def _setup(self):
# initialize helper classes for interacting with GCE, GCS
self.auth_http = instance.oauth_authorization(self.config)
self.gce_helper = gce.Gce(self.auth_http, self.config, project_id=self.config['compute']['project'])
self.gcs_helper = gcs.Gcs(self.auth_http, self.config, project_id=self.config['compute']['project'])
self.instance_manager = instance.VMInstanceManager()
self.instances = []
self.launched_instances = {}
self.user_terminated = False
# get job id
self.id = gen_job_id()
self.job_name_root = 'job-%s' %(self.id)
if self.config['update']:
self.job_name_root = 'job-updater-%s' %(self.id)
# setup instance completion api call
service_not_ready = True
while service_not_ready:
try:
service = discovery.build('storage', self.config['compute']['api_version'], http=self.gce_helper.auth_http)
self.bucket_req = service.objects().list(bucket=self.config['compute']['bucket'], prefix=self.job_name_root)
service_not_ready = False
except (ValueError, Exception) as e:
logging.info('Connection failed. Retrying...')
def stage(self):
""" Stages changes by creating update disks, if they don't already exist. """
# setup vars
compute_config = self.config_['compute']
created_snapshots = False
if not self.update_data_disks_:
self.compute_update_data_disks()
# authorize access to GCE api
auth_http = instance.oauth_authorization(self.config_)
gce_helper = gce.Gce(auth_http, self.config_, project_id=compute_config['project'])
# for all zones, create a disk snapshot if they don't already exist
for zone, disk, update_disk_name in zip(compute_config['zones'], compute_config['data_disks'], self.update_data_disks_):
# check for existence of the update disk (taken as a flag for the existence of an update node)
disk_valid = gce_helper.get_disk(update_disk_name, zone)
if not disk_valid:
# create a snapshot of the current disk
logging.info('Snapshotting disk %s' %(disk))
snapshot_response = gce_helper.snapshot_disk(disk, compute_config['project'], zone)
# create a disk from the snapshot
logging.info('Creating update disk %s from snapshot %s' %(update_disk_name, snapshot_response['snapshot_name']))
gce_helper.create_disk(update_disk_name, zone=zone, size_gb=compute_config['disk_size_gb'],
source_snapshot=snapshot_response['snapshot_name'])
# delete the snapshot
ss_del_response = gce_helper.delete_snapshot(snapshot_name=snapshot_response['snapshot_name'], project=compute_config['project'])
created_snapshots = True
return created_snapshots
def _setup(self):
# initialize helper classes for interacting with GCE, GCS
self.auth_http = instance.oauth_authorization(self.config)
self.gce_helper = gce.Gce(self.auth_http,
self.config,
project_id=self.config['compute']['project'])
self.gcs_helper = gcs.Gcs(self.auth_http,
self.config,
project_id=self.config['compute']['project'])
self.instance_manager = instance.VMInstanceManager()
self.instances = []
self.launched_instances = {}
self.user_terminated = False
# get job id
self.id = gen_job_id()
self.job_name_root = 'job-%s' % (self.id)
if self.config['update']:
self.job_name_root = 'job-updater-%s' % (self.id)
# setup instance completion api call
service_not_ready = True
while service_not_ready:
try:
service = discovery.build(
'storage',
self.config['compute']['api_version'],
http=self.gce_helper.auth_http)
self.bucket_req = service.objects().list(
bucket=self.config['compute']['bucket'],
prefix=self.job_name_root)
service_not_ready = False
except (ValueError, Exception) as e:
logging.info('Connection failed. Retrying...')
def push(self):
""" Pushed changes by replacing original disks with update disks. Super critical section. """
# setup vars
compute_config = self.config_['compute']
dt_now = dt.datetime.now()
if not self.update_data_disks_:
self.compute_update_data_disks()
# authorize access to GCE api
auth_http = instance.oauth_authorization(self.config_)
gce_helper = gce.Gce(auth_http, self.config_, project_id=compute_config['project'])
for zone, disk, update_disk in zip(compute_config['zones'], compute_config['data_disks'], self.update_data_disks_):
# check for update disk existence
disk_response = gce_helper.get_disk(update_disk, zone)
if not disk_response:
logging.error('Update disk %s does not exist' %(update_disk))
continue
# generate backup disk filename
backup_disk = '%s-backup-%s-%s-%s-%sh-%sm-%ss' %(disk, dt_now.month, dt_now.day, dt_now.year, dt_now.hour, dt_now.minute, dt_now.second)
# snapshot the updated data disks
snapshot_response = gce_helper.snapshot_disk(update_disk, compute_config['project'], zone)
# delete previous disk and replace, if not in use
disk_response = gce_helper.get_disk(disk, zone)
if disk_response:
if USERS_KEY not in disk_response.keys() or (USERS_KEY in disk_response.keys() and len(disk_response[USERS_KEY]) == 0):
# create new disk from snapshot
gce_helper.delete_disk(disk)
gce_helper.create_disk(disk, zone=zone, size_gb=compute_config['disk_size_gb'],
source_snapshot=snapshot_response['snapshot_name'])
# delete update disk (don't delete if push can't be done now, otherwise changes won't be overwritten)
gce_helper.delete_disk(update_disk)
elif USERS_KEY in disk_response.keys() and len(disk_response[USERS_KEY]) > 0:
# stage the push for a future time
logging.info('Master disk in use. Staging backup disk for a future push')
push_queue_filename = os.path.join(self.cache_dir_, PUSH_QUEUE_FILE)
f = open(push_queue_filename, 'a')
f.write(backup_disk + '\n')
else:
logging.warning('Master disk was not found')
# create backup disk from snapshot
gce_helper.create_disk(backup_disk, zone=zone, size_gb=compute_config['disk_size_gb'],
source_snapshot=snapshot_response['snapshot_name'])
# delete the snapshot
ss_del_response = gce_helper.delete_snapshot(snapshot_name=snapshot_response['snapshot_name'], project=compute_config['project'])
return True
def stage(self):
""" Stages changes by creating update disks, if they don't already exist. """
# setup vars
compute_config = self.config_['compute']
created_snapshots = False
if not self.update_data_disks_:
self.compute_update_data_disks()
# authorize access to GCE api
auth_http = instance.oauth_authorization(self.config_)
gce_helper = gce.Gce(auth_http,
self.config_,
project_id=compute_config['project'])
# for all zones, create a disk snapshot if they don't already exist
for zone, disk, update_disk_name in zip(compute_config['zones'],
compute_config['data_disks'],
self.update_data_disks_):
# check for existence of the update disk (taken as a flag for the existence of an update node)
disk_valid = gce_helper.get_disk(update_disk_name, zone)
if not disk_valid:
# create a snapshot of the current disk
logging.info('Snapshotting disk %s' % (disk))
snapshot_response = gce_helper.snapshot_disk(
disk, compute_config['project'], zone)
# create a disk from the snapshot
logging.info(
'Creating update disk %s from snapshot %s' %
(update_disk_name, snapshot_response['snapshot_name']))
gce_helper.create_disk(
update_disk_name,
zone=zone,
size_gb=compute_config['disk_size_gb'],
source_snapshot=snapshot_response['snapshot_name'])
# delete the snapshot
ss_del_response = gce_helper.delete_snapshot(
snapshot_name=snapshot_response['snapshot_name'],
project=compute_config['project'])
created_snapshots = True
return created_snapshots
def push(self):
""" Pushed changes by replacing original disks with update disks. Super critical section. """
# setup vars
compute_config = self.config_['compute']
dt_now = dt.datetime.now()
if not self.update_data_disks_:
self.compute_update_data_disks()
# authorize access to GCE api
auth_http = instance.oauth_authorization(self.config_)
gce_helper = gce.Gce(auth_http,
self.config_,
project_id=compute_config['project'])
for zone, disk, update_disk in zip(compute_config['zones'],
compute_config['data_disks'],
self.update_data_disks_):
# check for update disk existence
disk_response = gce_helper.get_disk(update_disk, zone)
if not disk_response:
logging.error('Update disk %s does not exist' % (update_disk))
continue
# generate backup disk filename
backup_disk = '%s-backup-%s-%s-%s-%sh-%sm-%ss' % (
disk, dt_now.month, dt_now.day, dt_now.year, dt_now.hour,
dt_now.minute, dt_now.second)
# snapshot the updated data disks
snapshot_response = gce_helper.snapshot_disk(
update_disk, compute_config['project'], zone)
# delete previous disk and replace, if not in use
disk_response = gce_helper.get_disk(disk, zone)
if disk_response:
if USERS_KEY not in disk_response.keys() or (
USERS_KEY in disk_response.keys()
and len(disk_response[USERS_KEY]) == 0):
# create new disk from snapshot
gce_helper.delete_disk(disk)
gce_helper.create_disk(
disk,
zone=zone,
size_gb=compute_config['disk_size_gb'],
source_snapshot=snapshot_response['snapshot_name'])
# delete update disk (don't delete if push can't be done now, otherwise changes won't be overwritten)
gce_helper.delete_disk(update_disk)
elif USERS_KEY in disk_response.keys() and len(
disk_response[USERS_KEY]) > 0:
# stage the push for a future time
logging.info(
'Master disk in use. Staging backup disk for a future push'
)
push_queue_filename = os.path.join(self.cache_dir_,
PUSH_QUEUE_FILE)
f = open(push_queue_filename, 'a')
f.write(backup_disk + '\n')
else:
logging.warning('Master disk was not found')
# create backup disk from snapshot
gce_helper.create_disk(
backup_disk,
zone=zone,
size_gb=compute_config['disk_size_gb'],
source_snapshot=snapshot_response['snapshot_name'])
# delete the snapshot
ss_del_response = gce_helper.delete_snapshot(
snapshot_name=snapshot_response['snapshot_name'],
project=compute_config['project'])
return True