def read(self, param):
try:
with open(os.path.join(self.path, param)) as f:
for line in f.readlines():
yield line.split()
except IOError, e:
msg = "Cgroup read error: %s/%s: %s" % (self.path, param, e)
logger.error(msg)
def write(self, param, value):
try:
with open(os.path.join(self.path, param), 'w') as f:
f.write(value)
logger.info("Writing cgroup: %s: %s" % (param, value))
except IOError, e:
msg = "Cgroup write error: %s/%s: %s" % (self.path, param, e)
logger.error(msg)
def remove(self, path):
for i in range(0, 10):
try:
return execute('lvremove', path, force=None)
except ProcessError:
sleep(1)
continue
logger.error("Failed to delete volume '%s' after 10 tries" % path)
raise
def _assign_node(self, volume, backup, source, nodes):
"""
Assigns the new volume to a node.
:returns: dict, node response on successful placement
:raises: HTTPError
"""
request_params = {
'account': self.account_id,
'size': volume.size,
'read_iops': volume.volume_type.read_iops,
'write_iops': volume.volume_type.write_iops,
}
if volume.image_id:
request_params['image_id'] = volume.image_id
if backup:
request_params['backup_source_volume_id'] = backup.volume.name
request_params['backup_id'] = backup.id
volume.restore_of = backup.id
if source:
request_params['source_volume_id'] = source.name
request_params['source_host'] = source.node.hostname
request_params['source_port'] = source.node.port
last_node_error = None
for node in nodes:
volume.node = node
self.db.commit() # prevent duplicate/lost volumes
try:
path = '/volumes/%s' % volume.name
return self.node_request(node, 'PUT', path, **request_params)
except NodeError, e:
last_node_error = e
# log server errors and continue
if (e.code // 100) == 5:
logger.error(str(e))
continue
# pass client error up to user, immediately
break
def get_registration_exceptions(local_info, node_info):
exceptions = {}
for k, v in local_info.items():
if 'hostname' in k and node_info[k] != v:
try:
node_value = socket.gethostbyname(node_info[k])
except socket.error:
# skip hostname translation on failure
pass
else:
try:
node_value = node_info[k]
except KeyError, e:
logger.error("During registration; missing '%s' key in api "
"server response" % k)
continue
if node_value != v:
logger.warning("Invalid '%s' registered "
"as %r != %r" % (k, node_value, v))
exceptions[k] = v
else:
logger.info("Verified '%s' registered as '%s'" % (k, v))
def create_clone(self, volume_id, clone_id, iqn, iscsi_ip, iscsi_port,
callback=None, lock=None, cinder=None):
volume = self.get(volume_id)
size = volume['size'] / 1024 / 1024 / 1024
logger.info("Cloning source '%s' to volume '%s'" %
(volume_id, clone_id))
snapshot_name = uuid.uuid4()
snapshot = self.create_snapshot(volume_id, snapshot_name,
clone_id=clone_id, type_='clone')
logger.info("Snapshot to clone id: '%s'" % snapshot['id'])
try:
new_volume = ISCSIDevice(iqn, iscsi_ip, iscsi_port)
new_volume.connect()
except (ISCSILoginFailed, ISCSINotConnected):
msg = "Unable to open iscsi connection to %s:%s - %s" % \
(iscsi_ip, iscsi_port, iqn)
logger.error(msg)
self.delete(snapshot['id'])
raise ServiceUnavailable(msg)
spawn(lock, self._copy_clone, snapshot, clone_id, size, new_volume,
cinder, callback=callback, skip_fork=self.skip_fork)
def __call__(self, request):
"""Handle WSGI request."""
try:
try:
# Call the implementation to handle the action
return self.encode_response(self.call(request))
except HTTPException:
raise
except Exception, e:
logger.exception("Caught unknown exception, traceback dumped")
raise HTTPInternalServerError("Internal controller error")
except HTTPException, e:
# Avoid logging HTTPOk Exceptions
if isinstance(e, HTTPError):
logger.error("%s" % e)
# (thrawn01) returning a HTTPError exception to WSGI
# container would result in a text/html content type
# provided by webob, this is not desired
body = {
'reason': e.detail,
'request-id': request.headers.get('x-request-id', '-')
}
return self.encode_response(Response(body=body, status=e.status))
def _do_create(self, volume_id, size_str, tag,
backup_source_volume_id=None):
try:
out = execute('lvcreate', self.volume_group,
name=volume_id, size=size_str, addtag=tag)
except ProcessError, e:
if not e.errcode == 5 and 'already exists' not in e.err:
raise
# We ran out of space on the storage node!
if "Insufficient free extents" in e.err \
or "insufficient free space" in e.err:
logger.error(e.err)
raise ServiceUnavailable("LVM reports insufficient "
"free space on drive")
# If we are requesting a restore, and the existing volume is this
# same failed restore, it's not an error.
if backup_source_volume_id and backup_source_volume_id == \
self.get(volume_id).get('backup_source_volume_id', False):
logger.info("Restarting failed restore on '%s'" % volume_id)
else:
raise AlreadyExists("Unable to create a new volume named "
"'%s' because one already exists." %
volume_id)
path = '/volumes/%s/clones/%s' % (source['id'], volume['id'])
node_params = {
'account': req.params.get('account', ''),
'iqn': export['name'],
'iscsi_ip': self.helper.storage_host,
'iscsi_port': self.helper.storage_port,
# First dirty method to close the export.
'mgmt_host': self.helper.management_host,
'mgmt_port': self.helper.management_port,
'cinder_host': self.helper.cinder_host,
}
try:
self.helper.node_request(source['host'], source['port'], 'PUT',
path, **node_params)
except NodeError, e:
logger.error('Clone node request failed: %s' % e)
self.helper.exports.delete(volume['id'])
self.helper.volumes.delete(volume['id'], lock=lock)
raise
volume['status'] = 'CLONING'
# Create from image
elif req.params.get('image_id'):
image_id = params['image_id'] = req.params['image_id']
account = params['account'] = req.params.get('account')
params['callback'] = self._create_from_image_cb(req, iops)
params['scrub_callback'] = self._post_scrub_cb()
try:
self.helper.volumes.create(self.id, **params)
except InvalidImage, e:
path = '/volumes/%s/clones/%s' % (source['id'], volume['id'])
node_params = {
'account': req.params.get('account', ''),
'iqn': export['name'],
'iscsi_ip': self.helper.storage_host,
'iscsi_port': self.helper.storage_port,
# First dirty method to close the export.
'mgmt_host': self.helper.management_host,
'mgmt_port': self.helper.management_port,
'cinder_host': self.helper.cinder_host,
}
try:
self.helper.node_request(source['host'], source['port'],
'PUT', path, **node_params)
except NodeError, e:
logger.error('Clone node request failed: %s' % e)
self.helper.exports.delete(volume['id'])
self.helper.volumes.delete(volume['id'], lock=lock)
raise
volume['status'] = 'CLONING'
# Create from image
elif req.params.get('image_id'):
image_id = params['image_id'] = req.params['image_id']
account = params['account'] = req.params.get('account')
params['callback'] = self._create_from_image_cb(req, iops)
params['scrub_callback'] = self._post_scrub_cb()
try:
self.helper.volumes.create(self.id, **params)
except InvalidImage, e:
else:
convert_gbs = self.convert_gbs
tmp_vol = self.create_convert_scratch(image, convert_gbs)
except GlanceError, e:
logger.warning("Error fetching glance image: %s" % e)
raise InvalidImage("Error fetching image: %s" % image_id)
# Create a tag to apply to the lvm volume
tag = encode_tag(backup_source_volume_id=backup_source_volume_id,
backup_id=backup_id)
try:
self._do_create(volume_id, size_str, tag, backup_source_volume_id)
except Exception, e:
# If we ran out of space due to the tmp_vol
logger.error('Failed to create volume: %s' % e)
# Update cinder immediately.
if callback:
callback()
if tmp_vol:
spawn(NullResource(), self.remove_lvm_volume, tmp_vol,
callback=scrub_callback, skip_fork=self.skip_fork)
raise
def log_duration():
duration = time() - op_start
parts = ['STAT: Create Volume']
if volume_id:
parts.append('Volume_ID: %s' % (volume_id,))
if backup_id:
class CgroupHelper(object):
cgroup_parameters = [
'blkio.throttle.read_iops_device',
'blkio.throttle.write_iops_device',
]
def __init__(self, conf):
sys_fs_path = conf.string('cgroup', 'cgroup_path',
'/sys/fs/cgroup/blkio/sysdefault')
self.cgroup_fs = CgroupFs(sys_fs_path)
run_dir = conf.string('storage', 'run_dir', conf.path('run'))
self.cgroups_path = os.path.join(run_dir, 'cgroups')
def _updates_path(self):
return os.path.join(self.cgroups_path, 'updates')
def all_cgroups(self):
data = defaultdict(dict)
for name in self.cgroup_parameters:
for device, throttle in self.cgroup_fs.read(name):
data[name][device] = throttle
return data
def get(self, volume):
device = volume['device_number']
data = {}
cgroups = self.all_cgroups()
for name, device_map in cgroups.items():
if device in device_map:
data[name] = device_map[device]
return data
def set_read_iops(self, volume, throttle):
self.set(volume, throttle, 'blkio.throttle.read_iops_device')
def set_write_iops(self, volume, throttle):
self.set(volume, throttle, 'blkio.throttle.write_iops_device')
def set(self, volume, throttle, param=None):
if not param:
params = self.cgroup_parameters
else:
params = [param]
throttle = int(throttle)
if throttle < 0:
raise ValueError("Throttle cannot be negative")
device = volume['device_number']
for name in params:
value = "%s %s" % (device, throttle)
self.cgroup_fs.write(name, value)
self.save_update(volume, name, throttle)
def save_update(self, volume, name, throttle):
try:
os.makedirs(self.cgroups_path)
except OSError, e:
if e.errno != errno.EEXIST:
raise
try:
with open(self._updates_path(), 'a') as f:
fcntl.lockf(f.fileno(), fcntl.LOCK_EX)
entry = '%s %s %s\n' % (volume['id'], name, throttle)
f.write(entry)
except IOError, e:
logger.error('Failed writing cgroup update: %s' % e)
else:
convert_gbs = self.convert_gbs
tmp_vol = self.create_convert_scratch(image, convert_gbs)
except GlanceError, e:
logger.warning("Error fetching glance image: %s" % e)
raise InvalidImage("Error fetching image: %s" % image_id)
# Create a tag to apply to the lvm volume
tag = encode_tag(backup_source_volume_id=backup_source_volume_id,
backup_id=backup_id)
try:
self._do_create(volume_id, size_str, tag, backup_source_volume_id)
except Exception, e:
# If we ran out of space due to the tmp_vol
logger.error('Failed to create volume: %s' % e)
# Update cinder immediately.
if callback:
callback()
if tmp_vol:
spawn(NullResource(),
self.remove_lvm_volume,
tmp_vol,
callback=scrub_callback,
skip_fork=self.skip_fork)
raise
def log_duration():
duration = time() - op_start
parts = ['STAT: Create Volume']