Windows instances will often take a few minutes setting up the image on first boot and then reboot. We should be more patient for those systems as well check if the domid changes so we can send agent requests to the current domid.

1 files changed, 47 insertions, 11 deletions

diff --git a/nova/virt/xenapi/vmops.py b/nova/virt/xenapi/vmops.py
index 2f4286184..53d2d2cec 100644
--- a/nova/virt/xenapi/vmops.py
+++ b/nova/virt/xenapi/vmops.py
@@ -25,6 +25,7 @@ import M2Crypto
 import os
 import pickle
 import subprocess
+import time
 import uuid
 
 from nova import context
@@ -44,7 +45,11 @@ from nova.virt.xenapi.vm_utils import ImageType
 
 XenAPI = None
 LOG = logging.getLogger("nova.virt.xenapi.vmops")
+
 FLAGS = flags.FLAGS
+flags.DEFINE_integer('windows_version_timeout', 300,
+                     'number of seconds to wait for windows agent to be '
+                     'fully operational')
 
 
 def cmp_version(a, b):
@@ -247,7 +252,15 @@ class VMOps(object):
                         'architecture': instance.architecture})
 
         def _check_agent_version():
-            version = self.get_agent_version(instance)
+            if instance.os_type == 'windows':
+                # Windows will generally perform a setup process on first boot
+                # that can take a couple of minutes and then reboot. So we
+                # need to be more patient than normal as well as watch for
+                # domid changes
+                version = self.get_agent_version(instance,
+                                  timeout=FLAGS.windows_version_timeout)
+            else:
+                version = self.get_agent_version(instance)
             if not version:
                 LOG.info(_('No agent version returned by instance'))
                 return
@@ -502,18 +515,41 @@ class VMOps(object):
         task = self._session.call_xenapi('Async.VM.clean_reboot', vm_ref)
         self._session.wait_for_task(task, instance.id)
 
-    def get_agent_version(self, instance):
+    def get_agent_version(self, instance, timeout=None):
         """Get the version of the agent running on the VM instance."""
 
-        # Send the encrypted password
-        transaction_id = str(uuid.uuid4())
-        args = {'id': transaction_id}
-        resp = self._make_agent_call('version', instance, '', args)
-        if resp is None:
-            # No response from the agent
-            return
-        resp_dict = json.loads(resp)
-        return resp_dict['message']
+        def _call():
+            # Send the encrypted password
+            transaction_id = str(uuid.uuid4())
+            args = {'id': transaction_id}
+            resp = self._make_agent_call('version', instance, '', args)
+            if resp is None:
+                # No response from the agent
+                return
+            resp_dict = json.loads(resp)
+            return resp_dict['message']
+
+        if timeout:
+            vm_ref = self._get_vm_opaque_ref(instance)
+            vm_rec = self._session.get_xenapi().VM.get_record(vm_ref)
+
+            domid = vm_rec['domid']
+
+            expiration = time.time() + timeout
+            while time.time() < expiration:
+                ret = _call()
+                if ret:
+                    return ret
+
+                vm_rec = self._session.get_xenapi().VM.get_record(vm_ref)
+                if vm_rec['domid'] != domid:
+                    LOG.info(_('domid changed from %(olddomid)s to '
+                               '%(newdomid)s') % {
+                                   'olddomid': domid,
+                                    'newdomid': vm_rec['domid']})
+                    domid = vm_rec['domid']
+        else:
+            return _call()
 
     def agent_update(self, instance, url, md5sum):
         """Update agent on the VM instance."""