From d0daaaa320ecb408e721d5e856676a87d7e98ea6 Mon Sep 17 00:00:00 2001 From: Vishvananda Ishaya Date: Wed, 14 Sep 2011 15:28:22 -0700 Subject: write out xml for rescue --- nova/virt/libvirt/connection.py | 52 +++++++++++++++++------------------------ 1 file changed, 21 insertions(+), 31 deletions(-) diff --git a/nova/virt/libvirt/connection.py b/nova/virt/libvirt/connection.py index c7445fe50..56a0f3bb0 100644 --- a/nova/virt/libvirt/connection.py +++ b/nova/virt/libvirt/connection.py @@ -466,22 +466,20 @@ class LibvirtConnection(driver.ComputeDriver): shutil.rmtree(temp_dir) @exception.wrap_exception() - def reboot(self, instance, network_info, regenerate_xml=False): + def reboot(self, instance, network_info, xml=None): """Reboot a virtual machine, given an instance reference. This method actually destroys and re-creates the domain to ensure the reboot happens, as the guest OS cannot ignore this action. """ - # NOTE(vish): this should accept block device info virt_dom = self._conn.lookupByName(instance['name']) # NOTE(itoumsn): Use XML delived from the running instance # instead of using to_xml(instance, network_info). This is almost # the ultimate stupid workaround. - if regenerate_xml: - xml = self.to_xml(instance, network_info, False) - else: + if not xml: xml = virt_dom.XMLDesc(0) + # NOTE(itoumsn): self.shutdown() and wait instead of self.destroy() is # better because we cannot ensure flushing dirty buffers # in the guest OS. But, in case of KVM, shutdown() does not work... @@ -545,7 +543,15 @@ class LibvirtConnection(driver.ComputeDriver): data recovery. """ - self.destroy(instance, network_info, cleanup=False) + + virt_dom = self._conn.lookupByName(instance['name']) + unrescue_xml = virt_dom.XMLDesc(0) + unrescue_xml_path = os.path.join(FLAGS.instances_path, + instance['name'], + 'unrescue.xml') + f = open(unrescue_xml_path, 'w') + f.write(unrescue_xml) + f.close() xml = self.to_xml(instance, network_info, rescue=True) rescue_images = { @@ -555,29 +561,7 @@ class LibvirtConnection(driver.ComputeDriver): } self._create_image(context, instance, xml, '.rescue', rescue_images, network_info=network_info) - self.firewall_driver.setup_basic_filtering(instance, network_info) - self.firewall_driver.prepare_instance_filter(instance, network_info) - self._create_new_domain(xml) - self.firewall_driver.apply_instance_filter(instance, network_info) - - def _wait_for_rescue(): - """Called at an interval until the VM is running again.""" - instance_name = instance['name'] - - try: - state = self.get_info(instance_name)['state'] - except exception.NotFound: - msg = _("During rescue, %s disappeared.") % instance_name - LOG.error(msg) - raise utils.LoopingCallDone - - if state == power_state.RUNNING: - msg = _("Instance %s rescued successfully.") % instance_name - LOG.info(msg) - raise utils.LoopingCallDone - - timer = utils.LoopingCall(_wait_for_rescue) - return timer.start(interval=0.5, now=True) + self.reboot(instance, network_info, xml=xml) @exception.wrap_exception() def unrescue(self, instance, callback, network_info): @@ -587,8 +571,14 @@ class LibvirtConnection(driver.ComputeDriver): simply call reboot. """ - # NOTE(vish): this should accept block device info - self.reboot(instance, network_info, regenerate_xml=True) + unrescue_xml_path = os.path.join(FLAGS.instances_path, + instance['name'], + 'unrescue.xml') + f = open(unrescue_xml_path) + unrescue_xml = f.read() + f.close() + os.remove(unrescue_xml_path) + self.reboot(instance, network_info, xml=unrescue_xml) @exception.wrap_exception() def poll_rescued_instances(self, timeout): -- cgit