# Copyright (c) 2010 Citrix Systems, Inc. # Copyright 2010 OpenStack Foundation # # 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. """ Management class for VM-related functions (spawn, reboot, etc). """ import base64 import functools import time import zlib import eventlet from eventlet import greenthread import netaddr from oslo_log import log as logging from oslo_serialization import jsonutils from oslo_utils import excutils from oslo_utils import importutils from oslo_utils import netutils from oslo_utils import strutils from oslo_utils import timeutils from oslo_utils import units import six from nova import block_device from nova import compute from nova.compute import power_state from nova.compute import task_states from nova.compute import vm_mode from nova.compute import vm_states import nova.conf from nova.console import type as ctype from nova import context as nova_context from nova import exception from nova.i18n import _, _LE, _LI, _LW from nova import objects from nova.objects import migrate_data as migrate_data_obj from nova.pci import manager as pci_manager from nova import utils from nova.virt import configdrive from nova.virt import driver as virt_driver from nova.virt import firewall from nova.virt.xenapi import agent as xapi_agent from nova.virt.xenapi import pool_states from nova.virt.xenapi import vm_utils from nova.virt.xenapi import volume_utils from nova.virt.xenapi import volumeops LOG = logging.getLogger(__name__) CONF = nova.conf.CONF DEFAULT_FIREWALL_DRIVER = "%s.%s" % ( firewall.__name__, firewall.IptablesFirewallDriver.__name__) RESIZE_TOTAL_STEPS = 5 DEVICE_ROOT = '0' DEVICE_RESCUE = '1' DEVICE_SWAP = '2' DEVICE_CONFIGDRIVE = '3' # Note(johngarbutt) HVM guests only support four devices # until the PV tools activate, when others before available # As such, ephemeral disk only available once PV tools load # Note(johngarbutt) When very large ephemeral storage is required, # multiple disks may be added. In this case the device id below # is the used for the first disk. The second disk will be given # next device id, i.e. 5, and so on, until enough space is added. DEVICE_EPHEMERAL = '4' # Note(johngarbutt) Currently don't support ISO boot during rescue # and we must have the ISO visible before the PV drivers start DEVICE_CD = '1' def make_step_decorator(context, instance, update_instance_progress, total_offset=0): """Factory to create a decorator that records instance progress as a series of discrete steps. Each time the decorator is invoked we bump the total-step-count, so after:: @step def step1(): ... @step def step2(): ... we have a total-step-count of 2. Each time the step-function (not the step-decorator!) is invoked, we bump the current-step-count by 1, so after:: step1() the current-step-count would be 1 giving a progress of ``1 / 2 * 100`` or 50%. """ step_info = dict(total=total_offset, current=0) def bump_progress(): step_info['current'] += 1 update_instance_progress(context, instance, step_info['current'], step_info['total']) def step_decorator(f): step_info['total'] += 1 @functools.wraps(f) def inner(*args, **kwargs): rv = f(*args, **kwargs) bump_progress() return rv return inner return step_decorator class VMOps(object): """Management class for VM-related tasks.""" def __init__(self, session, virtapi): self.compute_api = compute.API() self._session = session self._virtapi = virtapi self._volumeops = volumeops.VolumeOps(self._session) self.firewall_driver = firewall.load_driver( DEFAULT_FIREWALL_DRIVER, xenapi_session=self._session) vif_impl = importutils.import_class(CONF.xenserver.vif_driver) self.vif_driver = vif_impl(xenapi_session=self._session) self.default_root_dev = '/dev/sda' LOG.debug("Importing image upload handler: %s", CONF.xenserver.image_upload_handler) self.image_upload_handler = importutils.import_object( CONF.xenserver.image_upload_handler) def agent_enabled(self, instance): if CONF.xenserver.disable_agent: return False return xapi_agent.should_use_agent(instance) def _get_agent(self, instance, vm_ref): if self.agent_enabled(instance): return xapi_agent.XenAPIBasedAgent(self._session, self._virtapi, instance, vm_ref) raise exception.NovaException(_("Error: Agent is disabled")) def instance_exists(self, name_label): return vm_utils.lookup(self._session, name_label) is not None def list_instances(self): """List VM instances.""" # TODO(justinsb): Should we just always use the details method? # Seems to be the same number of API calls.. name_labels = [] for vm_ref, vm_rec in vm_utils.list_vms(self._session): name_labels.append(vm_rec["name_label"]) return name_labels def list_instance_uuids(self): """Get the list of nova instance uuids for VMs found on the hypervisor. """ nova_uuids = [] for vm_ref, vm_rec in vm_utils.list_vms(self._session): other_config = vm_rec['other_config'] nova_uuid = other_config.get('nova_uuid') if nova_uuid: nova_uuids.append(nova_uuid) return nova_uuids def confirm_migration(self, migration, instance, network_info): self._destroy_orig_vm(instance, network_info) def _destroy_orig_vm(self, instance, network_info): name_label = self._get_orig_vm_name_label(instance) vm_ref = vm_utils.lookup(self._session, name_label) return self._destroy(instance, vm_ref, network_info=network_info) def _attach_mapped_block_devices(self, instance, block_device_info): # We are attaching these volumes before start (no hotplugging) # because some guests (windows) don't load PV drivers quickly block_device_mapping = virt_driver.block_device_info_get_mapping( block_device_info) for vol in block_device_mapping: if vol['mount_device'] == instance['root_device_name']: # NOTE(alaski): The root device should be attached already continue connection_info = vol['connection_info'] mount_device = vol['mount_device'].rpartition("/")[2] self._volumeops.attach_volume(connection_info, instance['name'], mount_device, hotplug=False) def finish_revert_migration(self, context, instance, block_device_info=None, power_on=True): self._restore_orig_vm_and_cleanup_orphan(instance, block_device_info, power_on) def _restore_orig_vm_and_cleanup_orphan(self, instance, block_device_info=None, power_on=True): # NOTE(sirp): the original vm was suffixed with '-orig'; find it using # the old suffix, remove the suffix, then power it back on. name_label = self._get_orig_vm_name_label(instance) vm_ref = vm_utils.lookup(self._session, name_label) # NOTE(danms): if we're reverting migration in the failure case, # make sure we don't have a conflicting vm still running here, # as might be the case in a failed migrate-to-same-host situation new_ref = vm_utils.lookup(self._session, instance['name']) if vm_ref is not None: if new_ref is not None: self._destroy(instance, new_ref) # Remove the '-orig' suffix (which was added in case the # resized VM ends up on the source host, common during # testing) name_label = instance['name'] vm_utils.set_vm_name_label(self._session, vm_ref, name_label) self._attach_mapped_block_devices(instance, block_device_info) elif new_ref is not None: # We crashed before the -orig backup was made vm_ref = new_ref if power_on and vm_utils.is_vm_shutdown(self._session, vm_ref): self._start(instance, vm_ref) def finish_migration(self, context, migration, instance, disk_info, network_info, image_meta, resize_instance, block_device_info=None, power_on=True): def null_step_decorator(f): return f def create_disks_step(undo_mgr, disk_image_type, image_meta, name_label): import_root = True root_vol_vdi = None if block_device_info: LOG.debug("Block device information present: %s", block_device_info, instance=instance) # NOTE(alaski): Follows the basic procedure of # vm_utils.get_vdis_for_instance() used by spawn() for bdm in block_device_info['block_device_mapping']: if bdm['mount_device'] == instance['root_device_name']: connection_info = bdm['connection_info'] _sr, root_vol_vdi = self._volumeops.connect_volume( connection_info) import_root = False break # TODO(johngarbutt) clean up if this is not run vdis = vm_utils.import_all_migrated_disks(self._session, instance, import_root=import_root) if root_vol_vdi: vol_vdi_ref = self._session.call_xenapi('VDI.get_by_uuid', root_vol_vdi) vdis['root'] = dict(uuid=root_vol_vdi, file=None, ref=vol_vdi_ref, osvol=True) def undo_create_disks(): eph_vdis = vdis['ephemerals'] root_vdi = vdis['root'] vdi_refs = [vdi['ref'] for vdi in eph_vdis.values()] if not root_vdi.get('osvol', False): vdi_refs.append(root_vdi['ref']) else: self._volumeops.safe_cleanup_from_vdis(root_vdi['ref']) vm_utils.safe_destroy_vdis(self._session, vdi_refs) undo_mgr.undo_with(undo_create_disks) return vdis def completed_callback(): self._update_instance_progress(context, instance, step=5, total_steps=RESIZE_TOTAL_STEPS) self._spawn(context, instance, image_meta, null_step_decorator, create_disks_step, first_boot=False, injected_files=None, admin_password=None, network_info=network_info, block_device_info=block_device_info, name_label=None, rescue=False, power_on=power_on, resize=resize_instance, completed_callback=completed_callback) def _start(self, instance, vm_ref=None, bad_volumes_callback=None, start_pause=False): """Power on a VM instance.""" vm_ref = vm_ref or self._get_vm_opaque_ref(instance) LOG.debug("Starting instance", instance=instance) # Attached volumes that have become non-responsive will prevent a VM # from starting, so scan for these before attempting to start # # In order to make sure this detach is consistent (virt, BDM, cinder), # we only detach in the virt-layer if a callback is provided. if bad_volumes_callback: bad_devices = self._volumeops.find_bad_volumes(vm_ref) for device_name in bad_devices: self._volumeops.detach_volume( None, instance['name'], device_name) self._session.call_xenapi('VM.start_on', vm_ref, self._session.host_ref, start_pause, False) # Allow higher-layers a chance to detach bad-volumes as well (in order # to cleanup BDM entries and detach in Cinder) if bad_volumes_callback and bad_devices: bad_volumes_callback(bad_devices) # Do some operations which have to be done after start: # e.g. The vif's interim bridge won't be created until VM starts. # So the operations on the interim bridge have be done after # start. self._post_start_actions(instance) def _post_start_actions(self, instance): vm_ref = vm_utils.lookup(self._session, instance['name']) vif_refs = self._session.call_xenapi("VM.get_VIFs", vm_ref) for vif_ref in vif_refs: self.vif_driver.post_start_actions(instance, vif_ref) def _get_vdis_for_instance(self, context, instance, name_label, image_meta, image_type, block_device_info): """Create or connect to all virtual disks for this instance.""" vdis = self._connect_cinder_volumes(instance, block_device_info) # If we didn't get a root VDI from volumes, # then use the Glance image as the root device if 'root' not in vdis: create_image_vdis = vm_utils.create_image(context, self._session, instance, name_label, image_meta.id, image_type) vdis.update(create_image_vdis) # Fetch VDI refs now so we don't have to fetch the ref multiple times for vdi in six.itervalues(vdis): vdi['ref'] = self._session.call_xenapi('VDI.get_by_uuid', vdi['uuid']) return vdis def _connect_cinder_volumes(self, instance, block_device_info): """Attach all the cinder volumes described in block_device_info.""" vdis = {} if block_device_info: msg = "block device info: %s" % block_device_info # NOTE(mriedem): block_device_info can contain an auth_password # so we have to scrub the message before logging it. LOG.debug(strutils.mask_password(msg), instance=instance) root_device_name = block_device_info['root_device_name'] for bdm in block_device_info['block_device_mapping']: if (block_device.strip_prefix(bdm['mount_device']) == block_device.strip_prefix(root_device_name)): # If we're a root-device, record that fact so we don't # download a root image via Glance type_ = 'root' else: # Otherwise, use mount_device as `type_` so that we have # easy access to it in _attach_disks to create the VBD type_ = bdm['mount_device'] conn_info = bdm['connection_info'] _sr, vdi_uuid = self._volumeops.connect_volume(conn_info) if vdi_uuid: vdis[type_] = dict(uuid=vdi_uuid, file=None, osvol=True) return vdis def _update_last_dom_id(self, vm_ref): other_config = self._session.VM.get_other_config(vm_ref) other_config['last_dom_id'] = self._session.VM.get_domid(vm_ref) self._session.VM.set_other_config(vm_ref, other_config) def spawn(self, context, instance, image_meta, injected_files, admin_password, network_info=None, block_device_info=None, name_label=None, rescue=False): if block_device_info: LOG.debug("Block device information present: %s", block_device_info, instance=instance) if block_device_info and not block_device_info['root_device_name']: block_device_info['root_device_name'] = self.default_root_dev step = make_step_decorator(context, instance, self._update_instance_progress) @step def create_disks_step(undo_mgr, disk_image_type, image_meta, name_label): vdis = self._get_vdis_for_instance(context, instance, name_label, image_meta, disk_image_type, block_device_info) def undo_create_disks(): vdi_refs = [vdi['ref'] for vdi in vdis.values() if not vdi.get('osvol')] vm_utils.safe_destroy_vdis(self._session, vdi_refs) vol_vdi_refs = [vdi['ref'] for vdi in vdis.values() if vdi.get('osvol')] self._volumeops.safe_cleanup_from_vdis(vol_vdi_refs) undo_mgr.undo_with(undo_create_disks) return vdis self._spawn(context, instance, image_meta, step, create_disks_step, True, injected_files, admin_password, network_info, block_device_info, name_label, rescue) def _spawn(self, context, instance, image_meta, step, create_disks_step, first_boot, injected_files=None, admin_password=None, network_info=None, block_device_info=None, name_label=None, rescue=False, power_on=True, resize=True, completed_callback=None): if name_label is None: name_label = instance['name'] self._ensure_instance_name_unique(name_label) self._ensure_enough_free_mem(instance) def attach_disks(undo_mgr, vm_ref, vdis, disk_image_type): if image_meta.properties.get('hw_ipxe_boot', False): if 'iso' in vdis: vm_utils.handle_ipxe_iso( self._session, instance, vdis['iso'], network_info) else: LOG.warning(_LW('ipxe_boot is True but no ISO image ' 'found'), instance=instance) if resize: self._resize_up_vdis(instance, vdis) self._attach_disks(instance, image_meta, vm_ref, name_label, vdis, disk_image_type, network_info, rescue, admin_password, injected_files) if not first_boot: self._attach_mapped_block_devices(instance, block_device_info) def attach_pci_devices(undo_mgr, vm_ref): dev_to_passthrough = "" devices = pci_manager.get_instance_pci_devs(instance) for d in devices: pci_address = d["address"] if pci_address.count(":") == 1: pci_address = "0000:" + pci_address dev_to_passthrough += ",0/" + pci_address # Remove the first comma if string is not empty. # Note(guillaume-thouvenin): If dev_to_passthrough is empty, we # don't need to update other_config. if dev_to_passthrough: vm_utils.set_other_config_pci(self._session, vm_ref, dev_to_passthrough[1:]) @step def determine_disk_image_type_step(undo_mgr): return vm_utils.determine_disk_image_type(image_meta) @step def create_kernel_ramdisk_step(undo_mgr): kernel_file, ramdisk_file = vm_utils.create_kernel_and_ramdisk( context, self._session, instance, name_label) def undo_create_kernel_ramdisk(): vm_utils.destroy_kernel_ramdisk(self._session, instance, kernel_file, ramdisk_file) undo_mgr.undo_with(undo_create_kernel_ramdisk) return kernel_file, ramdisk_file @step def create_vm_record_step(undo_mgr, disk_image_type, kernel_file, ramdisk_file): vm_ref = self._create_vm_record(context, instance, name_label, disk_image_type, kernel_file, ramdisk_file, image_meta, rescue) def undo_create_vm(): self._destroy(instance, vm_ref, network_info=network_info) undo_mgr.undo_with(undo_create_vm) return vm_ref @step def attach_devices_step(undo_mgr, vm_ref, vdis, disk_image_type): attach_disks(undo_mgr, vm_ref, vdis, disk_image_type) attach_pci_devices(undo_mgr, vm_ref) if rescue: # NOTE(johannes): Attach disks from original VM to rescue VM now, # before booting the VM, since we can't hotplug block devices # on non-PV guests @step def attach_orig_disks_step(undo_mgr, vm_ref): vbd_refs = self._attach_orig_disks(instance, vm_ref) def undo_attach_orig_disks(): # Destroy the VBDs in preparation to re-attach the VDIs # to its original VM. (does not delete VDI) for vbd_ref in vbd_refs: vm_utils.destroy_vbd(self._session, vbd_ref) undo_mgr.undo_with(undo_attach_orig_disks) @step def inject_instance_data_step(undo_mgr, vm_ref, vdis): self._inject_instance_metadata(instance, vm_ref) self._inject_auto_disk_config(instance, vm_ref) # NOTE: We add the hostname here so windows PV tools # can pick it up during booting if first_boot: self._inject_hostname(instance, vm_ref, rescue) self._file_inject_vm_settings(instance, vm_ref, vdis, network_info) self.inject_network_info(instance, network_info, vm_ref) @step def setup_network_step(undo_mgr, vm_ref): self._create_vifs(instance, vm_ref, network_info) self._prepare_instance_filter(instance, network_info) @step def start_paused_step(undo_mgr, vm_ref): if power_on: self._start(instance, vm_ref, start_pause=True) @step def boot_and_configure_instance_step(undo_mgr, vm_ref): self._unpause_and_wait(vm_ref, instance, power_on) if first_boot: self._configure_new_instance_with_agent(instance, vm_ref, injected_files, admin_password) self._remove_hostname(instance, vm_ref) @step def apply_security_group_filters_step(undo_mgr): self.firewall_driver.apply_instance_filter(instance, network_info) undo_mgr = utils.UndoManager() try: # NOTE(sirp): The create_disks() step will potentially take a # *very* long time to complete since it has to fetch the image # over the network and images can be several gigs in size. To # avoid progress remaining at 0% for too long, make sure the # first step is something that completes rather quickly. disk_image_type = determine_disk_image_type_step(undo_mgr) vdis = create_disks_step(undo_mgr, disk_image_type, image_meta, name_label) kernel_file, ramdisk_file = create_kernel_ramdisk_step(undo_mgr) vm_ref = create_vm_record_step(undo_mgr, disk_image_type, kernel_file, ramdisk_file) attach_devices_step(undo_mgr, vm_ref, vdis, disk_image_type) inject_instance_data_step(undo_mgr, vm_ref, vdis) # if use neutron, prepare waiting event from neutron # first_boot is True in new booted instance # first_boot is False in migration and we don't waiting # for neutron event regardless of whether or not it is # migrated to another host, if unplug VIFs locally, the # port status may not changed in neutron side and we # cannot get the vif plug event from neturon timeout = CONF.vif_plugging_timeout events = self._get_neutron_events(network_info, power_on, first_boot) try: with self._virtapi.wait_for_instance_event( instance, events, deadline=timeout, error_callback=self._neutron_failed_callback): LOG.debug("wait for instance event:%s", events) setup_network_step(undo_mgr, vm_ref) if rescue: attach_orig_disks_step(undo_mgr, vm_ref) start_paused_step(undo_mgr, vm_ref) except eventlet.timeout.Timeout: self._handle_neutron_event_timeout(instance, undo_mgr) apply_security_group_filters_step(undo_mgr) boot_and_configure_instance_step(undo_mgr, vm_ref) if completed_callback: completed_callback() except Exception: msg = _("Failed to spawn, rolling back") undo_mgr.rollback_and_reraise(msg=msg, instance=instance) def _handle_neutron_event_timeout(self, instance, undo_mgr): # We didn't get callback from Neutron within given time LOG.warning(_LW('Timeout waiting for vif plugging callback'), instance=instance) if CONF.vif_plugging_is_fatal: raise exception.VirtualInterfaceCreateException() def _unpause_and_wait(self, vm_ref, instance, power_on): if power_on: LOG.debug("Update instance when power on", instance=instance) self._session.VM.unpause(vm_ref) self._wait_for_instance_to_start(instance, vm_ref) self._update_last_dom_id(vm_ref) def _neutron_failed_callback(self, event_name, instance): LOG.warning(_LW('Neutron Reported failure on event %(event)s'), {'event': event_name}, instance=instance) if CONF.vif_plugging_is_fatal: raise exception.VirtualInterfaceCreateException() def _get_neutron_events(self, network_info, power_on, first_boot): # Only get network-vif-plugged events with VIF's status is not active. # With VIF whose status is active, neutron may not notify such event. timeout = CONF.vif_plugging_timeout if (utils.is_neutron() and power_on and timeout and first_boot): return [('network-vif-plugged', vif['id']) for vif in network_info if vif.get('active', True) is False] else: return [] def _attach_orig_disks(self, instance, vm_ref): orig_vm_ref = vm_utils.lookup(self._session, instance['name']) orig_vdi_refs = self._find_vdi_refs(orig_vm_ref, exclude_volumes=True) # Attach original root disk root_vdi_ref = orig_vdi_refs.get(DEVICE_ROOT) if not root_vdi_ref: raise exception.NotFound(_("Unable to find root VBD/VDI for VM")) vbd_ref = vm_utils.create_vbd(self._session, vm_ref, root_vdi_ref, DEVICE_RESCUE, bootable=False) vbd_refs = [vbd_ref] # Attach original swap disk swap_vdi_ref = orig_vdi_refs.get(DEVICE_SWAP) if swap_vdi_ref: vbd_ref = vm_utils.create_vbd(self._session, vm_ref, swap_vdi_ref, DEVICE_SWAP, bootable=False) vbd_refs.append(vbd_ref) # Attach original ephemeral disks for userdevice, vdi_ref in six.iteritems(orig_vdi_refs): if userdevice >= DEVICE_EPHEMERAL: vbd_ref = vm_utils.create_vbd(self._session, vm_ref, vdi_ref, userdevice, bootable=False) vbd_refs.append(vbd_ref) return vbd_refs def _file_inject_vm_settings(self, instance, vm_ref, vdis, network_info): if CONF.flat_injected: vm_utils.preconfigure_instance(self._session, instance, vdis['root']['ref'], network_info) def _ensure_instance_name_unique(self, name_label): vm_ref = vm_utils.lookup(self._session, name_label) if vm_ref is not None: raise exception.InstanceExists(name=name_label) def _ensure_enough_free_mem(self, instance): if not vm_utils.is_enough_free_mem(self._session, instance): raise exception.InsufficientFreeMemory(uuid=instance['uuid']) def _create_vm_record(self, context, instance, name_label, disk_image_type, kernel_file, ramdisk_file, image_meta, rescue=False): """Create the VM record in Xen, making sure that we do not create a duplicate name-label. Also do a rough sanity check on memory to try to short-circuit a potential failure later. (The memory check only accounts for running VMs, so it can miss other builds that are in progress.) """ mode = vm_utils.determine_vm_mode(instance, disk_image_type) # NOTE(tpownall): If rescue mode then we should try to pull the vm_mode # value from the image properties to ensure the vm is built properly. if rescue: rescue_vm_mode = image_meta.properties.get('hw_vm_mode', None) if rescue_vm_mode is None: LOG.debug("vm_mode not found in rescue image properties." "Setting vm_mode to %s", mode, instance=instance) else: mode = vm_mode.canonicalize(rescue_vm_mode) if instance.vm_mode != mode: # Update database with normalized (or determined) value instance.vm_mode = mode instance.save() device_id = vm_utils.get_vm_device_id(self._session, image_meta) use_pv_kernel = (mode == vm_mode.XEN) LOG.debug("Using PV kernel: %s", use_pv_kernel, instance=instance) vm_ref = vm_utils.create_vm(self._session, instance, name_label, kernel_file, ramdisk_file, use_pv_kernel, device_id) return vm_ref def _attach_disks(self, instance, image_meta, vm_ref, name_label, vdis, disk_image_type, network_info, rescue=False, admin_password=None, files=None): flavor = instance.get_flavor() # Attach (required) root disk if disk_image_type == vm_utils.ImageType.DISK_ISO: # DISK_ISO needs two VBDs: the ISO disk and a blank RW disk root_disk_size = flavor.root_gb if root_disk_size > 0: vm_utils.generate_iso_blank_root_disk(self._session, instance, vm_ref, DEVICE_ROOT, name_label, root_disk_size) cd_vdi = vdis.pop('iso') vm_utils.attach_cd(self._session, vm_ref, cd_vdi['ref'], DEVICE_CD) else: root_vdi = vdis['root'] auto_disk_config = instance['auto_disk_config'] # NOTE(tpownall): If rescue mode we need to ensure that we're # pulling the auto_disk_config value from the image properties so # that we can pull it from the rescue_image_ref. if rescue: if not image_meta.properties.obj_attr_is_set( "hw_auto_disk_config"): LOG.debug("'hw_auto_disk_config' value not found in" "rescue image_properties. Setting value to %s", auto_disk_config, instance=instance) else: auto_disk_config = strutils.bool_from_string( image_meta.properties.hw_auto_disk_config) if auto_disk_config: LOG.debug("Auto configuring disk, attempting to " "resize root disk...", instance=instance) vm_utils.try_auto_configure_disk(self._session, root_vdi['ref'], flavor.root_gb) vm_utils.create_vbd(self._session, vm_ref, root_vdi['ref'], DEVICE_ROOT, bootable=True, osvol=root_vdi.get('osvol')) # Attach (optional) additional block-devices for type_, vdi_info in vdis.items(): # Additional block-devices for boot use their device-name as the # type. if not type_.startswith('/dev'): continue # Convert device name to user device number, e.g. /dev/xvdb -> 1 userdevice = ord(block_device.strip_prefix(type_)) - ord('a') vm_utils.create_vbd(self._session, vm_ref, vdi_info['ref'], userdevice, bootable=False, osvol=vdi_info.get('osvol')) # For rescue, swap and ephemeral disks get attached in # _attach_orig_disks # Attach (optional) swap disk swap_mb = flavor.swap if not rescue and swap_mb: vm_utils.generate_swap(self._session, instance, vm_ref, DEVICE_SWAP, name_label, swap_mb) ephemeral_gb = flavor.ephemeral_gb if not rescue and ephemeral_gb: ephemeral_vdis = vdis.get('ephemerals') if ephemeral_vdis: # attach existing (migrated) ephemeral disks for userdevice, ephemeral_vdi in six.iteritems(ephemeral_vdis): vm_utils.create_vbd(self._session, vm_ref, ephemeral_vdi['ref'], userdevice, bootable=False) else: # create specified ephemeral disks vm_utils.generate_ephemeral(self._session, instance, vm_ref, DEVICE_EPHEMERAL, name_label, ephemeral_gb) # Attach (optional) configdrive v2 disk if configdrive.required_by(instance): vm_utils.generate_configdrive(self._session, instance, vm_ref, DEVICE_CONFIGDRIVE, network_info, admin_password=admin_password, files=files) def _wait_for_instance_to_start(self, instance, vm_ref): LOG.debug('Waiting for instance state to become running', instance=instance) expiration = time.time() + CONF.xenserver.running_timeout while time.time() < expiration: state = vm_utils.get_power_state(self._session, vm_ref) if state == power_state.RUNNING: break greenthread.sleep(0.5) def _configure_new_instance_with_agent(self, instance, vm_ref, injected_files, admin_password): if not self.agent_enabled(instance): LOG.debug("Skip agent setup, not enabled.", instance=instance) return agent = self._get_agent(instance, vm_ref) version = agent.get_version() if not version: LOG.debug("Skip agent setup, unable to contact agent.", instance=instance) return LOG.debug('Detected agent version: %s', version, instance=instance) # NOTE(johngarbutt) the agent object allows all of # the following steps to silently fail agent.inject_ssh_key() if injected_files: agent.inject_files(injected_files) if admin_password: agent.set_admin_password(admin_password) agent.resetnetwork() agent.update_if_needed(version) def _prepare_instance_filter(self, instance, network_info): try: self.firewall_driver.setup_basic_filtering( instance, network_info) except NotImplementedError: # NOTE(salvatore-orlando): setup_basic_filtering might be # empty or not implemented at all, as basic filter could # be implemented with VIF rules created by xapi plugin pass self.firewall_driver.prepare_instance_filter(instance, network_info) def _get_vm_opaque_ref(self, instance, check_rescue=False): """Get xapi OpaqueRef from a db record. :param check_rescue: if True will return the 'name'-rescue vm if it exists, instead of just 'name' """ vm_ref = vm_utils.lookup(self._session, instance['name'], check_rescue) if vm_ref is None: raise exception.InstanceNotFound(instance_id=instance['name']) return vm_ref def _acquire_bootlock(self, vm): """Prevent an instance from booting.""" self._session.call_xenapi( "VM.set_blocked_operations", vm, {"start": ""}) def _release_bootlock(self, vm): """Allow an instance to boot.""" self._session.call_xenapi( "VM.remove_from_blocked_operations", vm, "start") def snapshot(self, context, instance, image_id, update_task_state): """Create snapshot from a running VM instance. :param context: request context :param instance: instance to be snapshotted :param image_id: id of image to upload to Steps involved in a XenServer snapshot: 1. XAPI-Snapshot: Snapshotting the instance using XenAPI. This creates: Snapshot (Template) VM, Snapshot VBD, Snapshot VDI, Snapshot VHD 2. Wait-for-coalesce: The Snapshot VDI and Instance VDI both point to a 'base-copy' VDI. The base_copy is immutable and may be chained with other base_copies. If chained, the base_copies coalesce together, so, we must wait for this coalescing to occur to get a stable representation of the data on disk. 3. Push-to-data-store: Once coalesced, we call 'image_upload_handler' to upload the images. """ vm_ref = self._get_vm_opaque_ref(instance) label = "%s-snapshot" % instance['name'] start_time = timeutils.utcnow() with vm_utils.snapshot_attached_here( self._session, instance, vm_ref, label, post_snapshot_callback=update_task_state) as vdi_uuids: update_task_state(task_state=task_states.IMAGE_UPLOADING, expected_state=task_states.IMAGE_PENDING_UPLOAD) self.image_upload_handler.upload_image(context, self._session, instance, image_id, vdi_uuids, ) duration = timeutils.delta_seconds(start_time, timeutils.utcnow()) LOG.debug("Finished snapshot and upload for VM, duration: " "%(duration).2f secs for image %(image_id)s", {'image_id': image_id, 'duration': duration}, instance=instance) def post_interrupted_snapshot_cleanup(self, context, instance): """Cleans up any resources left after a failed snapshot.""" vm_ref = self._get_vm_opaque_ref(instance) vm_utils.remove_old_snapshots(self._session, instance, vm_ref) def _get_orig_vm_name_label(self, instance): return instance['name'] + '-orig' def _update_instance_progress(self, context, instance, step, total_steps): """Update instance progress percent to reflect current step number """ # FIXME(sirp): for now we're taking a KISS approach to instance # progress: # Divide the action's workflow into discrete steps and "bump" the # instance's progress field as each step is completed. # # For a first cut this should be fine, however, for large VM images, # the get_vdis_for_instance step begins to dominate the equation. A # better approximation would use the percentage of the VM image that # has been streamed to the destination host. progress = round(float(step) / total_steps * 100) LOG.debug("Updating progress to %d", progress, instance=instance) instance.progress = progress instance.save() def _resize_ensure_vm_is_shutdown(self, instance, vm_ref): if vm_utils.is_vm_shutdown(self._session, vm_ref): LOG.debug("VM was already shutdown.", instance=instance) return if not vm_utils.clean_shutdown_vm(self._session, instance, vm_ref): LOG.debug("Clean shutdown did not complete successfully, " "trying hard shutdown.", instance=instance) if not vm_utils.hard_shutdown_vm(self._session, instance, vm_ref): raise exception.ResizeError( reason=_("Unable to terminate instance.")) def _migrate_disk_resizing_down(self, context, instance, dest, flavor, vm_ref, sr_path): step = make_step_decorator(context, instance, self._update_instance_progress, total_offset=1) @step def fake_step_to_match_resizing_up(): pass @step def rename_and_power_off_vm(undo_mgr): self._resize_ensure_vm_is_shutdown(instance, vm_ref) self._apply_orig_vm_name_label(instance, vm_ref) def restore_orig_vm(): # Do not need to restore block devices, not yet been removed self._restore_orig_vm_and_cleanup_orphan(instance) undo_mgr.undo_with(restore_orig_vm) @step def create_copy_vdi_and_resize(undo_mgr, old_vdi_ref): new_vdi_ref, new_vdi_uuid = vm_utils.resize_disk(self._session, instance, old_vdi_ref, flavor) def cleanup_vdi_copy(): vm_utils.destroy_vdi(self._session, new_vdi_ref) undo_mgr.undo_with(cleanup_vdi_copy) return new_vdi_ref, new_vdi_uuid @step def transfer_vhd_to_dest(new_vdi_ref, new_vdi_uuid): vm_utils.migrate_vhd(self._session, instance, new_vdi_uuid, dest, sr_path, 0) # Clean up VDI now that it's been copied vm_utils.destroy_vdi(self._session, new_vdi_ref) undo_mgr = utils.UndoManager() try: fake_step_to_match_resizing_up() rename_and_power_off_vm(undo_mgr) old_vdi_ref, _ignore = vm_utils.get_vdi_for_vm_safely( self._session, vm_ref) new_vdi_ref, new_vdi_uuid = create_copy_vdi_and_resize( undo_mgr, old_vdi_ref) transfer_vhd_to_dest(new_vdi_ref, new_vdi_uuid) except Exception as error: LOG.exception(_LE("_migrate_disk_resizing_down failed. " "Restoring orig vm"), instance=instance) undo_mgr._rollback() raise exception.InstanceFaultRollback(error) def _migrate_disk_resizing_up(self, context, instance, dest, vm_ref, sr_path): step = make_step_decorator(context, instance, self._update_instance_progress, total_offset=1) """ NOTE(johngarbutt) Understanding how resize up works. For resize up, we attempt to minimize the amount of downtime for users by copying snapshots of their disks, while their VM is still running. It is worth noting, that migrating the snapshot, means migrating the whole VHD chain up to, but not including, the leaf VHD the VM is still writing to. Once the snapshots have been migrated, we power down the VM and migrate all the disk changes since the snapshots were taken. In addition, the snapshots are taken at the latest possible point, to help minimize the time it takes to migrate the disk changes after the VM has been turned off. Before starting to migrate any of the disks, we rename the VM, to -orig, in case we attempt to migrate the VM back onto this host, and so once we have completed the migration of the disk, confirm/rollback migrate can work in the usual way. If there is a failure at any point, we need to rollback to the position we were in before starting to migrate. In particular, we need to delete and snapshot VDIs that may have been created, and restore the VM back to its original name. """ @step def fake_step_to_show_snapshot_complete(): pass @step def transfer_immutable_vhds(root_vdi_uuids): immutable_root_vdi_uuids = root_vdi_uuids[1:] for vhd_num, vdi_uuid in enumerate(immutable_root_vdi_uuids, start=1): vm_utils.migrate_vhd(self._session, instance, vdi_uuid, dest, sr_path, vhd_num) LOG.debug("Migrated root base vhds", instance=instance) def _process_ephemeral_chain_recursive(ephemeral_chains, active_vdi_uuids): # This method is called several times, recursively. # The first phase snapshots the ephemeral disks, and # migrates the read only VHD files. # The final call into this method calls # power_down_and_transfer_leaf_vhds # to turn off the VM and copy the rest of the VHDs. number_of_chains = len(ephemeral_chains) if number_of_chains == 0: # If we get here, we have snapshotted and migrated # all the ephemeral disks, so its time to power down # and complete the migration of the diffs since the snapshot LOG.debug("Migrated all base vhds.", instance=instance) return power_down_and_transfer_leaf_vhds( active_root_vdi_uuid, active_vdi_uuids) remaining_chains = [] if number_of_chains > 1: remaining_chains = ephemeral_chains[1:] ephemeral_disk_index = len(active_vdi_uuids) userdevice = int(DEVICE_EPHEMERAL) + ephemeral_disk_index # Here we take a snapshot of the ephemeral disk, # and migrate all VHDs in the chain that are not being written to # Once that is completed, we call back into this method to either: # - migrate any remaining ephemeral disks # - or, if all disks are migrated, we power down and complete # the migration but copying the diffs since all the snapshots # were taken with vm_utils.snapshot_attached_here(self._session, instance, vm_ref, label, str(userdevice)) as chain_vdi_uuids: # remember active vdi, we will migrate these later vdi_ref, vm_vdi_rec = vm_utils.get_vdi_for_vm_safely( self._session, vm_ref, str(userdevice)) active_uuid = vm_vdi_rec['uuid'] active_vdi_uuids.append(active_uuid) # migrate inactive vhds inactive_vdi_uuids = chain_vdi_uuids[1:] ephemeral_disk_number = ephemeral_disk_index + 1 for seq_num, vdi_uuid in enumerate(inactive_vdi_uuids, start=1): vm_utils.migrate_vhd(self._session, instance, vdi_uuid, dest, sr_path, seq_num, ephemeral_disk_number) LOG.debug("Read-only migrated for disk: %s", userdevice, instance=instance) # This is recursive to simplify the taking and cleaning up # of all the ephemeral disk snapshots return _process_ephemeral_chain_recursive(remaining_chains, active_vdi_uuids) @step def transfer_ephemeral_disks_then_all_leaf_vdis(): ephemeral_chains = vm_utils.get_all_vdi_uuids_for_vm( self._session, vm_ref, min_userdevice=int(DEVICE_EPHEMERAL)) if ephemeral_chains: ephemeral_chains = list(ephemeral_chains) else: ephemeral_chains = [] _process_ephemeral_chain_recursive(ephemeral_chains, []) @step def power_down_and_transfer_leaf_vhds(root_vdi_uuid, ephemeral_vdi_uuids=None): self._resize_ensure_vm_is_shutdown(instance, vm_ref) if root_vdi_uuid is not None: vm_utils.migrate_vhd(self._session, instance, root_vdi_uuid, dest, sr_path, 0) if ephemeral_vdi_uuids: for ephemeral_disk_number, ephemeral_vdi_uuid in enumerate( ephemeral_vdi_uuids, start=1): vm_utils.migrate_vhd(self._session, instance, ephemeral_vdi_uuid, dest, sr_path, 0, ephemeral_disk_number) self._apply_orig_vm_name_label(instance, vm_ref) try: label = "%s-snapshot" % instance['name'] if volume_utils.is_booted_from_volume(self._session, vm_ref): LOG.debug('Not snapshotting root disk since it is a volume', instance=instance) # NOTE(alaski): This is done twice to match the number of # defined steps. fake_step_to_show_snapshot_complete() fake_step_to_show_snapshot_complete() # NOTE(alaski): This is set to None to avoid transferring the # VHD in power_down_and_transfer_leaf_vhds. active_root_vdi_uuid = None # snapshot and transfer all ephemeral disks # then power down and transfer any diffs since # the snapshots were taken transfer_ephemeral_disks_then_all_leaf_vdis() return with vm_utils.snapshot_attached_here( self._session, instance, vm_ref, label) as root_vdi_uuids: # NOTE(johngarbutt) snapshot attached here will delete # the snapshot if an error occurs fake_step_to_show_snapshot_complete() # transfer all the non-active VHDs in the root disk chain transfer_immutable_vhds(root_vdi_uuids) vdi_ref, vm_vdi_rec = vm_utils.get_vdi_for_vm_safely( self._session, vm_ref) active_root_vdi_uuid = vm_vdi_rec['uuid'] # snapshot and transfer all ephemeral disks # then power down and transfer any diffs since # the snapshots were taken transfer_ephemeral_disks_then_all_leaf_vdis() except Exception as error: LOG.exception(_LE("_migrate_disk_resizing_up failed. " "Restoring orig vm due_to: %s."), error, instance=instance) try: self._restore_orig_vm_and_cleanup_orphan(instance) # TODO(johngarbutt) should also cleanup VHDs at destination except Exception as rollback_error: LOG.warning(_LW("_migrate_disk_resizing_up failed to " "rollback: %s"), rollback_error, instance=instance) raise exception.InstanceFaultRollback(error) def _apply_orig_vm_name_label(self, instance, vm_ref): # NOTE(sirp): in case we're resizing to the same host (for dev # purposes), apply a suffix to name-label so the two VM records # extant until a confirm_resize don't collide. name_label = self._get_orig_vm_name_label(instance) vm_utils.set_vm_name_label(self._session, vm_ref, name_label) def _ensure_not_resize_down_ephemeral(self, instance, flavor): old_gb = instance["ephemeral_gb"] new_gb = flavor.ephemeral_gb if old_gb > new_gb: reason = _("Can't resize down ephemeral disks.") raise exception.ResizeError(reason) def migrate_disk_and_power_off(self, context, instance, dest, flavor, block_device_info): """Copies a VHD from one host machine to another, possibly resizing filesystem beforehand. :param instance: the instance that owns the VHD in question. :param dest: the destination host machine. :param flavor: flavor to resize to """ self._ensure_not_resize_down_ephemeral(instance, flavor) # 0. Zero out the progress to begin self._update_instance_progress(context, instance, step=0, total_steps=RESIZE_TOTAL_STEPS) old_gb = instance['root_gb'] new_gb = flavor.root_gb resize_down = old_gb > new_gb if new_gb == 0 and old_gb != 0: reason = _("Can't resize a disk to 0 GB.") raise exception.ResizeError(reason=reason) vm_ref = self._get_vm_opaque_ref(instance) sr_path = vm_utils.get_sr_path(self._session) if resize_down: self._migrate_disk_resizing_down( context, instance, dest, flavor, vm_ref, sr_path) else: self._migrate_disk_resizing_up( context, instance, dest, vm_ref, sr_path) self._detach_block_devices_from_orig_vm(instance, block_device_info) # NOTE(sirp): disk_info isn't used by the xenapi driver, instead it # uses a staging-area (/images/instance) and sequence-numbered # VHDs to figure out how to reconstruct the VDI chain after syncing disk_info = {} return disk_info def _detach_block_devices_from_orig_vm(self, instance, block_device_info): block_device_mapping = virt_driver.block_device_info_get_mapping( block_device_info) name_label = self._get_orig_vm_name_label(instance) for vol in block_device_mapping: connection_info = vol['connection_info'] mount_device = vol['mount_device'].rpartition("/")[2] self._volumeops.detach_volume(connection_info, name_label, mount_device) def _resize_up_vdis(self, instance, vdis): new_root_gb = instance['root_gb'] root_vdi = vdis.get('root') if new_root_gb and root_vdi: if root_vdi.get('osvol', False): # Don't resize root volumes. LOG.debug("Not resizing the root volume.", instance=instance) else: vdi_ref = root_vdi['ref'] vm_utils.update_vdi_virtual_size(self._session, instance, vdi_ref, new_root_gb) ephemeral_vdis = vdis.get('ephemerals') if not ephemeral_vdis: # NOTE(johngarbutt) no existing (migrated) ephemeral disks # to resize, so nothing more to do here. return total_ephemeral_gb = instance['ephemeral_gb'] if total_ephemeral_gb: sizes = vm_utils.get_ephemeral_disk_sizes(total_ephemeral_gb) # resize existing (migrated) ephemeral disks, # and add any extra disks if required due to a # larger total_ephemeral_gb (resize down is not supported). for userdevice, new_size in enumerate(sizes, start=int(DEVICE_EPHEMERAL)): vdi = ephemeral_vdis.get(str(userdevice)) if vdi: vdi_ref = vdi['ref'] vm_utils.update_vdi_virtual_size(self._session, instance, vdi_ref, new_size) else: LOG.debug("Generating new ephemeral vdi %d during resize", userdevice, instance=instance) # NOTE(johngarbutt) we generate but don't attach # the new disk to make up any additional ephemeral space vdi_ref = vm_utils.generate_single_ephemeral( self._session, instance, None, userdevice, new_size) vdis[str(userdevice)] = {'ref': vdi_ref, 'generated': True} def reboot(self, instance, reboot_type, bad_volumes_callback=None): """Reboot VM instance.""" # Note (salvatore-orlando): security group rules are not re-enforced # upon reboot, since this action on the XenAPI drivers does not # remove existing filters vm_ref = self._get_vm_opaque_ref(instance, check_rescue=True) try: if reboot_type == "HARD": self._session.call_xenapi('VM.hard_reboot', vm_ref) else: self._session.call_xenapi('VM.clean_reboot', vm_ref) except self._session.XenAPI.Failure as exc: details = exc.details if (details[0] == 'VM_BAD_POWER_STATE' and details[-1] == 'halted'): LOG.info(_LI("Starting halted instance found during reboot"), instance=instance) self._start(instance, vm_ref=vm_ref, bad_volumes_callback=bad_volumes_callback) return elif details[0] == 'SR_BACKEND_FAILURE_46': LOG.warning(_LW("Reboot failed due to bad volumes, detaching " "bad volumes and starting halted instance"), instance=instance) self._start(instance, vm_ref=vm_ref, bad_volumes_callback=bad_volumes_callback) return else: raise def set_admin_password(self, instance, new_pass): """Set the root/admin password on the VM instance.""" if self.agent_enabled(instance): vm_ref = self._get_vm_opaque_ref(instance) agent = self._get_agent(instance, vm_ref) agent.set_admin_password(new_pass) else: raise NotImplementedError() def inject_file(self, instance, path, contents): """Write a file to the VM instance.""" if self.agent_enabled(instance): vm_ref = self._get_vm_opaque_ref(instance) agent = self._get_agent(instance, vm_ref) agent.inject_file(path, contents) else: raise NotImplementedError() @staticmethod def _sanitize_xenstore_key(key): """Xenstore only allows the following characters as keys: ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmnopqrstuvwxyz 0123456789-/_@ So convert the others to _ Also convert / to _, because that is somewhat like a path separator. """ allowed_chars = ("ABCDEFGHIJKLMNOPQRSTUVWXYZ" "abcdefghijklmnopqrstuvwxyz" "0123456789-_@") return ''.join([x in allowed_chars and x or '_' for x in key]) def _inject_instance_metadata(self, instance, vm_ref): """Inject instance metadata into xenstore.""" @utils.synchronized('xenstore-' + instance['uuid']) def store_meta(topdir, data_dict): for key, value in data_dict.items(): key = self._sanitize_xenstore_key(key) value = value or '' self._add_to_param_xenstore(vm_ref, '%s/%s' % (topdir, key), jsonutils.dumps(value)) # Store user metadata store_meta('vm-data/user-metadata', utils.instance_meta(instance)) def _inject_auto_disk_config(self, instance, vm_ref): """Inject instance's auto_disk_config attribute into xenstore.""" @utils.synchronized('xenstore-' + instance['uuid']) def store_auto_disk_config(key, value): value = value and True or False self._add_to_param_xenstore(vm_ref, key, str(value)) store_auto_disk_config('vm-data/auto-disk-config', instance['auto_disk_config']) def change_instance_metadata(self, instance, diff): """Apply changes to instance metadata to xenstore.""" try: vm_ref = self._get_vm_opaque_ref(instance) except exception.NotFound: # NOTE(johngarbutt) race conditions mean we can still get here # during operations where the VM is not present, like resize. # Skip the update when not possible, as the updated metadata will # get added when the VM is being booted up at the end of the # resize or rebuild. LOG.warning(_LW("Unable to update metadata, VM not found."), instance=instance, exc_info=True) return def process_change(location, change): if change[0] == '-': self._remove_from_param_xenstore(vm_ref, location) try: self._delete_from_xenstore(instance, location, vm_ref=vm_ref) except exception.InstanceNotFound: # If the VM is not running then no need to update # the live xenstore - the param xenstore will be # used next time the VM is booted pass elif change[0] == '+': self._add_to_param_xenstore(vm_ref, location, jsonutils.dumps(change[1])) try: self._write_to_xenstore(instance, location, change[1], vm_ref=vm_ref) except exception.InstanceNotFound: # If the VM is not running then no need to update # the live xenstore pass @utils.synchronized('xenstore-' + instance['uuid']) def update_meta(): for key, change in diff.items(): key = self._sanitize_xenstore_key(key) location = 'vm-data/user-metadata/%s' % key process_change(location, change) update_meta() def _find_vdi_refs(self, vm_ref, exclude_volumes=False): """Find and return the root and ephemeral vdi refs for a VM.""" if not vm_ref: return {} vdi_refs = {} for vbd_ref in self._session.call_xenapi("VM.get_VBDs", vm_ref): vbd = self._session.call_xenapi("VBD.get_record", vbd_ref) if not exclude_volumes or 'osvol' not in vbd['other_config']: vdi_refs[vbd['userdevice']] = vbd['VDI'] return vdi_refs def _destroy_vdis(self, instance, vm_ref): """Destroys all VDIs associated with a VM.""" LOG.debug("Destroying VDIs", instance=instance) vdi_refs = vm_utils.lookup_vm_vdis(self._session, vm_ref) if not vdi_refs: return for vdi_ref in vdi_refs: try: vm_utils.destroy_vdi(self._session, vdi_ref) except exception.StorageError as exc: LOG.error(exc) def _destroy_kernel_ramdisk(self, instance, vm_ref): """Three situations can occur: 1. We have neither a ramdisk nor a kernel, in which case we are a RAW image and can omit this step 2. We have one or the other, in which case, we should flag as an error 3. We have both, in which case we safely remove both the kernel and the ramdisk. """ instance_uuid = instance['uuid'] if not instance['kernel_id'] and not instance['ramdisk_id']: # 1. No kernel or ramdisk LOG.debug("Using RAW or VHD, skipping kernel and ramdisk " "deletion", instance=instance) return if not (instance['kernel_id'] and instance['ramdisk_id']): # 2. We only have kernel xor ramdisk raise exception.InstanceUnacceptable(instance_id=instance_uuid, reason=_("instance has a kernel or ramdisk but not both")) # 3. We have both kernel and ramdisk (kernel, ramdisk) = vm_utils.lookup_kernel_ramdisk(self._session, vm_ref) if kernel or ramdisk: vm_utils.destroy_kernel_ramdisk(self._session, instance, kernel, ramdisk) LOG.debug("kernel/ramdisk files removed", instance=instance) def _destroy_rescue_instance(self, rescue_vm_ref, original_vm_ref): """Destroy a rescue instance.""" # Shutdown Rescue VM state = vm_utils.get_power_state(self._session, rescue_vm_ref) if state != power_state.SHUTDOWN: self._session.call_xenapi("VM.hard_shutdown", rescue_vm_ref) # Destroy Rescue VDIs vdi_refs = vm_utils.lookup_vm_vdis(self._session, rescue_vm_ref) # Don't destroy any VDIs belonging to the original VM orig_vdi_refs = self._find_vdi_refs(original_vm_ref) vdi_refs = set(vdi_refs) - set(orig_vdi_refs.values()) vm_utils.safe_destroy_vdis(self._session, vdi_refs) # Destroy Rescue VM self._session.call_xenapi("VM.destroy", rescue_vm_ref) def destroy(self, instance, network_info, block_device_info=None, destroy_disks=True): """Destroy VM instance. This is the method exposed by xenapi_conn.destroy(). The rest of the destroy_* methods are internal. """ LOG.info(_LI("Destroying VM"), instance=instance) # We don't use _get_vm_opaque_ref because the instance may # truly not exist because of a failure during build. A valid # vm_ref is checked correctly where necessary. vm_ref = vm_utils.lookup(self._session, instance['name']) rescue_vm_ref = vm_utils.lookup(self._session, "%s-rescue" % instance['name']) if rescue_vm_ref: self._destroy_rescue_instance(rescue_vm_ref, vm_ref) # NOTE(sirp): information about which volumes should be detached is # determined by the VBD.other_config['osvol'] attribute # NOTE(alaski): `block_device_info` is used to efficiently determine if # there's a volume attached, or which volumes to cleanup if there is # no VM present. return self._destroy(instance, vm_ref, network_info=network_info, destroy_disks=destroy_disks, block_device_info=block_device_info) def _destroy(self, instance, vm_ref, network_info=None, destroy_disks=True, block_device_info=None): """Destroys VM instance by performing: 1. A shutdown 2. Destroying associated VDIs. 3. Destroying kernel and ramdisk files (if necessary). 4. Destroying that actual VM record. """ if vm_ref is None: LOG.warning(_LW("VM is not present, skipping destroy..."), instance=instance) # NOTE(alaski): There should not be a block device mapping here, # but if there is it very likely means there was an error cleaning # it up previously and there is now an orphaned sr/pbd. This will # prevent both volume and instance deletes from completing. bdms = block_device_info['block_device_mapping'] or [] if not bdms: return for bdm in bdms: volume_id = bdm['connection_info']['data']['volume_id'] # Note(bobba): Check for the old-style SR first; if this # doesn't find the SR, also look for the new-style from # parse_sr_info sr_uuid = 'FA15E-D15C-%s' % volume_id sr_ref = None try: sr_ref = volume_utils.find_sr_by_uuid(self._session, sr_uuid) if not sr_ref: connection_data = bdm['connection_info']['data'] (sr_uuid, _, _) = volume_utils.parse_sr_info( connection_data) sr_ref = volume_utils.find_sr_by_uuid(self._session, sr_uuid) except Exception: LOG.exception(_LE('Failed to find an SR for volume %s'), volume_id, instance=instance) try: if sr_ref: volume_utils.forget_sr(self._session, sr_ref) else: LOG.error(_LE('Volume %s is associated with the ' 'instance but no SR was found for it'), volume_id, instance=instance) except Exception: LOG.exception(_LE('Failed to forget the SR for volume %s'), volume_id, instance=instance) return # NOTE(alaski): Attempt clean shutdown first if there's an attached # volume to reduce the risk of corruption. if block_device_info and block_device_info['block_device_mapping']: if not vm_utils.clean_shutdown_vm(self._session, instance, vm_ref): LOG.debug("Clean shutdown did not complete successfully, " "trying hard shutdown.", instance=instance) vm_utils.hard_shutdown_vm(self._session, instance, vm_ref) else: vm_utils.hard_shutdown_vm(self._session, instance, vm_ref) if destroy_disks: self._volumeops.detach_all(vm_ref) self._destroy_vdis(instance, vm_ref) self._destroy_kernel_ramdisk(instance, vm_ref) self.unplug_vifs(instance, network_info, vm_ref) self.firewall_driver.unfilter_instance( instance, network_info=network_info) vm_utils.destroy_vm(self._session, instance, vm_ref) def pause(self, instance): """Pause VM instance.""" vm_ref = self._get_vm_opaque_ref(instance) self._session.call_xenapi('VM.pause', vm_ref) def unpause(self, instance): """Unpause VM instance.""" vm_ref = self._get_vm_opaque_ref(instance) self._session.call_xenapi('VM.unpause', vm_ref) def suspend(self, instance): """Suspend the specified instance.""" vm_ref = self._get_vm_opaque_ref(instance) self._acquire_bootlock(vm_ref) self._session.call_xenapi('VM.suspend', vm_ref) def resume(self, instance): """Resume the specified instance.""" vm_ref = self._get_vm_opaque_ref(instance) self._release_bootlock(vm_ref) self._session.call_xenapi('VM.resume', vm_ref, False, True) def rescue(self, context, instance, network_info, image_meta, rescue_password): """Rescue the specified instance. - shutdown the instance VM. - set 'bootlock' to prevent the instance from starting in rescue. - spawn a rescue VM (the vm name-label will be instance-N-rescue). """ rescue_name_label = '%s-rescue' % instance.name rescue_vm_ref = vm_utils.lookup(self._session, rescue_name_label) if rescue_vm_ref: raise RuntimeError(_("Instance is already in Rescue Mode: %s") % instance.name) vm_ref = self._get_vm_opaque_ref(instance) vm_utils.hard_shutdown_vm(self._session, instance, vm_ref) self._acquire_bootlock(vm_ref) self.spawn(context, instance, image_meta, [], rescue_password, network_info, name_label=rescue_name_label, rescue=True) def set_bootable(self, instance, is_bootable): """Set the ability to power on/off an instance.""" vm_ref = self._get_vm_opaque_ref(instance) if is_bootable: self._release_bootlock(vm_ref) else: self._acquire_bootlock(vm_ref) def unrescue(self, instance): """Unrescue the specified instance. - unplug the instance VM's disk from the rescue VM. - teardown the rescue VM. - release the bootlock to allow the instance VM to start. """ rescue_vm_ref = vm_utils.lookup(self._session, "%s-rescue" % instance.name) if not rescue_vm_ref: raise exception.InstanceNotInRescueMode( instance_id=instance.uuid) original_vm_ref = self._get_vm_opaque_ref(instance) self._destroy_rescue_instance(rescue_vm_ref, original_vm_ref) self._release_bootlock(original_vm_ref) self._start(instance, original_vm_ref) def soft_delete(self, instance): """Soft delete the specified instance.""" try: vm_ref = self._get_vm_opaque_ref(instance) except exception.NotFound: LOG.warning(_LW("VM is not present, skipping soft delete..."), instance=instance) else: vm_utils.hard_shutdown_vm(self._session, instance, vm_ref) self._acquire_bootlock(vm_ref) def restore(self, instance): """Restore the specified instance.""" vm_ref = self._get_vm_opaque_ref(instance) self._release_bootlock(vm_ref) self._start(instance, vm_ref) def power_off(self, instance): """Power off the specified instance.""" vm_ref = self._get_vm_opaque_ref(instance) vm_utils.hard_shutdown_vm(self._session, instance, vm_ref) def power_on(self, instance): """Power on the specified instance.""" vm_ref = self._get_vm_opaque_ref(instance) self._start(instance, vm_ref) def _cancel_stale_tasks(self, timeout, task): """Cancel the given tasks that are older than the given timeout.""" task_refs = self._session.call_xenapi("task.get_by_name_label", task) for task_ref in task_refs: task_rec = self._session.call_xenapi("task.get_record", task_ref) task_created = timeutils.parse_strtime(task_rec["created"].value, "%Y%m%dT%H:%M:%SZ") if timeutils.is_older_than(task_created, timeout): self._session.call_xenapi("task.cancel", task_ref) def poll_rebooting_instances(self, timeout, instances): """Look for rebooting instances that can be expired. - issue a "hard" reboot to any instance that has been stuck in a reboot state for >= the given timeout """ # NOTE(jk0): All existing clean_reboot tasks must be canceled before # we can kick off the hard_reboot tasks. self._cancel_stale_tasks(timeout, 'VM.clean_reboot') ctxt = nova_context.get_admin_context() instances_info = dict(instance_count=len(instances), timeout=timeout) if instances_info["instance_count"] > 0: LOG.info(_LI("Found %(instance_count)d hung reboots " "older than %(timeout)d seconds"), instances_info) for instance in instances: LOG.info(_LI("Automatically hard rebooting"), instance=instance) self.compute_api.reboot(ctxt, instance, "HARD") def get_info(self, instance, vm_ref=None): """Return data about VM instance.""" vm_ref = vm_ref or self._get_vm_opaque_ref(instance) return vm_utils.compile_info(self._session, vm_ref) def get_diagnostics(self, instance): """Return data about VM diagnostics.""" vm_ref = self._get_vm_opaque_ref(instance) vm_rec = self._session.call_xenapi("VM.get_record", vm_ref) return vm_utils.compile_diagnostics(vm_rec) def get_instance_diagnostics(self, instance): """Return data about VM diagnostics using the common API.""" vm_ref = self._get_vm_opaque_ref(instance) vm_rec = self._session.VM.get_record(vm_ref) return vm_utils.compile_instance_diagnostics(instance, vm_rec) def _get_vif_device_map(self, vm_rec): vif_map = {} for vif in [self._session.call_xenapi("VIF.get_record", vrec) for vrec in vm_rec['VIFs']]: vif_map[vif['device']] = vif['MAC'] return vif_map def get_all_bw_counters(self): """Return running bandwidth counter for each interface on each running VM. """ counters = vm_utils.fetch_bandwidth(self._session) bw = {} for vm_ref, vm_rec in vm_utils.list_vms(self._session): vif_map = self._get_vif_device_map(vm_rec) name = vm_rec['name_label'] if 'nova_uuid' not in vm_rec['other_config']: continue dom = vm_rec.get('domid') if dom is None or dom not in counters: continue vifs_bw = bw.setdefault(name, {}) for vif_num, vif_data in six.iteritems(counters[dom]): mac = vif_map[vif_num] vif_data['mac_address'] = mac vifs_bw[mac] = vif_data return bw def get_console_output(self, instance): """Return last few lines of instance console.""" dom_id = self._get_last_dom_id(instance, check_rescue=True) try: raw_console_data = self._session.call_plugin('console', 'get_console_log', {'dom_id': dom_id}) except self._session.XenAPI.Failure: LOG.exception(_LE("Guest does not have a console available")) raise exception.ConsoleNotAvailable() return zlib.decompress(base64.b64decode(raw_console_data)) def get_vnc_console(self, instance): """Return connection info for a vnc console.""" if instance.vm_state == vm_states.RESCUED: name = '%s-rescue' % instance.name vm_ref = vm_utils.lookup(self._session, name) if vm_ref is None: # The rescue instance might not be ready at this point. raise exception.InstanceNotReady(instance_id=instance.uuid) else: vm_ref = vm_utils.lookup(self._session, instance.name) if vm_ref is None: # The compute manager expects InstanceNotFound for this case. raise exception.InstanceNotFound(instance_id=instance.uuid) session_id = self._session.get_session_id() path = "/console?ref=%s&session_id=%s" % (str(vm_ref), session_id) # NOTE: XS5.6sp2+ use http over port 80 for xenapi com return ctype.ConsoleVNC( host=CONF.vnc.vncserver_proxyclient_address, port=80, internal_access_path=path) def _vif_xenstore_data(self, vif): """convert a network info vif to injectable instance data.""" def get_ip(ip): if not ip: return None return ip['address'] def fixed_ip_dict(ip, subnet): if ip['version'] == 4: netmask = str(subnet.as_netaddr().netmask) else: netmask = subnet.as_netaddr()._prefixlen return {'ip': ip['address'], 'enabled': '1', 'netmask': netmask, 'gateway': get_ip(subnet['gateway'])} def convert_route(route): return {'route': str(netaddr.IPNetwork(route['cidr']).network), 'netmask': str(netaddr.IPNetwork(route['cidr']).netmask), 'gateway': get_ip(route['gateway'])} network = vif['network'] v4_subnets = [subnet for subnet in network['subnets'] if subnet['version'] == 4] v6_subnets = [subnet for subnet in network['subnets'] if subnet['version'] == 6] # NOTE(tr3buchet): routes and DNS come from all subnets routes = [convert_route(route) for subnet in network['subnets'] for route in subnet['routes']] dns = [get_ip(ip) for subnet in network['subnets'] for ip in subnet['dns']] info_dict = {'label': network['label'], 'mac': vif['address']} if v4_subnets: # NOTE(tr3buchet): gateway and broadcast from first subnet # primary IP will be from first subnet # subnets are generally unordered :( info_dict['gateway'] = get_ip(v4_subnets[0]['gateway']) info_dict['broadcast'] = str(v4_subnets[0].as_netaddr().broadcast) info_dict['ips'] = [fixed_ip_dict(ip, subnet) for subnet in v4_subnets for ip in subnet['ips']] if v6_subnets: # NOTE(tr3buchet): gateway from first subnet # primary IP will be from first subnet # subnets are generally unordered :( info_dict['gateway_v6'] = get_ip(v6_subnets[0]['gateway']) info_dict['ip6s'] = [fixed_ip_dict(ip, subnet) for subnet in v6_subnets for ip in subnet['ips']] if routes: info_dict['routes'] = routes if dns: info_dict['dns'] = list(set(dns)) return info_dict def inject_network_info(self, instance, network_info, vm_ref=None): """Generate the network info and make calls to place it into the xenstore and the xenstore param list. vm_ref can be passed in because it will sometimes be different than what vm_utils.lookup(session, instance['name']) will find (ex: rescue) """ vm_ref = vm_ref or self._get_vm_opaque_ref(instance) LOG.debug("Injecting network info to xenstore", instance=instance) @utils.synchronized('xenstore-' + instance['uuid']) def update_nwinfo(): for vif in network_info: xs_data = self._vif_xenstore_data(vif) location = ('vm-data/networking/%s' % vif['address'].replace(':', '')) self._add_to_param_xenstore(vm_ref, location, jsonutils.dumps(xs_data)) try: self._write_to_xenstore(instance, location, xs_data, vm_ref=vm_ref) except exception.InstanceNotFound: # If the VM is not running, no need to update the # live xenstore pass update_nwinfo() def _create_vifs(self, instance, vm_ref, network_info): """Creates vifs for an instance.""" LOG.debug("Creating vifs", instance=instance) vif_refs = [] # this function raises if vm_ref is not a vm_opaque_ref self._session.call_xenapi("VM.get_domid", vm_ref) for device, vif in enumerate(network_info): LOG.debug('Create VIF %s', vif, instance=instance) vif_ref = self.vif_driver.plug(instance, vif, vm_ref=vm_ref, device=device) vif_refs.append(vif_ref) LOG.debug('Created the vif_refs: %(vifs)s for VM name: %(name)s', {'vifs': vif_refs, 'name': instance['name']}, instance=instance) def plug_vifs(self, instance, network_info): """Set up VIF networking on the host.""" for device, vif in enumerate(network_info): self.vif_driver.plug(instance, vif, device=device) def unplug_vifs(self, instance, network_info, vm_ref): if network_info: for vif in network_info: self.vif_driver.unplug(instance, vif, vm_ref) def reset_network(self, instance, rescue=False): """Calls resetnetwork method in agent.""" if self.agent_enabled(instance): vm_ref = self._get_vm_opaque_ref(instance) agent = self._get_agent(instance, vm_ref) self._inject_hostname(instance, vm_ref, rescue) agent.resetnetwork() self._remove_hostname(instance, vm_ref) else: raise NotImplementedError() def _inject_hostname(self, instance, vm_ref, rescue): """Inject the hostname of the instance into the xenstore.""" hostname = instance['hostname'] if rescue: hostname = 'RESCUE-%s' % hostname if instance['os_type'] == "windows": # NOTE(jk0): Windows host names can only be <= 15 chars. hostname = hostname[:15] LOG.debug("Injecting hostname (%s) into xenstore", hostname, instance=instance) @utils.synchronized('xenstore-' + instance['uuid']) def update_hostname(): self._add_to_param_xenstore(vm_ref, 'vm-data/hostname', hostname) update_hostname() def _remove_hostname(self, instance, vm_ref): LOG.debug("Removing hostname from xenstore", instance=instance) @utils.synchronized('xenstore-' + instance['uuid']) def update_hostname(): self._remove_from_param_xenstore(vm_ref, 'vm-data/hostname') update_hostname() def _write_to_xenstore(self, instance, path, value, vm_ref=None): """Writes the passed value to the xenstore record for the given VM at the specified location. A XenAPIPlugin.PluginError will be raised if any error is encountered in the write process. """ return self._make_plugin_call('xenstore.py', 'write_record', instance, vm_ref=vm_ref, path=path, value=jsonutils.dumps(value)) def _read_from_xenstore(self, instance, path, ignore_missing_path=True, vm_ref=None): """Reads the passed location from xenstore for the given vm. Missing paths are ignored, unless explicitly stated not to, which causes xenstore to raise an exception. A XenAPIPlugin.PluginError is raised if any error is encountered in the read process. """ # NOTE(sulo): These need to be string for valid field type # for xapi. if ignore_missing_path: ignore_missing = 'True' else: ignore_missing = 'False' return self._make_plugin_call('xenstore.py', 'read_record', instance, vm_ref=vm_ref, path=path, ignore_missing_path=ignore_missing) def _delete_from_xenstore(self, instance, path, vm_ref=None): """Deletes the value from the xenstore record for the given VM at the specified location. A XenAPIPlugin.PluginError will be raised if any error is encountered in the delete process. """ return self._make_plugin_call('xenstore.py', 'delete_record', instance, vm_ref=vm_ref, path=path) def _make_plugin_call(self, plugin, method, instance=None, vm_ref=None, **addl_args): """Abstracts out the process of calling a method of a xenapi plugin. Any errors raised by the plugin will in turn raise a RuntimeError here. """ args = {} if instance or vm_ref: args['dom_id'] = self._get_dom_id(instance, vm_ref) args.update(addl_args) try: return self._session.call_plugin(plugin, method, args) except self._session.XenAPI.Failure as e: err_msg = e.details[-1].splitlines()[-1] if 'TIMEOUT:' in err_msg: LOG.error(_LE('TIMEOUT: The call to %(method)s timed out. ' 'args=%(args)r'), {'method': method, 'args': args}, instance=instance) return {'returncode': 'timeout', 'message': err_msg} elif 'NOT IMPLEMENTED:' in err_msg: LOG.error(_LE('NOT IMPLEMENTED: The call to %(method)s is not' ' supported by the agent. args=%(args)r'), {'method': method, 'args': args}, instance=instance) return {'returncode': 'notimplemented', 'message': err_msg} else: LOG.error(_LE('The call to %(method)s returned an error: ' '%(e)s. args=%(args)r'), {'method': method, 'args': args, 'e': e}, instance=instance) return {'returncode': 'error', 'message': err_msg} def _get_dom_id(self, instance, vm_ref=None, check_rescue=False): vm_ref = vm_ref or self._get_vm_opaque_ref(instance, check_rescue) domid = self._session.call_xenapi("VM.get_domid", vm_ref) if not domid or domid == "-1": raise exception.InstanceNotFound(instance_id=instance['name']) return domid def _get_last_dom_id(self, instance, vm_ref=None, check_rescue=False): vm_ref = vm_ref or self._get_vm_opaque_ref(instance, check_rescue) other_config = self._session.call_xenapi("VM.get_other_config", vm_ref) if 'last_dom_id' not in other_config: raise exception.InstanceNotFound(instance_id=instance['name']) return other_config['last_dom_id'] def _add_to_param_xenstore(self, vm_ref, key, val): """Takes a key/value pair and adds it to the xenstore parameter record for the given vm instance. If the key exists in xenstore, it is overwritten """ self._remove_from_param_xenstore(vm_ref, key) self._session.call_xenapi('VM.add_to_xenstore_data', vm_ref, key, val) def _remove_from_param_xenstore(self, vm_ref, key): """Takes a single key and removes it from the xenstore parameter record data for the given VM. If the key doesn't exist, the request is ignored. """ self._session.call_xenapi('VM.remove_from_xenstore_data', vm_ref, key) def refresh_security_group_rules(self, security_group_id): """recreates security group rules for every instance.""" self.firewall_driver.refresh_security_group_rules(security_group_id) def refresh_instance_security_rules(self, instance): """recreates security group rules for specified instance.""" self.firewall_driver.refresh_instance_security_rules(instance) def unfilter_instance(self, instance_ref, network_info): """Removes filters for each VIF of the specified instance.""" self.firewall_driver.unfilter_instance(instance_ref, network_info=network_info) def _get_host_uuid_from_aggregate(self, context, hostname): aggregate_list = objects.AggregateList.get_by_host( context, CONF.host, key=pool_states.POOL_FLAG) reason = _('Destination host:%s must be in the same ' 'aggregate as the source server') % hostname if len(aggregate_list) == 0: raise exception.MigrationPreCheckError(reason=reason) if hostname not in aggregate_list[0].metadata: raise exception.MigrationPreCheckError(reason=reason) return aggregate_list[0].metadata[hostname] def _ensure_host_in_aggregate(self, context, hostname): self._get_host_uuid_from_aggregate(context, hostname) def _get_host_opaque_ref(self, context, hostname): host_uuid = self._get_host_uuid_from_aggregate(context, hostname) return self._session.call_xenapi("host.get_by_uuid", host_uuid) def _migrate_receive(self, ctxt): destref = self._session.host_ref # Get the network to for migrate. # This is the one associated with the pif marked management. From cli: # uuid=`xe pif-list --minimal management=true` # xe pif-param-get param-name=network-uuid uuid=$uuid expr = 'field "management" = "true"' pifs = self._session.call_xenapi('PIF.get_all_records_where', expr) if len(pifs) != 1: msg = _('No suitable network for migrate') raise exception.MigrationPreCheckError(reason=msg) pifkey = list(pifs.keys())[0] if not (netutils.is_valid_ipv4(pifs[pifkey]['IP']) or netutils.is_valid_ipv6(pifs[pifkey]['IPv6'])): msg = (_('PIF %s does not contain IP address') % pifs[pifkey]['uuid']) raise exception.MigrationPreCheckError(reason=msg) nwref = pifs[list(pifs.keys())[0]]['network'] try: options = {} migrate_data = self._session.call_xenapi("host.migrate_receive", destref, nwref, options) except self._session.XenAPI.Failure: LOG.exception(_LE('Migrate Receive failed')) msg = _('Migrate Receive failed') raise exception.MigrationPreCheckError(reason=msg) return migrate_data def _get_iscsi_srs(self, ctxt, instance_ref): vm_ref = self._get_vm_opaque_ref(instance_ref) vbd_refs = self._session.call_xenapi("VM.get_VBDs", vm_ref) iscsi_srs = [] for vbd_ref in vbd_refs: vdi_ref = self._session.call_xenapi("VBD.get_VDI", vbd_ref) # Check if it's on an iSCSI SR sr_ref = self._session.call_xenapi("VDI.get_SR", vdi_ref) if self._session.call_xenapi("SR.get_type", sr_ref) == 'iscsi': iscsi_srs.append(sr_ref) return iscsi_srs def check_can_live_migrate_destination(self, ctxt, instance_ref, block_migration=False, disk_over_commit=False): """Check if it is possible to execute live migration. :param ctxt: security context :param instance_ref: nova.db.sqlalchemy.models.Instance object :param block_migration: if true, prepare for block migration if None, calculate it from driver :param disk_over_commit: if true, allow disk over commit """ dest_check_data = objects.XenapiLiveMigrateData() # Notes(eliqiao): if block_migration is None, we calculate it # by checking if src and dest node are in same aggregate if block_migration is None: src = instance_ref['host'] try: self._ensure_host_in_aggregate(ctxt, src) except exception.MigrationPreCheckError: block_migration = True else: sr_ref = vm_utils.safe_find_sr(self._session) sr_rec = self._session.get_rec('SR', sr_ref) block_migration = not sr_rec["shared"] if block_migration: dest_check_data.block_migration = True dest_check_data.migrate_send_data = self._migrate_receive(ctxt) dest_check_data.destination_sr_ref = vm_utils.safe_find_sr( self._session) else: dest_check_data.block_migration = False src = instance_ref['host'] # TODO(eilqiao): There is still one case that block_migration is # passed from admin user, so we need this check until # block_migration flag is removed from API self._ensure_host_in_aggregate(ctxt, src) # TODO(johngarbutt) we currently assume # instance is on a SR shared with other destination # block migration work will be able to resolve this return dest_check_data def _is_xsm_sr_check_relaxed(self): try: return self.cached_xsm_sr_relaxed except AttributeError: config_value = None try: config_value = self._make_plugin_call('config_file', 'get_val', key='relax-xsm-sr-check') except Exception: LOG.exception(_LE('Plugin config_file get_val failed')) self.cached_xsm_sr_relaxed = config_value == "true" return self.cached_xsm_sr_relaxed def check_can_live_migrate_source(self, ctxt, instance_ref, dest_check_data): """Check if it's possible to execute live migration on the source side. :param ctxt: security context :param instance_ref: nova.db.sqlalchemy.models.Instance object :param dest_check_data: data returned by the check on the destination, includes block_migration flag """ if len(self._get_iscsi_srs(ctxt, instance_ref)) > 0: # XAPI must support the relaxed SR check for live migrating with # iSCSI VBDs if not self._is_xsm_sr_check_relaxed(): raise exception.MigrationError(reason=_('XAPI supporting ' 'relax-xsm-sr-check=true required')) if not isinstance(dest_check_data, migrate_data_obj.LiveMigrateData): obj = objects.XenapiLiveMigrateData() obj.from_legacy_dict(dest_check_data) dest_check_data = obj if ('block_migration' in dest_check_data and dest_check_data.block_migration): vm_ref = self._get_vm_opaque_ref(instance_ref) try: self._call_live_migrate_command( "VM.assert_can_migrate", vm_ref, dest_check_data) except self._session.XenAPI.Failure as exc: reason = exc.details[0] msg = _('assert_can_migrate failed because: %s') % reason LOG.debug(msg, exc_info=True) raise exception.MigrationPreCheckError(reason=msg) return dest_check_data def _ensure_pv_driver_info_for_live_migration(self, instance, vm_ref): """Checks if pv drivers are present for this instance. If it is present but not reported, try to fake the info for live-migration. """ if self._pv_driver_version_reported(instance, vm_ref): # Since driver version is reported we do not need to do anything return if self._pv_device_reported(instance, vm_ref): LOG.debug("PV device present but missing pv driver info. " "Attempting to insert missing info in xenstore.", instance=instance) self._write_fake_pv_version(instance, vm_ref) else: LOG.debug("Could not determine the presence of pv device. " "Skipping inserting pv driver info.", instance=instance) def _pv_driver_version_reported(self, instance, vm_ref): xs_path = "attr/PVAddons/MajorVersion" major_version = self._read_from_xenstore(instance, xs_path, vm_ref=vm_ref) LOG.debug("Major Version: %s reported.", major_version, instance=instance) # xenstore reports back string only, if the path is missing we get # None as string, since missing paths are ignored. if major_version == '"None"': return False else: return True def _pv_device_reported(self, instance, vm_ref): vm_rec = self._session.VM.get_record(vm_ref) vif_list = [self._session.call_xenapi("VIF.get_record", vrec) for vrec in vm_rec['VIFs']] net_devices = [vif['device'] for vif in vif_list] # NOTE(sulo): We infer the presence of pv driver # by the presence of a pv network device. If xenstore reports # device status as connected (status=4) we take that as the presence # of pv driver. Any other status will likely cause migration to fail. for device in net_devices: xs_path = "device/vif/%s/state" % device ret = self._read_from_xenstore(instance, xs_path, vm_ref=vm_ref) LOG.debug("PV Device vif.%(vif_ref)s reporting state %(ret)s", {'vif_ref': device, 'ret': ret}, instance=instance) if strutils.is_int_like(ret) and int(ret) == 4: return True return False def _write_fake_pv_version(self, instance, vm_ref): version = self._session.product_version LOG.debug("Writing pvtools version major: %(major)s minor: %(minor)s " "micro: %(micro)s", {'major': version[0], 'minor': version[1], 'micro': version[2]}, instance=instance) major_ver = "attr/PVAddons/MajorVersion" self._write_to_xenstore(instance, major_ver, version[0], vm_ref=vm_ref) minor_ver = "attr/PVAddons/MinorVersion" self._write_to_xenstore(instance, minor_ver, version[1], vm_ref=vm_ref) micro_ver = "attr/PVAddons/MicroVersion" self._write_to_xenstore(instance, micro_ver, version[2], vm_ref=vm_ref) xs_path = "data/updated" self._write_to_xenstore(instance, xs_path, "1", vm_ref=vm_ref) def _generate_vdi_map(self, destination_sr_ref, vm_ref, sr_ref=None): """generate a vdi_map for _call_live_migrate_command.""" if sr_ref is None: sr_ref = vm_utils.safe_find_sr(self._session) vm_vdis = vm_utils.get_instance_vdis_for_sr(self._session, vm_ref, sr_ref) return {vdi: destination_sr_ref for vdi in vm_vdis} def _call_live_migrate_command(self, command_name, vm_ref, migrate_data): """unpack xapi specific parameters, and call a live migrate command.""" # NOTE(coreywright): though a nullable object field, migrate_send_data # is required for XenAPI live migration commands migrate_send_data = None if 'migrate_send_data' in migrate_data: migrate_send_data = migrate_data.migrate_send_data if not migrate_send_data: raise exception.InvalidParameterValue( 'XenAPI requires destination migration data') # NOTE(coreywright): convert to xmlrpc marshallable type migrate_send_data = dict(migrate_send_data) destination_sr_ref = migrate_data.destination_sr_ref vdi_map = self._generate_vdi_map(destination_sr_ref, vm_ref) # Add destination SR refs for all of the VDIs that we created # as part of the pre migration callback sr_uuid_map = None if "sr_uuid_map" in migrate_data: sr_uuid_map = migrate_data.sr_uuid_map if sr_uuid_map: for sr_uuid in sr_uuid_map: # Source and destination SRs have the same UUID, so get the # reference for the local SR sr_ref = self._session.call_xenapi("SR.get_by_uuid", sr_uuid) vdi_map.update( self._generate_vdi_map( sr_uuid_map[sr_uuid], vm_ref, sr_ref)) vif_map = {} options = {} self._session.call_xenapi(command_name, vm_ref, migrate_send_data, True, vdi_map, vif_map, options) def pre_live_migration(self, context, instance, block_device_info, network_info, disk_info, migrate_data): if migrate_data is None: raise exception.InvalidParameterValue( 'XenAPI requires migrate data') migrate_data.sr_uuid_map = self.connect_block_device_volumes( block_device_info) return migrate_data def live_migrate(self, context, instance, destination_hostname, post_method, recover_method, block_migration, migrate_data=None): try: vm_ref = self._get_vm_opaque_ref(instance) # NOTE(sulo): We try to ensure that PV driver information is # present in xenstore for the instance we are trying to # live-migrate, if the process of faking pv version info fails, # we simply log it and carry on with the rest of the process. # Any xapi error due to PV version are caught and migration # will be safely reverted by the rollback process. try: self._ensure_pv_driver_info_for_live_migration(instance, vm_ref) except Exception as e: LOG.warning(e) if migrate_data is not None: (kernel, ramdisk) = vm_utils.lookup_kernel_ramdisk( self._session, vm_ref) migrate_data.kernel_file = kernel migrate_data.ramdisk_file = ramdisk if migrate_data is not None and migrate_data.block_migration: iscsi_srs = self._get_iscsi_srs(context, instance) try: self._call_live_migrate_command( "VM.migrate_send", vm_ref, migrate_data) except self._session.XenAPI.Failure: LOG.exception(_LE('Migrate Send failed')) raise exception.MigrationError( reason=_('Migrate Send failed')) # Tidy up the iSCSI SRs for sr_ref in iscsi_srs: volume_utils.forget_sr(self._session, sr_ref) else: host_ref = self._get_host_opaque_ref(context, destination_hostname) self._session.call_xenapi("VM.pool_migrate", vm_ref, host_ref, {"live": "true"}) post_method(context, instance, destination_hostname, block_migration, migrate_data) except Exception: with excutils.save_and_reraise_exception(): recover_method(context, instance, destination_hostname, block_migration) def post_live_migration(self, context, instance, migrate_data=None): if migrate_data is not None: vm_utils.destroy_kernel_ramdisk(self._session, instance, migrate_data.kernel_file, migrate_data.ramdisk_file) def post_live_migration_at_destination(self, context, instance, network_info, block_migration, block_device_info): # FIXME(johngarbutt): we should block all traffic until we have # applied security groups, however this requires changes to XenServer self._prepare_instance_filter(instance, network_info) self.firewall_driver.apply_instance_filter(instance, network_info) vm_utils.create_kernel_and_ramdisk(context, self._session, instance, instance['name']) # NOTE(johngarbutt) workaround XenServer bug CA-98606 vm_ref = self._get_vm_opaque_ref(instance) vm_utils.strip_base_mirror_from_vdis(self._session, vm_ref) def rollback_live_migration_at_destination(self, instance, block_device_info): bdms = block_device_info['block_device_mapping'] or [] for bdm in bdms: conn_data = bdm['connection_info']['data'] uuid, label, params = volume_utils.parse_sr_info(conn_data) try: sr_ref = volume_utils.find_sr_by_uuid(self._session, uuid) if sr_ref: volume_utils.forget_sr(self._session, sr_ref) except Exception: LOG.exception(_LE('Failed to forget the SR for volume %s'), params['id'], instance=instance) def get_per_instance_usage(self): """Get usage info about each active instance.""" usage = {} def _is_active(vm_rec): power_state = vm_rec['power_state'].lower() return power_state in ['running', 'paused'] def _get_uuid(vm_rec): other_config = vm_rec['other_config'] return other_config.get('nova_uuid', None) for vm_ref, vm_rec in vm_utils.list_vms(self._session): uuid = _get_uuid(vm_rec) if _is_active(vm_rec) and uuid is not None: memory_mb = int(vm_rec['memory_static_max']) / units.Mi usage[uuid] = {'memory_mb': memory_mb, 'uuid': uuid} return usage def connect_block_device_volumes(self, block_device_info): sr_uuid_map = {} try: if block_device_info is not None: for block_device_map in block_device_info[ 'block_device_mapping']: sr_uuid, _ = self._volumeops.connect_volume( block_device_map['connection_info']) sr_ref = self._session.call_xenapi('SR.get_by_uuid', sr_uuid) sr_uuid_map[sr_uuid] = sr_ref except Exception: with excutils.save_and_reraise_exception(): # Disconnect the volumes we just connected for sr_ref in six.itervalues(sr_uuid_map): volume_utils.forget_sr(self._session, sr_ref) return sr_uuid_map