diff options
| author | Ilya Alekseyev <ialekseev@griddynamics.com> | 2011-04-18 15:25:44 +0400 |
|---|---|---|
| committer | Ilya Alekseyev <ialekseev@griddynamics.com> | 2011-04-18 15:25:44 +0400 |
| commit | 1e02e152aea76d5af30c42708caaee2fb1f05ede (patch) | |
| tree | 57d6498fe5122f291f3e4899b1816487fc5bbff3 /nova/virt | |
| parent | 519c885a4e3622417cd78655a211a2f23033d610 (diff) | |
| parent | 9761e7690ab8f13a76c761ba6055372a0d29eaff (diff) | |
trunk merged. conflict resolved.
Diffstat (limited to 'nova/virt')
| -rw-r--r-- | nova/virt/libvirt_conn.py | 227 | ||||
| -rw-r--r-- | nova/virt/vmwareapi_conn.py | 4 | ||||
| -rw-r--r-- | nova/virt/xenapi/vm_utils.py | 27 | ||||
| -rw-r--r-- | nova/virt/xenapi/vmops.py | 43 | ||||
| -rw-r--r-- | nova/virt/xenapi_conn.py | 9 |
5 files changed, 230 insertions, 80 deletions
diff --git a/nova/virt/libvirt_conn.py b/nova/virt/libvirt_conn.py index 1eec55e5f..bc54835ff 100644 --- a/nova/virt/libvirt_conn.py +++ b/nova/virt/libvirt_conn.py @@ -117,6 +117,8 @@ flags.DEFINE_integer('live_migration_bandwidth', 0, 'Define live migration behavior') flags.DEFINE_string('qemu_img', 'qemu-img', 'binary to use for qemu-img commands') +flags.DEFINE_bool('start_guests_on_host_boot', False, + 'Whether to restart guests when the host reboots') def get_connection(read_only): @@ -172,36 +174,36 @@ def _get_network_info(instance): instance['instance_type']) network_info = [] - def ip_dict(ip): - return { - "ip": ip.address, - "netmask": network["netmask"], - "enabled": "1"} - - def ip6_dict(ip6): - prefix = ip6.network.cidr_v6 - mac = instance.mac_address - return { - "ip": utils.to_global_ipv6(prefix, mac), - "netmask": ip6.network.netmask_v6, - "gateway": ip6.network.gateway_v6, - "enabled": "1"} - for network in networks: network_ips = [ip for ip in ip_addresses - if ip.network_id == network.id] + if ip['network_id'] == network['id']] + + def ip_dict(ip): + return { + 'ip': ip['address'], + 'netmask': network['netmask'], + 'enabled': '1'} + + def ip6_dict(): + prefix = network['cidr_v6'] + mac = instance['mac_address'] + return { + 'ip': utils.to_global_ipv6(prefix, mac), + 'netmask': network['netmask_v6'], + 'enabled': '1'} mapping = { 'label': network['label'], 'gateway': network['gateway'], 'broadcast': network['broadcast'], - 'mac': instance.mac_address, + 'mac': instance['mac_address'], 'rxtx_cap': flavor['rxtx_cap'], 'dns': [network['dns']], 'ips': [ip_dict(ip) for ip in network_ips]} if FLAGS.use_ipv6: - mapping['ip6s'] = [ip6_dict(ip) for ip in network_ips] + mapping['ip6s'] = [ip6_dict()] + mapping['gateway6'] = network['gateway_v6'] network_info.append((network, mapping)) return network_info @@ -236,12 +238,8 @@ class LibvirtConnection(driver.ComputeDriver): {'name': instance['name'], 'state': state}) db.instance_set_state(ctxt, instance['id'], state) - if state == power_state.SHUTOFF: - # TODO(soren): This is what the compute manager does when you - # terminate # an instance. At some point I figure we'll have a - # "terminated" state and some sort of cleanup job that runs - # occasionally, cleaning them out. - db.instance_destroy(ctxt, instance['id']) + # NOTE(justinsb): We no longer delete SHUTOFF instances, + # the user may want to power them back on if state != power_state.RUNNING: continue @@ -316,12 +314,57 @@ class LibvirtConnection(driver.ComputeDriver): return infos def destroy(self, instance, cleanup=True): + instance_name = instance['name'] + + # TODO(justinsb): Refactor all lookupByName calls for error-handling try: - virt_dom = self._conn.lookupByName(instance['name']) - virt_dom.destroy() - except Exception as _err: - pass - # If the instance is already terminated, we're still happy + virt_dom = self._conn.lookupByName(instance_name) + except libvirt.libvirtError as e: + errcode = e.get_error_code() + if errcode == libvirt.VIR_ERR_NO_DOMAIN: + virt_dom = None + else: + LOG.warning(_("Error from libvirt during lookup of " + "%(instance_name)s. Code=%(errcode)s " + "Error=%(e)s") % + locals()) + raise + + # If the instance is already terminated, we're still happy + # Otherwise, destroy it + if virt_dom is not None: + try: + virt_dom.destroy() + except libvirt.libvirtError as e: + is_okay = False + errcode = e.get_error_code() + if errcode == libvirt.VIR_ERR_OPERATION_INVALID: + # If the instance if already shut off, we get this: + # Code=55 Error=Requested operation is not valid: + # domain is not running + (state, _max_mem, _mem, _cpus, _t) = virt_dom.info() + if state == power_state.SHUTOFF: + is_okay = True + + if not is_okay: + LOG.warning(_("Error from libvirt during destroy of " + "%(instance_name)s. Code=%(errcode)s " + "Error=%(e)s") % + locals()) + raise + + try: + # NOTE(justinsb): We remove the domain definition. We probably + # would do better to keep it if cleanup=False (e.g. volumes?) + # (e.g. #2 - not losing machines on failure) + virt_dom.undefine() + except libvirt.libvirtError as e: + errcode = e.get_error_code() + LOG.warning(_("Error from libvirt during undefine of " + "%(instance_name)s. Code=%(errcode)s " + "Error=%(e)s") % + locals()) + raise # We'll save this for when we do shutdown, # instead of destroy - but destroy returns immediately @@ -332,12 +375,18 @@ class LibvirtConnection(driver.ComputeDriver): state = self.get_info(instance['name'])['state'] db.instance_set_state(context.get_admin_context(), instance['id'], state) - if state == power_state.SHUTDOWN: + if state == power_state.SHUTOFF: break - except Exception: + + # Let's not hammer on the DB + time.sleep(1) + except Exception as ex: + msg = _("Error encountered when destroying instance '%(id)s': " + "%(ex)s") % {"id": instance["id"], "ex": ex} + LOG.debug(msg) db.instance_set_state(context.get_admin_context(), instance['id'], - power_state.SHUTDOWN) + power_state.SHUTOFF) break self.firewall_driver.unfilter_instance(instance) @@ -429,7 +478,6 @@ class LibvirtConnection(driver.ComputeDriver): 'container_format': base['container_format'], 'is_public': False, 'properties': {'architecture': base['architecture'], - 'type': base['type'], 'name': '%s.%s' % (base['name'], image_id), 'kernel_id': instance['kernel_id'], 'image_location': 'snapshot', @@ -480,7 +528,7 @@ class LibvirtConnection(driver.ComputeDriver): xml = self.to_xml(instance) self.firewall_driver.setup_basic_filtering(instance) self.firewall_driver.prepare_instance_filter(instance) - self._conn.createXML(xml, 0) + self._create_new_domain(xml) self.firewall_driver.apply_instance_filter(instance) timer = utils.LoopingCall(f=None) @@ -528,7 +576,7 @@ class LibvirtConnection(driver.ComputeDriver): 'kernel_id': FLAGS.rescue_kernel_id, 'ramdisk_id': FLAGS.rescue_ramdisk_id} self._create_image(instance, xml, '.rescue', rescue_images) - self._conn.createXML(xml, 0) + self._create_new_domain(xml) timer = utils.LoopingCall(f=None) @@ -571,10 +619,15 @@ class LibvirtConnection(driver.ComputeDriver): self.firewall_driver.setup_basic_filtering(instance, network_info) self.firewall_driver.prepare_instance_filter(instance, network_info) self._create_image(instance, xml, network_info) - self._conn.createXML(xml, 0) + domain = self._create_new_domain(xml) LOG.debug(_("instance %s: is running"), instance['name']) self.firewall_driver.apply_instance_filter(instance) + if FLAGS.start_guests_on_host_boot: + LOG.debug(_("instance %s: setting autostart ON") % + instance['name']) + domain.setAutostart(1) + timer = utils.LoopingCall(f=None) def _wait_for_boot(): @@ -802,7 +855,10 @@ class LibvirtConnection(driver.ComputeDriver): root_fname = '%08x' % int(disk_images['image_id']) size = FLAGS.minimum_root_size - if inst['instance_type'] == 'm1.tiny' or suffix == '.rescue': + + inst_type_id = inst['instance_type_id'] + inst_type = instance_types.get_instance_type(inst_type_id) + if inst_type['name'] == 'm1.tiny' or suffix == '.rescue': size = None root_fname += "_sm" @@ -814,14 +870,13 @@ class LibvirtConnection(driver.ComputeDriver): user=user, project=project, size=size) - type_data = instance_types.get_instance_type(inst['instance_type']) - if type_data['local_gb']: + if inst_type['local_gb']: self._cache_image(fn=self._create_local, target=basepath('disk.local'), - fname="local_%s" % type_data['local_gb'], + fname="local_%s" % inst_type['local_gb'], cow=FLAGS.use_cow_images, - local_gb=type_data['local_gb']) + local_gb=inst_type['local_gb']) # For now, we assume that if we're not using a kernel, we're using a # partitioned disk image where the target partition is the first @@ -847,7 +902,7 @@ class LibvirtConnection(driver.ComputeDriver): for (network_ref, mapping) in network_info: ifc_num += 1 - if not 'injected' in network_ref: + if not network_ref['injected']: continue have_injected_networks = True @@ -955,8 +1010,8 @@ class LibvirtConnection(driver.ComputeDriver): nics.append(self._get_nic_for_xml(network, mapping)) # FIXME(vish): stick this in db - instance_type_name = instance['instance_type'] - instance_type = instance_types.get_instance_type(instance_type_name) + inst_type_id = instance['instance_type_id'] + inst_type = instance_types.get_instance_type(inst_type_id) if FLAGS.use_cow_images: driver_type = 'qcow2' @@ -967,15 +1022,16 @@ class LibvirtConnection(driver.ComputeDriver): 'name': instance['name'], 'basepath': os.path.join(FLAGS.instances_path, instance['name']), - 'memory_kb': instance_type['memory_mb'] * 1024, - 'vcpus': instance_type['vcpus'], + 'memory_kb': inst_type['memory_mb'] * 1024, + 'vcpus': inst_type['vcpus'], 'rescue': rescue, - 'local': instance_type['local_gb'], + 'local': inst_type['local_gb'], 'driver_type': driver_type, 'nics': nics} if FLAGS.vnc_enabled: - xml_info['vncserver_host'] = FLAGS.vncserver_host + if FLAGS.libvirt_type != 'lxc': + xml_info['vncserver_host'] = FLAGS.vncserver_host if not rescue: if instance['kernel_id']: xml_info['kernel'] = xml_info['basepath'] + "/kernel" @@ -991,11 +1047,22 @@ class LibvirtConnection(driver.ComputeDriver): return xml def get_info(self, instance_name): + # NOTE(justinsb): When libvirt isn't running / can't connect, we get: + # libvir: Remote error : unable to connect to + # '/var/run/libvirt/libvirt-sock', libvirtd may need to be started: + # No such file or directory try: virt_dom = self._conn.lookupByName(instance_name) - except: - raise exception.NotFound(_("Instance %s not found") - % instance_name) + except libvirt.libvirtError as e: + errcode = e.get_error_code() + if errcode == libvirt.VIR_ERR_NO_DOMAIN: + raise exception.NotFound(_("Instance %s not found") + % instance_name) + LOG.warning(_("Error from libvirt during lookup. " + "Code=%(errcode)s Error=%(e)s") % + locals()) + raise + (state, max_mem, mem, num_cpu, cpu_time) = virt_dom.info() return {'state': state, 'max_mem': max_mem, @@ -1003,6 +1070,24 @@ class LibvirtConnection(driver.ComputeDriver): 'num_cpu': num_cpu, 'cpu_time': cpu_time} + def _create_new_domain(self, xml, persistent=True, launch_flags=0): + # NOTE(justinsb): libvirt has two types of domain: + # * a transient domain disappears when the guest is shutdown + # or the host is rebooted. + # * a permanent domain is not automatically deleted + # NOTE(justinsb): Even for ephemeral instances, transient seems risky + + if persistent: + # To create a persistent domain, first define it, then launch it. + domain = self._conn.defineXML(xml) + + domain.createWithFlags(launch_flags) + else: + # createXML call creates a transient domain + domain = self._conn.createXML(xml, launch_flags) + + return domain + def get_diagnostics(self, instance_name): raise exception.ApiError(_("diagnostics are not supported " "for libvirt")) @@ -1406,18 +1491,13 @@ class LibvirtConnection(driver.ComputeDriver): # wait for completion timeout_count = range(FLAGS.live_migration_retry_count) while timeout_count: - try: - filter_name = 'nova-instance-%s' % instance_ref.name - self._conn.nwfilterLookupByName(filter_name) + if self.firewall_driver.instance_filter_exists(instance_ref): break - except libvirt.libvirtError: - timeout_count.pop() - if len(timeout_count) == 0: - ec2_id = instance_ref['hostname'] - iname = instance_ref.name - msg = _('Timeout migrating for %(ec2_id)s(%(iname)s)') - raise exception.Error(msg % locals()) - time.sleep(1) + timeout_count.pop() + if len(timeout_count) == 0: + msg = _('Timeout migrating for %s. nwfilter not found.') + raise exception.Error(msg % instance_ref.name) + time.sleep(1) def live_migration(self, ctxt, instance_ref, dest, post_method, recover_method): @@ -1546,6 +1626,10 @@ class FirewallDriver(object): """ raise NotImplementedError() + def instance_filter_exists(self, instance): + """Check nova-instance-instance-xxx exists""" + raise NotImplementedError() + class NWFilterFirewall(FirewallDriver): """ @@ -1853,6 +1937,21 @@ class NWFilterFirewall(FirewallDriver): return 'nova-instance-%s' % (instance['name']) return 'nova-instance-%s-%s' % (instance['name'], nic_id) + def instance_filter_exists(self, instance): + """Check nova-instance-instance-xxx exists""" + network_info = _get_network_info(instance) + for (network, mapping) in network_info: + nic_id = mapping['mac'].replace(':', '') + instance_filter_name = self._instance_filter_name(instance, nic_id) + try: + self._conn.nwfilterLookupByName(instance_filter_name) + except libvirt.libvirtError: + name = instance.name + LOG.debug(_('The nwfilter(%(instance_filter_name)s) for' + '%(name)s is not found.') % locals()) + return False + return True + class IptablesFirewallDriver(FirewallDriver): def __init__(self, execute=None, **kwargs): @@ -2042,6 +2141,10 @@ class IptablesFirewallDriver(FirewallDriver): return ipv4_rules, ipv6_rules + def instance_filter_exists(self, instance): + """Check nova-instance-instance-xxx exists""" + return self.nwfilter.instance_filter_exists(instance) + def refresh_security_group_members(self, security_group): pass diff --git a/nova/virt/vmwareapi_conn.py b/nova/virt/vmwareapi_conn.py index 20c1b2b45..1c6d2572d 100644 --- a/nova/virt/vmwareapi_conn.py +++ b/nova/virt/vmwareapi_conn.py @@ -42,6 +42,7 @@ from nova import exception from nova import flags
from nova import log as logging
from nova import utils
+from nova.virt import driver
from nova.virt.vmwareapi import error_util
from nova.virt.vmwareapi import vim
from nova.virt.vmwareapi import vim_util
@@ -104,11 +105,12 @@ def get_connection(_): api_retry_count)
-class VMWareESXConnection(object):
+class VMWareESXConnection(driver.ComputeDriver):
"""The ESX host connection object."""
def __init__(self, host_ip, host_username, host_password,
api_retry_count, scheme="https"):
+ super(VMWareESXConnection, self).__init__()
session = VMWareAPISession(host_ip, host_username, host_password,
api_retry_count, scheme=scheme)
self._vmops = VMWareVMOps(session)
diff --git a/nova/virt/xenapi/vm_utils.py b/nova/virt/xenapi/vm_utils.py index d07d60800..d2045a557 100644 --- a/nova/virt/xenapi/vm_utils.py +++ b/nova/virt/xenapi/vm_utils.py @@ -49,6 +49,8 @@ LOG = logging.getLogger("nova.virt.xenapi.vm_utils") FLAGS = flags.FLAGS flags.DEFINE_string('default_os_type', 'linux', 'Default OS type') +flags.DEFINE_integer('block_device_creation_timeout', 10, + 'time to wait for a block device to be created') XENAPI_POWER_STATE = { 'Halted': power_state.SHUTDOWN, @@ -101,8 +103,8 @@ class VMHelper(HelperBase): 3. Using hardware virtualization """ - instance_type = instance_types.\ - get_instance_type(instance.instance_type) + inst_type_id = instance.instance_type_id + instance_type = instance_types.get_instance_type(inst_type_id) mem = str(long(instance_type['memory_mb']) * 1024 * 1024) vcpus = str(instance_type['vcpus']) rec = { @@ -169,8 +171,8 @@ class VMHelper(HelperBase): @classmethod def ensure_free_mem(cls, session, instance): - instance_type = instance_types.get_instance_type( - instance.instance_type) + inst_type_id = instance.instance_type_id + instance_type = instance_types.get_instance_type(inst_type_id) mem = long(instance_type['memory_mb']) * 1024 * 1024 #get free memory from host host = session.get_xenapi_host() @@ -896,6 +898,16 @@ def remap_vbd_dev(dev): return remapped_dev +def _wait_for_device(dev): + """Wait for device node to appear""" + for i in xrange(0, FLAGS.block_device_creation_timeout): + if os.path.exists('/dev/%s' % dev): + return + time.sleep(1) + + raise StorageError(_('Timeout waiting for device %s to be created') % dev) + + def with_vdi_attached_here(session, vdi_ref, read_only, f): this_vm_ref = get_this_vm_ref(session) vbd_rec = {} @@ -924,6 +936,11 @@ def with_vdi_attached_here(session, vdi_ref, read_only, f): if dev != orig_dev: LOG.debug(_('VBD %(vbd_ref)s plugged into wrong dev, ' 'remapping to %(dev)s') % locals()) + if dev != 'autodetect': + # NOTE(johannes): Unit tests will end up with a device called + # 'autodetect' which obviously won't exist. It's not ideal, + # but the alternatives were much messier + _wait_for_device(dev) return f(dev) finally: LOG.debug(_('Destroying VBD for VDI %s ... '), vdi_ref) @@ -1130,7 +1147,7 @@ def _prepare_injectables(inst, networks_info): 'dns': dns, 'address_v6': ip_v6 and ip_v6['ip'] or '', 'netmask_v6': ip_v6 and ip_v6['netmask'] or '', - 'gateway_v6': ip_v6 and ip_v6['gateway'] or '', + 'gateway_v6': ip_v6 and info['gateway6'] or '', 'use_ipv6': FLAGS.use_ipv6} interfaces_info.append(interface_info) diff --git a/nova/virt/xenapi/vmops.py b/nova/virt/xenapi/vmops.py index c96c35a6e..7c7aa8e98 100644 --- a/nova/virt/xenapi/vmops.py +++ b/nova/virt/xenapi/vmops.py @@ -176,7 +176,7 @@ class VMOps(object): vdi_ref, network_info) self.create_vifs(vm_ref, network_info) - self.inject_network_info(instance, vm_ref, network_info) + self.inject_network_info(instance, network_info, vm_ref) return vm_ref def _spawn(self, instance, vm_ref): @@ -802,8 +802,10 @@ class VMOps(object): instance['id']) networks = db.network_get_all_by_instance(admin_context, instance['id']) - flavor = db.instance_type_get_by_name(admin_context, - instance['instance_type']) + + inst_type = db.instance_type_get_by_id(admin_context, + instance['instance_type_id']) + network_info = [] for network in networks: network_IPs = [ip for ip in IPs if ip.network_id == network.id] @@ -814,12 +816,11 @@ class VMOps(object): "netmask": network["netmask"], "enabled": "1"} - def ip6_dict(ip6): + def ip6_dict(): return { "ip": utils.to_global_ipv6(network['cidr_v6'], instance['mac_address']), "netmask": network['netmask_v6'], - "gateway": network['gateway_v6'], "enabled": "1"} info = { @@ -827,23 +828,41 @@ class VMOps(object): 'gateway': network['gateway'], 'broadcast': network['broadcast'], 'mac': instance.mac_address, - 'rxtx_cap': flavor['rxtx_cap'], + 'rxtx_cap': inst_type['rxtx_cap'], 'dns': [network['dns']], 'ips': [ip_dict(ip) for ip in network_IPs]} if network['cidr_v6']: - info['ip6s'] = [ip6_dict(ip) for ip in network_IPs] + info['ip6s'] = [ip6_dict()] + if network['gateway_v6']: + info['gateway6'] = network['gateway_v6'] network_info.append((network, info)) return network_info - def inject_network_info(self, instance, vm_ref, network_info): + #TODO{tr3buchet) remove this shim with nova-multi-nic + def inject_network_info(self, instance, network_info=None, vm_ref=None): + """ + shim in place which makes inject_network_info work without being + passed network_info. + shim goes away after nova-multi-nic + """ + if not network_info: + network_info = self._get_network_info(instance) + self._inject_network_info(instance, network_info, vm_ref) + + 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 VMHelper.lookup(session, instance.name) will find (ex: rescue) """ logging.debug(_("injecting network info to xs for vm: |%s|"), vm_ref) - # this function raises if vm_ref is not a vm_opaque_ref - self._session.get_xenapi().VM.get_record(vm_ref) + if vm_ref: + # this function raises if vm_ref is not a vm_opaque_ref + self._session.get_xenapi().VM.get_record(vm_ref) + else: + vm_ref = VMHelper.lookup(self._session, instance.name) for (network, info) in network_info: location = 'vm-data/networking/%s' % info['mac'].replace(':', '') @@ -875,8 +894,10 @@ class VMOps(object): VMHelper.create_vif(self._session, vm_ref, network_ref, mac_address, device, rxtx_cap) - def reset_network(self, instance, vm_ref): + def reset_network(self, instance, vm_ref=None): """Creates uuid arg to pass to make_agent_call and calls it.""" + if not vm_ref: + vm_ref = VMHelper.lookup(self._session, instance.name) args = {'id': str(uuid.uuid4())} # TODO(tr3buchet): fix function call after refactor #resp = self._make_agent_call('resetnetwork', instance, '', args) diff --git a/nova/virt/xenapi_conn.py b/nova/virt/xenapi_conn.py index 99fd35c61..0cabccf08 100644 --- a/nova/virt/xenapi_conn.py +++ b/nova/virt/xenapi_conn.py @@ -63,6 +63,7 @@ import xmlrpclib from eventlet import event from eventlet import tpool +from eventlet import timeout from nova import context from nova import db @@ -140,6 +141,9 @@ flags.DEFINE_bool('xenapi_remap_vbd_dev', False, flags.DEFINE_string('xenapi_remap_vbd_dev_prefix', 'sd', 'Specify prefix to remap VBD dev to ' '(ex. /dev/xvdb -> /dev/sdb)') +flags.DEFINE_integer('xenapi_login_timeout', + 10, + 'Timeout in seconds for XenAPI login.') def get_connection(_): @@ -318,7 +322,10 @@ class XenAPISession(object): def __init__(self, url, user, pw): self.XenAPI = self.get_imported_xenapi() self._session = self._create_session(url) - self._session.login_with_password(user, pw) + exception = self.XenAPI.Failure(_("Unable to log in to XenAPI " + "(is the Dom0 disk full?)")) + with timeout.Timeout(FLAGS.xenapi_login_timeout, exception): + self._session.login_with_password(user, pw) self.loop = None def get_imported_xenapi(self): |
