: %s", e)
if response and 'targetId' in response:
target_id = response['targetId']
elif response:
return
else:
#print 'no response'
return
#if job didn't set a keep alive use the clouds default
if not vm_keepalive and self.keep_alive:
vm_keepalive = self.keep_alive
new_vm = cluster_tools.VM(name=next_instance_name, vmtype=vm_type, user=vm_user,
clusteraddr=self.network_address, id=target_id,
cloudtype=self.cloud_type, network=vm_networkassoc,
hostname=self.construct_hostname(next_instance_name),
image=vm_image, flavor=vm_instance_type,
memory=vm_mem, cpucores=vm_cores, storage=vm_storage,
keep_alive=vm_keepalive, job_per_core=job_per_core)
try:
self.resource_checkout(new_vm)
except:
log.exception("Unexpected Error checking out resources when creating a VM. Programming error?")
self.vm_destroy(new_vm, reason="Failed Resource checkout", return_resources=False)
return self.ERROR
self.vms.append(new_vm)
#log.info("added a new vm %s"%new_vm)
return 0
def vm_create(self,
vm_name,
vm_image,
vm_mem,
vm_cores,
vm_type,
vm_user,
vm_keepalive=0,
network='default',
customizations=None,
pre_customization=None,
extra_userdata="",
key_name=""):
import libvirt
###Create a VM on LocalHost.###
conn = libvirt.open(None)
if conn is None:
log.debug("Failed to open connection to hypervisior")
sys.exit(1)
#get VM name and create a working directory
name = self._generate_next_name()
config_tmp = tempfile.mkdtemp(suffix="-" + name)
log.debug(config_val.get('global', 'condor_host_on_vm'))
log.debug(config_val.get('global', 'condor_context_file'))
if customizations:
user_data = cloud_init_util.build_write_files_cloud_init(
customizations)
else:
user_data = ""
if pre_customization:
user_data = cloud_init_util.inject_customizations(
pre_customization, user_data)
if os.path.exists('/etc/cloudscheduler/auth-key.yaml'):
extra_userdata = ['/etc/cloudscheduler/auth-key.yaml'
] + extra_userdata
if len(extra_userdata) > 0:
# need to use the multi-mime type functions
user_data = cloud_init_util.build_multi_mime_message(
[(user_data, 'cloud-config', 'cloud_conf.yaml')],
extra_userdata)
if not user_data:
log.error("Problem building cloud-config user data.")
return self.ERROR
raw_user = user_data
user_data = utilities.gzip_userdata(user_data)
try:
if self.name in vm_image.keys():
image = vm_image[self.name]
elif self.network_address in vm_image.keys():
image = vm_image[self.network_address]
else:
image = vm_image['default']
except Exception as error:
log.error("Could not determine image: %s", error)
###INSIDE A CONTAINER: The default image repo is /jobs/repo, will check there that the image exists
#to-do: add optional specification for image repo location
#check image exists in repo
if os.path.exists('/jobs/instances/base/' + image):
path = '/jobs/instances/base/' + image
elif os.path.exists(image):
path = image
image = os.path.basename(path)
else:
log.error(
'Could not find image %s: Does not exists in image repository',
image)
return 1
#check image format
if image.endswith('.img'):
image_copy = image.rstrip('.img')
image_copy = image_copy + '-' + name + '.qcow2'
subprocess.call('qemu-img create -f qcow2 -b ' + path +
' /jobs/instances/' + image_copy,
shell=True)
image = image_copy
path = '/jobs/instances/' + image
elif image.endswith('.qcow2'):
image_copy = image.rstrip('.qcow2')
base = image_copy + '.img'
if not os.path.exists('/jobs/instances/base/' + base):
subprocess.call('qemu-img convert -f qcow2 -O raw ' + path +
' /jobs/instances/base/' + base,
shell=True)
image_copy = image_copy + '-' + name + '.qcow2'
subprocess.call(
'qemu-img create -f qcow2 -b /jobs/instances/base/' + base +
' /jobs/instances/' + image_copy,
shell=True)
image = image_copy
path = '/jobs/instances/' + image
#Create the config directory for metadata server
metapath = self._generate_meta(name)
subprocess.call('mv ' + metapath + ' ' + config_tmp + '/meta-data',
shell=True)
subprocess.call('rm -f ' + metapath, shell=True)
with open(config_tmp + "/user-data", 'w') as ufile:
ufile.write(user_data)
with open(config_tmp + '/raw-user', 'w') as raw:
raw.write(raw_user)
try:
subprocess.call("mkisofs -o " + config_tmp +
"/config.iso -V cidata -r -J --quiet " +
config_tmp + "/meta-data " + config_tmp +
"/user-data",
shell=True)
except Exception as error:
log.debug("Could not create config dir")
virt_call = "virt-install --name "+name+" --network "+network+ \
" --print-xml --dry-run -r "+str(vm_mem)+" --disk path="+path+ \
",sparse=true --disk path="+config_tmp+ \
"/config.iso,device=cdrom --import --serial file,path="+config_tmp+"/boot-log"
image_xml = subprocess.check_output(virt_call, shell=True)
dom = conn.createXML(image_xml, 0)
if dom is None:
log.debug("Failed to create domain from xml definiton")
sys.exit(1)
else:
if not vm_keepalive and self.keep_alive:
vm_keepalive = self.keep_alive
new_vm = cluster_tools.VM(name=name,
id=dom.ID(),
vmtype=vm_type,
hostname=self.name,
user=vm_user,
cloudtype=self.cloud_type,
network=network,
image=image,
memory=vm_mem,
cpucores=vm_cores,
keep_alive=vm_keepalive)
try:
self.resource_checkout(new_vm)
log.info("Launching 1 VM: %s on %s ", dom.ID(), self.name)
self.vms.append(new_vm)
except Exception as error:
log.error(
"Unexpected Error checking out resources when creating a VM. Programming error?: %s",
error)
self.vm_destroy(new_vm,
reason="Failed Resource checkout",
return_resources=False)
return self.ERROR
conn.close()
return 0
def vm_create(self,
vm_name,
vm_type,
vm_user,
vm_networkassoc,
vm_image,
vm_mem,
vm_cores,
vm_storage,
customization=None,
vm_keepalive=0,
instance_type="",
job_per_core=False,
securitygroup=None,
key_name="",
pre_customization=None,
use_cloud_init=False,
extra_userdata=None):
""" Create a VM on OpenStack."""
import novaclient.exceptions
use_cloud_init = use_cloud_init or config_val.getboolean(
'global', 'use_cloud_init')
nova = self._get_creds_nova_updated()
if securitygroup:
sec_group = []
for group in securitygroup:
if group in self.security_groups:
sec_group.append(group)
if len(sec_group) == 0:
log.debug("No defined security groups for job - "
"trying default value from cloud_resources.conf")
sec_group = self.security_groups
else:
sec_group = self.security_groups
log.debug("Using security group: %s", str(sec_group))
if key_name and len(key_name) > 0:
if not nova.keypairs.findall(name=key_name):
key_name = ""
else:
key_name = self.key_name if self.key_name else ""
if customization:
user_data = cloud_init_util.build_write_files_cloud_init(
customization)
else:
user_data = ""
if pre_customization:
if not use_cloud_init:
for item in pre_customization:
user_data = '\n'.join([item, user_data])
else:
user_data = cloud_init_util.inject_customizations(
pre_customization, user_data)
elif use_cloud_init:
user_data = cloud_init_util.inject_customizations([], user_data)
if extra_userdata:
# need to use the multi-mime type functions
user_data = cloud_init_util.build_multi_mime_message(
[(user_data, 'cloud-config', 'cloud_conf.yaml')],
extra_userdata)
if not user_data:
log.error("Problem building cloud-config user data.")
return self.ERROR
#with open('/tmp/userdata.yaml', 'w') as f:
# f.write(user_data)
# Compress the user data to try and get under the limit
user_data = utilities.gzip_userdata(user_data)
try:
if self.name in vm_image.keys():
image = vm_image[self.name]
elif self.network_address in vm_image.keys():
image = vm_image[self.network_address]
else:
image = vm_image['default']
except:
try:
vm_default_ami = _attr_list_to_dict(
config_val.get('job', 'default_VMAMI'))
if self.name in vm_default_ami.keys():
image = vm_default_ami[self.name]
else:
image = vm_default_ami[self.network_address]
except:
try:
image = vm_default_ami["default"]
except:
log.exception("Can't find a suitable AMI")
return
try:
imageobj = nova.glance.find_image(image)
except novaclient.exceptions.EndpointNotFound:
log.error(
"Endpoint not found, are your region settings correct for %s",
self.name)
return -4
except Exception as e:
log.warning(
"Exception occurred while trying to fetch image: %s %s", image,
e)
self.failed_image_set.add(image)
return
try:
if self.name in instance_type.keys():
i_type = instance_type[self.name]
elif self.network_address in instance_type.keys():
i_type = instance_type[self.network_address]
else:
i_type = instance_type['default']
except:
log.debug("No instance type for %s, trying default",
self.network_address)
try:
if self.name in self.DEFAULT_INSTANCE_TYPE_LIST.keys():
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[self.name]
else:
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[
self.network_address]
except:
log.debug(
"No default instance type found for %s, trying single default",
self.network_address)
i_type = self.DEFAULT_INSTANCE_TYPE
try:
flavor = nova.flavors.find(name=i_type)
except Exception as e:
log.warning(
"Exception occurred while trying to get flavor by name:"
" %s - will attempt to use name value as a uuid.", e)
try:
flavor = nova.flavors.get(i_type)
log.debug("Got flavor via uuid: %s", i_type)
except Exception as ex:
log.error(
"Exception occurred trying to get flavor by uuid: %s", ex)
return
self.flavor_set.add(flavor)
# find the network id to use if more than one network
if vm_networkassoc:
network = self._find_network(vm_networkassoc)
if network:
netid = [{'net-id': network.id}]
else:
log.debug("Unable to find network named: %s on %s",
vm_networkassoc, self.name)
if len(vm_networkassoc.split('-')) == 5: #uuid
netid = [{'net-id': vm_networkassoc}]
else:
netid = []
elif self.network_pools and len(self.network_pools) > 0:
network = self._find_network(self.network_pools[0])
if network:
netid = [{'net-id': network.id}]
else:
log.debug("Unable to find network named: %s on %s",
self.network_pools[0], self.name)
if len(self.network_pools[0].split('-')) == 5: #uuid
netid = [{'net-id': self.network_pools[0]}]
else:
netid = []
else:
netid = []
# Need to get the rotating hostname from the google code to use for here.
name = self._generate_next_name()
instance = None
if name:
try:
instance = nova.servers.create(
name=name,
image=imageobj,
flavor=flavor,
key_name=key_name,
availability_zone=self.placement_zone,
nics=netid,
userdata=user_data,
security_groups=sec_group)
except novaclient.exceptions.OverLimit as e:
log.info(
"Unable to create VM without exceeded quota on %s: %s",
self.name, e.message)
except Exception as e:
log.error("Unhandled exception while creating vm on %s: %s",
self.name, e)
if instance:
instance_id = instance.id
#if job didn't set a keep_alive use the clouds default
if not vm_keepalive and self.keep_alive:
vm_keepalive = self.keep_alive
new_vm = cluster_tools.VM(name=vm_name,
id=instance_id,
vmtype=vm_type,
user=vm_user,
clusteraddr=self.network_address,
hostname=name,
cloudtype=self.cloud_type,
network=vm_networkassoc,
image=vm_image,
flavor=flavor.name,
memory=vm_mem,
cpucores=vm_cores,
storage=vm_storage,
keep_alive=vm_keepalive,
job_per_core=job_per_core)
try:
self.resource_checkout(new_vm)
log.info("Launching 1 VM: %s on %s under tenant: %s",
instance_id, self.name, self.tenant_name)
except:
log.error(
"Unexpected Error checking out resources creating VM. Programming error?"
)
self.vm_destroy(new_vm,
reason="Failed Resource checkout",
return_resources=False)
return self.ERROR
self.vms.append(new_vm)
else:
log.debug("Failed to create instance on %s", self.name)
return self.ERROR
else:
log.debug("Unable to generate name for %s", self.name)
return self.ERROR
return 0
def vm_create(self, vm_name, vm_type, vm_user, vm_networkassoc,
vm_image, vm_mem, vm_cores, vm_storage, customization=None,
pre_customization=None, vm_keepalive=0, instance_type="",
job_per_core=False, securitygroup=[],
key_name="", use_cloud_init=False, extra_userdata=[]):
"""Attempt to boot a new VM on the cluster."""
use_cloud_init = use_cloud_init or config_val.get('global', 'use_cloud_init')
log.verbose("Trying to boot %s on %s", vm_type, self.network_address)
if len(securitygroup) != 0:
sec_group = []
for group in securitygroup:
if group in self.security_groups:
sec_group.append(group)
if len(sec_group) == 0:
log.debug("No matching security groups - trying default config")
sec_group = self.security_groups
else:
sec_group = self.security_groups
try:
if self.name in vm_image.keys():
vm_ami = vm_image[self.name]
else:
vm_ami = vm_image[self.network_address]
except:
log.debug("No AMI for %s, trying default", self.network_address)
try:
vm_default_ami = _attr_list_to_dict(config_val.get('job', 'default_VMAMI'))
if self.name in vm_default_ami.keys():
vm_ami = vm_default_ami[self.name]
else:
vm_ami = vm_default_ami[self.network_address]
except:
try:
vm_ami = vm_default_ami["default"]
except:
log.exception("Can't find a suitable AMI")
self.failed_image_set.add(vm_ami)
return
try:
if self.name in instance_type.keys():
i_type = instance_type[self.name]
elif self.network_address in instance_type.keys():
i_type = instance_type[self.network_address]
else:
i_type = instance_type["default"]
except:
log.debug("No instance type for %s, trying default", self.network_address)
try:
if self.name in self.DEFAULT_INSTANCE_TYPE_LIST.keys():
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[self.name]
else:
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[self.network_address]
except:
log.debug("No default instance type found for %s, trying single default", \
self.network_address)
i_type = self.DEFAULT_INSTANCE_TYPE
instance_type = i_type
if key_name == "" or key_name is None:
key_name = self.key_name if self.key_name else ""
if customization:
if not use_cloud_init:
user_data = nimbus_xml.ws_optional(customization)
else:
user_data = cloud_init_util.build_write_files_cloud_init(customization)
else:
user_data = ""
if pre_customization:
if not use_cloud_init:
for item in pre_customization:
user_data = '\n'.join([item, user_data])
else:
user_data = cloud_init_util.inject_customizations(pre_customization, user_data)
elif use_cloud_init:
user_data = cloud_init_util.inject_customizations([], user_data)[0]
if len(extra_userdata) > 0:
# need to use the multi-mime type functions
user_data = cloud_init_util.build_multi_mime_message([(user_data, 'cloud-config',
'cloud_conf.yaml')],
extra_userdata)
if "AmazonEC2" == self.cloud_type and vm_networkassoc != "public":
log.debug("You requested '%s' networking, but EC2 only supports 'public'", \
vm_networkassoc)
addressing_type = "public"
else:
addressing_type = vm_networkassoc
user_data = utilities.gzip_userdata(user_data)
try:
client = self._get_connection()
#Uncomment for debugging boto calls
#boto3.set_stream_logger('botocore')
#boto_file_handler = logging.handlers.WatchedFileHandler('/tmp/csboto3.yaml', )
#botolog = logging.getLogger('botocore')
#botolog.addHandler(boto_file_handler)
resp = client.run_instances(ImageId=vm_ami, MinCount=1, MaxCount=1,
InstanceType=instance_type, UserData=user_data,
KeyName=key_name, SecurityGroups=sec_group)
# will need to figure out how PlacementGroups will work
# probably just be Placement={"AvailabilityZone':placement_zone}
except Exception as e:
log.error("Problem creating instance %s", e.__dict__)
return self.ERROR
#if job didn't set a keep_alive use the clouds default
if not vm_keepalive and self.keep_alive:
vm_keepalive = self.keep_alive
if 'Instances' in resp.keys():
new_vm_id = resp['Instances'][0]['InstanceId']
else:
#print resp.keys()
new_vm_id = "unable to get id"
return self.ERROR
new_vm = cluster_tools.VM(name=vm_name, id=new_vm_id, vmtype=vm_type, user=vm_user,
clusteraddr=self.network_address,
cloudtype=self.cloud_type, network=vm_networkassoc,
image=vm_ami, flavor=instance_type,
memory=vm_mem, cpucores=vm_cores, storage=vm_storage,
keep_alive=vm_keepalive, job_per_core=job_per_core)
#try:
# new_vm.spot_id = spot_id
#except:
# log.verbose("No spot ID to add to VM %s" % instance_id)
try:
self.resource_checkout(new_vm)
except:
log.exception("Unexpected Error checking out resources when creating a VM. \
Programming error?")
self.vm_destroy(new_vm, reason="Failed Resource checkout", return_resources=False)
return self.ERROR
self.vms.append(new_vm)
return 0
def vm_create(self, vm_name, vm_type, vm_user,
vm_image, vm_mem, vm_cores, vm_storage, customization=None,
vm_keepalive=0, instance_type={}, job_per_core=False,
pre_customization=None, extra_userdata=None):
""" Create a VM on Azure."""
use_cloud_init = True
if customization:
user_data = cloud_init_util.build_write_files_cloud_init(customization)
else:
user_data = ""
if pre_customization:
user_data = cloud_init_util.inject_customizations(pre_customization, user_data)
elif use_cloud_init:
user_data = cloud_init_util.inject_customizations([], user_data)
if extra_userdata:
# need to use the multi-mime type functions
user_data = cloud_init_util.build_multi_mime_message([(user_data, 'cloud-config',
'cloud_conf.yaml')],
extra_userdata)
if not user_data:
log.error("Problem building cloud-config user data.")
return self.ERROR
# Compress the user data to try and get under the limit
user_data = utilities.gzip_userdata(user_data)
try:
if self.name in vm_image.keys():
image = vm_image[self.name]
elif self.network_address in vm_image.keys():
image = vm_image[self.network_address]
else:
image = vm_image['default']
except:
try:
vm_default_ami = _attr_list_to_dict(config_val.get('job', 'default_VMAMI'))
if self.name in vm_default_ami.keys():
image = vm_default_ami[self.name]
else:
image = vm_default_ami[self.network_address]
except:
try:
image = vm_default_ami["default"]
except:
log.exception("Can't find a suitable AMI")
return
try:
if self.name in instance_type.keys():
i_type = instance_type[self.name]
elif self.network_address in instance_type.keys():
i_type = instance_type[self.network_address]
else:
i_type = instance_type['default']
except:
log.debug("No instance type for %s, trying default", self.network_address)
try:
if self.name in self.DEFAULT_INSTANCE_TYPE_LIST.keys():
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[self.name]
else:
i_type = self.DEFAULT_INSTANCE_TYPE_LIST[self.network_address]
except:
log.debug("No default instance type found for %s, \
trying single default", self.network_address)
i_type = self.DEFAULT_INSTANCE_TYPE
name = self._generate_next_name()
instance = None
req = None
if name:
sms = self._get_service_connection()
try:
conf_set = azure.servicemanagement.\
LinuxConfigurationSet(host_name=name,
user_name=self.username,
user_password=self.password,
disable_ssh_password_authentication=False,
custom_data=user_data)
net_set = azure.servicemanagement.ConfigurationSet()
vm_ssh_port = 20000+self.count
net_set.input_endpoints.input_endpoints.append(
azure.servicemanagement.ConfigurationSetInputEndpoint(name='SSH',
protocol='TCP',
port=vm_ssh_port,
local_port=22))
self.count += 1
if self.count > 15000:
self.count = 0
os_hd = azure.servicemanagement.OSVirtualHardDisk(image, self.blob_url + name)
res = sms.check_hosted_service_name_availability(self.azure_service_name)
if res.result:
req = sms.create_hosted_service(self.azure_service_name,
self.azure_service_name,
location=self.regions[0])
sms.wait_for_operation_status(req.request_id)
if len(self.vms) == 0:
req = sms.create_virtual_machine_deployment(service_name=self.azure_service_name,
deployment_name=self.azure_service_name,
deployment_slot='production',
role_name=name, label=name,
system_config=conf_set,
network_config=net_set,
os_virtual_hard_disk=os_hd,
role_size=i_type)
try:
op_status = sms.wait_for_operation_status(req.request_id)
except Exception as e:
log.error("Problem creating VM on Azure: %s", e.result.error.message)
return 1
else:
req = sms.add_role(service_name=self.azure_service_name,
deployment_name=self.azure_service_name, role_name=name,
system_config=conf_set, network_config=net_set,
os_virtual_hard_disk=os_hd, role_size=i_type)
try:
op_status = sms.wait_for_operation_status(req.request_id)
except Exception as e:
log.error("Problem creating VM on Azure: %s", e.result.error.message)
return 1
except Exception as e:
log.error("Unhandled exception while creating vm on %s: %s", self.name, e)
return self.ERROR
if req:
#if job didn't set a keep_alive use the clouds default
if not vm_keepalive and self.keep_alive:
vm_keepalive = self.keep_alive
new_vm = cluster_tools.VM(name=vm_name, id=name, vmtype=vm_type, user=vm_user,
clusteraddr=self.network_address,
hostname=name,
cloudtype=self.cloud_type, network=None,
image=vm_image, flavor=i_type,
memory=vm_mem, cpucores=vm_cores, storage=vm_storage,
keep_alive=vm_keepalive, job_per_core=job_per_core,
ssh_port=vm_ssh_port)
try:
self.resource_checkout(new_vm)
log.info("Launching 1 VM: %s on %s", name, self.name)
except:
log.error("Unexpected Error checking out resources when creating a VM. \
Programming error?")
self.vm_destroy(new_vm, reason="Failed Resource checkout",
return_resources=False)
return self.ERROR
self.vms.append(new_vm)
else:
log.debug("Failed to create instance on %s", self.name)
return self.ERROR
else:
log.debug("Unable to generate name for %s", self.name)
return self.ERROR
return 0