Initial fork out of Nova.
This commit is contained in:
234
cinder/quota.py
Normal file
234
cinder/quota.py
Normal file
@@ -0,0 +1,234 @@
|
||||
# vim: tabstop=4 shiftwidth=4 softtabstop=4
|
||||
|
||||
# Copyright 2010 United States Government as represented by the
|
||||
# Administrator of the National Aeronautics and Space Administration.
|
||||
# All Rights Reserved.
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License"); you may
|
||||
# not use this file except in compliance with the License. You may obtain
|
||||
# a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
|
||||
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
|
||||
# License for the specific language governing permissions and limitations
|
||||
# under the License.
|
||||
|
||||
"""Quotas for instances, volumes, and floating ips."""
|
||||
|
||||
from cinder import db
|
||||
from cinder.openstack.common import cfg
|
||||
from cinder import flags
|
||||
|
||||
|
||||
quota_opts = [
|
||||
cfg.IntOpt('quota_instances',
|
||||
default=10,
|
||||
help='number of instances allowed per project'),
|
||||
cfg.IntOpt('quota_cores',
|
||||
default=20,
|
||||
help='number of instance cores allowed per project'),
|
||||
cfg.IntOpt('quota_ram',
|
||||
default=50 * 1024,
|
||||
help='megabytes of instance ram allowed per project'),
|
||||
cfg.IntOpt('quota_volumes',
|
||||
default=10,
|
||||
help='number of volumes allowed per project'),
|
||||
cfg.IntOpt('quota_gigabytes',
|
||||
default=1000,
|
||||
help='number of volume gigabytes allowed per project'),
|
||||
cfg.IntOpt('quota_floating_ips',
|
||||
default=10,
|
||||
help='number of floating ips allowed per project'),
|
||||
cfg.IntOpt('quota_metadata_items',
|
||||
default=128,
|
||||
help='number of metadata items allowed per instance'),
|
||||
cfg.IntOpt('quota_injected_files',
|
||||
default=5,
|
||||
help='number of injected files allowed'),
|
||||
cfg.IntOpt('quota_injected_file_content_bytes',
|
||||
default=10 * 1024,
|
||||
help='number of bytes allowed per injected file'),
|
||||
cfg.IntOpt('quota_injected_file_path_bytes',
|
||||
default=255,
|
||||
help='number of bytes allowed per injected file path'),
|
||||
cfg.IntOpt('quota_security_groups',
|
||||
default=10,
|
||||
help='number of security groups per project'),
|
||||
cfg.IntOpt('quota_security_group_rules',
|
||||
default=20,
|
||||
help='number of security rules per security group'),
|
||||
]
|
||||
|
||||
FLAGS = flags.FLAGS
|
||||
FLAGS.register_opts(quota_opts)
|
||||
|
||||
|
||||
quota_resources = ['metadata_items', 'injected_file_content_bytes',
|
||||
'volumes', 'gigabytes', 'ram', 'floating_ips', 'instances',
|
||||
'injected_files', 'cores', 'security_groups', 'security_group_rules']
|
||||
|
||||
|
||||
def _get_default_quotas():
|
||||
defaults = {
|
||||
'instances': FLAGS.quota_instances,
|
||||
'cores': FLAGS.quota_cores,
|
||||
'ram': FLAGS.quota_ram,
|
||||
'volumes': FLAGS.quota_volumes,
|
||||
'gigabytes': FLAGS.quota_gigabytes,
|
||||
'floating_ips': FLAGS.quota_floating_ips,
|
||||
'metadata_items': FLAGS.quota_metadata_items,
|
||||
'injected_files': FLAGS.quota_injected_files,
|
||||
'injected_file_content_bytes':
|
||||
FLAGS.quota_injected_file_content_bytes,
|
||||
'security_groups': FLAGS.quota_security_groups,
|
||||
'security_group_rules': FLAGS.quota_security_group_rules,
|
||||
}
|
||||
# -1 in the quota flags means unlimited
|
||||
return defaults
|
||||
|
||||
|
||||
def get_class_quotas(context, quota_class, defaults=None):
|
||||
"""Update defaults with the quota class values."""
|
||||
|
||||
if not defaults:
|
||||
defaults = _get_default_quotas()
|
||||
|
||||
quota = db.quota_class_get_all_by_name(context, quota_class)
|
||||
for key in defaults.keys():
|
||||
if key in quota:
|
||||
defaults[key] = quota[key]
|
||||
|
||||
return defaults
|
||||
|
||||
|
||||
def get_project_quotas(context, project_id):
|
||||
defaults = _get_default_quotas()
|
||||
if context.quota_class:
|
||||
get_class_quotas(context, context.quota_class, defaults)
|
||||
quota = db.quota_get_all_by_project(context, project_id)
|
||||
for key in defaults.keys():
|
||||
if key in quota:
|
||||
defaults[key] = quota[key]
|
||||
return defaults
|
||||
|
||||
|
||||
def _get_request_allotment(requested, used, quota):
|
||||
if quota == -1:
|
||||
return requested
|
||||
return quota - used
|
||||
|
||||
|
||||
def allowed_instances(context, requested_instances, instance_type):
|
||||
"""Check quota and return min(requested_instances, allowed_instances)."""
|
||||
project_id = context.project_id
|
||||
context = context.elevated()
|
||||
requested_cores = requested_instances * instance_type['vcpus']
|
||||
requested_ram = requested_instances * instance_type['memory_mb']
|
||||
usage = db.instance_data_get_for_project(context, project_id)
|
||||
used_instances, used_cores, used_ram = usage
|
||||
quota = get_project_quotas(context, project_id)
|
||||
allowed_instances = _get_request_allotment(requested_instances,
|
||||
used_instances,
|
||||
quota['instances'])
|
||||
allowed_cores = _get_request_allotment(requested_cores, used_cores,
|
||||
quota['cores'])
|
||||
allowed_ram = _get_request_allotment(requested_ram, used_ram, quota['ram'])
|
||||
if instance_type['vcpus']:
|
||||
allowed_instances = min(allowed_instances,
|
||||
allowed_cores // instance_type['vcpus'])
|
||||
if instance_type['memory_mb']:
|
||||
allowed_instances = min(allowed_instances,
|
||||
allowed_ram // instance_type['memory_mb'])
|
||||
|
||||
return min(requested_instances, allowed_instances)
|
||||
|
||||
|
||||
def allowed_volumes(context, requested_volumes, size):
|
||||
"""Check quota and return min(requested_volumes, allowed_volumes)."""
|
||||
project_id = context.project_id
|
||||
context = context.elevated()
|
||||
size = int(size)
|
||||
requested_gigabytes = requested_volumes * size
|
||||
used_volumes, used_gigabytes = db.volume_data_get_for_project(context,
|
||||
project_id)
|
||||
quota = get_project_quotas(context, project_id)
|
||||
allowed_volumes = _get_request_allotment(requested_volumes, used_volumes,
|
||||
quota['volumes'])
|
||||
allowed_gigabytes = _get_request_allotment(requested_gigabytes,
|
||||
used_gigabytes,
|
||||
quota['gigabytes'])
|
||||
if size != 0:
|
||||
allowed_volumes = min(allowed_volumes,
|
||||
int(allowed_gigabytes // size))
|
||||
return min(requested_volumes, allowed_volumes)
|
||||
|
||||
|
||||
def allowed_floating_ips(context, requested_floating_ips):
|
||||
"""Check quota and return min(requested, allowed) floating ips."""
|
||||
project_id = context.project_id
|
||||
context = context.elevated()
|
||||
used_floating_ips = db.floating_ip_count_by_project(context, project_id)
|
||||
quota = get_project_quotas(context, project_id)
|
||||
allowed_floating_ips = _get_request_allotment(requested_floating_ips,
|
||||
used_floating_ips,
|
||||
quota['floating_ips'])
|
||||
return min(requested_floating_ips, allowed_floating_ips)
|
||||
|
||||
|
||||
def allowed_security_groups(context, requested_security_groups):
|
||||
"""Check quota and return min(requested, allowed) security groups."""
|
||||
project_id = context.project_id
|
||||
context = context.elevated()
|
||||
used_sec_groups = db.security_group_count_by_project(context, project_id)
|
||||
quota = get_project_quotas(context, project_id)
|
||||
allowed_sec_groups = _get_request_allotment(requested_security_groups,
|
||||
used_sec_groups,
|
||||
quota['security_groups'])
|
||||
return min(requested_security_groups, allowed_sec_groups)
|
||||
|
||||
|
||||
def allowed_security_group_rules(context, security_group_id,
|
||||
requested_rules):
|
||||
"""Check quota and return min(requested, allowed) sec group rules."""
|
||||
project_id = context.project_id
|
||||
context = context.elevated()
|
||||
used_rules = db.security_group_rule_count_by_group(context,
|
||||
security_group_id)
|
||||
quota = get_project_quotas(context, project_id)
|
||||
allowed_rules = _get_request_allotment(requested_rules,
|
||||
used_rules,
|
||||
quota['security_group_rules'])
|
||||
return min(requested_rules, allowed_rules)
|
||||
|
||||
|
||||
def _calculate_simple_quota(context, resource, requested):
|
||||
"""Check quota for resource; return min(requested, allowed)."""
|
||||
quota = get_project_quotas(context, context.project_id)
|
||||
allowed = _get_request_allotment(requested, 0, quota[resource])
|
||||
return min(requested, allowed)
|
||||
|
||||
|
||||
def allowed_metadata_items(context, requested_metadata_items):
|
||||
"""Return the number of metadata items allowed."""
|
||||
return _calculate_simple_quota(context, 'metadata_items',
|
||||
requested_metadata_items)
|
||||
|
||||
|
||||
def allowed_injected_files(context, requested_injected_files):
|
||||
"""Return the number of injected files allowed."""
|
||||
return _calculate_simple_quota(context, 'injected_files',
|
||||
requested_injected_files)
|
||||
|
||||
|
||||
def allowed_injected_file_content_bytes(context, requested_bytes):
|
||||
"""Return the number of bytes allowed per injected file content."""
|
||||
resource = 'injected_file_content_bytes'
|
||||
return _calculate_simple_quota(context, resource, requested_bytes)
|
||||
|
||||
|
||||
def allowed_injected_file_path_bytes(context):
|
||||
"""Return the number of bytes allowed in an injected file path."""
|
||||
return FLAGS.quota_injected_file_path_bytes
|
||||
Reference in New Issue
Block a user