Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
259 lines (209 sloc) 10.8 KB
# Copyright (c) 2011 OpenStack Foundation
# Copyright (c) 2015 Spanish National Research Council (CSIC)
# Copyright 2015 INDIGO-DataCloud
# 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
# 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.
The FilterScheduler taking into account preemptible instances.
import random
from oslo_config import cfg
from oslo_log import log as logging
from nova import compute
from nova import exception
from nova.i18n import _, _LI
from nova import rpc
from nova.scheduler import filter_scheduler as nova_filter_scheduler
from nova.scheduler import scheduler_options
opts = [
help='Which weight class names to use for weighing the '
'selection of preemptible instances for termination'),
CONF.register_opts(opts, group="preemptible_instances_scheduler")
LOG = logging.getLogger(__name__)
class FilterScheduler(nova_filter_scheduler.FilterScheduler):
"""Scheduler that will take into account preemptible instances."""
def __init__(self, *args, **kwargs):
super(FilterScheduler, self).__init__(*args, **kwargs)
self.options = scheduler_options.SchedulerOptions()
self.notifier = rpc.get_notifier('scheduler')
self.compute_api = compute.API()
# self.weight_handler = weights.SpotWeightHandler()
# weigher_classes = self.weight_handler.get_matching_classes(
# CONF.preemptible.weight_classes)
# self.weighers = [cls() for cls in weigher_classes]
def select_destinations(self, context, spec_obj):
"""Selects a filtered set of hosts and nodes."""
context, 'scheduler.select_destinations.start',
num_instances = spec_obj.num_instances
selected_hosts = self._schedule(context, spec_obj)
# Couldn't fulfill the request_spec
if len(selected_hosts) < num_instances:
# NOTE(Rui Chen): If multiple creates failed, set the updated time
# of selected HostState to None so that these HostStates are
# refreshed according to database in next schedule, and release
# the resource consumed by instance in the process of selecting
# host.
for host in selected_hosts:
host.obj.updated = None
# Log the details but don't put those into the reason since
# we don't want to give away too much information about our
# actual environment.
LOG.debug('There are %(hosts)d hosts available but '
'%(num_instances)d instances requested to build.',
{'hosts': len(selected_hosts),
'num_instances': num_instances})
reason = _('There are not enough hosts available.')
raise exception.NoValidHost(reason=reason)
# NOTE(aloga): Detect if we are overcomitting here. If so, we need to
# delete some preemptible instances
dests = []
for host in selected_hosts:
if (self.detect_overcommit(host.obj) and
not self._is_preemptible_request(spec_obj)):
preemptibles = self.select_preemptibles_from_host(host.obj,
self.terminate_preemptible_instances(context, preemptibles)
dests.append(dict(, nodename=host.obj.nodename,
context, 'scheduler.select_destinations.end',
return dests
def terminate_preemptible_instances(self, context, instances):
"""Terminate the selected preemptible instances."""
# NOTE(aloga): we should not delete them directly, but probably send
# them a signal so that the user is able to save her work.
elevated = context.elevated()
for instance in instances:"Deleting %(uuid)s") % {"uuid": instance["uuid"]})
instance = self.compute_api.get(elevated,
self.compute_api.delete(elevated, instance)
def select_preemptibles_from_host(self, host, request):
"""Select preemptible instances to be killed for the request."""
preemptibles = [i for i in host.instances.values()
if i.system_metadata.get("preemptible")]
if not preemptibles:
# Log the details but don't put those into the reason since
# we don't want to give away too much information about our
# actual environment.
LOG.debug('Need to terminate preemptible instances, but there'
'are no preemptible instances on %(host)s' %
{'host': host})
reason = _('Cannot terminate enough preemptible instances.')
raise exception.NoValidHost(reason=reason)
# FIXME(aloga): This needs to be fixed, as we are assuming that killing
# one instance will free enough resources
return [preemptibles.pop()]
def detect_overcommit(self, host):
"""Detect overcommit of resources, according to configured ratios."""
ratio = host.ram_allocation_ratio or 1
ram_limit = host.total_usable_ram_mb * ratio
used_ram = host.total_usable_ram_mb - host.free_ram_mb
if used_ram > ram_limit:
return True
ratio = host.disk_allocation_ratio or 1
disk_limit = host.total_usable_disk_gb * ratio
used_disk = host.total_usable_disk_gb - host.free_disk_mb / 1024.
if used_disk > disk_limit:
return True
ratio = host.cpu_allocation_ratio or 1
cpus_limit = host.vcpus_total * ratio
if host.vcpus_used > cpus_limit:
return True
return False
def _schedule(self, context, spec_obj):
"""Returns a list of hosts that meet the required specs,
ordered by their fitness.
elevated = context.elevated()
# Find our local list of acceptable hosts by repeatedly
# filtering and weighing our options. Each time we choose a
# host, we virtually consume resources on it so subsequent
# selections can adjust accordingly.
# Note: remember, we are using an iterator here. So only
# traverse this list once. This can bite you if the hosts
# are being scanned in a filter or weighing function.
# If the request is for a preemptible instace, take into account all
# resources used on the host. However, if the request is for a normal
# instance, do not take into account the preemptible instances. This
# way we can schedule normal requests even when there is no room for
# them without doing a retry cycle.
if self._is_preemptible_request(spec_obj):
hosts = self._get_all_host_states(elevated, partial=False)
hosts = self._get_all_host_states(elevated, partial=True)
hosts_full_state = self._get_all_host_states(elevated, partial=False)
selected_hosts = []
num_instances = spec_obj.num_instances
for num in range(num_instances):
# Filter local hosts based on requirements ...
hosts = self.host_manager.get_filtered_hosts(hosts,
spec_obj, index=num)
if not hosts:
# Can't get any more locally.
LOG.debug("Filtered %(hosts)s", {'hosts': hosts})
# Get the full host states for weighing. The filtered list of
# hosts does not take into account preemptible instances, but we
# need them for weighing
hosts_full_state = list(hosts_full_state)
filtered_hosts = {(, h.nodename): h for h in hosts}
hosts_aux = [h for h in hosts_full_state
if (, h.nodename) in filtered_hosts]
weighed_hosts = self.host_manager.get_weighed_hosts(hosts_aux,
LOG.debug("Weighed %(hosts)s", {'hosts': weighed_hosts})
scheduler_host_subset_size = CONF.scheduler_host_subset_size
if scheduler_host_subset_size > len(weighed_hosts):
scheduler_host_subset_size = len(weighed_hosts)
if scheduler_host_subset_size < 1:
scheduler_host_subset_size = 1
chosen_host = random.choice(
LOG.debug("Selected host: %(host)s", {'host': chosen_host})
# Now consume the resources so the filter/weights
# will change for the next instance.
# First update the chosen host, that is from the full state list
# Now consume from the partial state list
host =
node = chosen_host.obj.nodename
state_key = (host, node)
# Now continue with the rest of the scheduling function
if spec_obj.instance_group is not None:
# hosts has to be not part of the updates when saving
return selected_hosts
def _get_all_host_states(self, context, partial=False):
"""Template method, so a subclass can implement caching."""
if partial:
return self.host_manager.get_all_host_partial_states(context)
return self.host_manager.get_all_host_states(context)
def _is_preemptible_request(self, spec_obj):
# NOTE(aloga): this is not lazy loadable
if not hasattr(spec_obj, "scheduler_hints"):
spec_obj.scheduler_hints = {}
hints = spec_obj.scheduler_hints.get("preemptible", [False])
return all([h == "True" for h in hints])
You can’t perform that action at this time.