Fix GPT partition tables after agent writes contents

Fixes errors that were being raised upon restarting the agent
directly written out software raid images as the raidset is
restarted for device consistency and partition updates later
on in the code path of deployment.

Story: 2007455
Task: 39187
Change-Id: I9abf51eb77b262932e70329af5ce1593106a3171

ironic_python_agent/extensions/standby.py

@@ -196,6 +196,11 @@ def _write_image(image_info, device):
     totaltime = time.time() - starttime
     LOG.info('Image {} written to device {} in {} seconds'.format(
              image, device, totaltime))
+    try:
+        disk_utils.fix_gpt_partition(device, node_uuid=None)
+    except exception.InstanceDeployFailure:
+        # Note: the catch internal to the helper method logs any errors.
+        pass
     return uuids
 
 
@@ -518,6 +523,12 @@ class StandbyExtension(base.BaseAgentExtension):
                  "seconds".format(device, totaltime))
         # Verify if the checksum of the streamed image is correct
         image_download.verify_image(device)
+        # Fix any gpt partition
+        try:
+            disk_utils.fix_gpt_partition(device, node_uuid=None)
+        except exception.InstanceDeployFailure:
+            # Note: the catch internal to the helper method logs any errors.
+            pass
 
     @base.async_command('cache_image', _validate_image_info)
     def cache_image(self, image_info=None, force=False):

ironic_python_agent/tests/unit/extensions/test_standby.py

@@ -15,6 +15,7 @@
 import os
 import tempfile
 
+from ironic_lib import exception
 import mock
 from oslo_concurrency import processutils
 
@@ -157,9 +158,10 @@ class TestStandbyExtension(base.IronicAgentTest):
         expected_loc = os.path.join(tempfile.gettempdir(), 'fake_id')
         self.assertEqual(expected_loc, location)
 
+    @mock.patch('ironic_lib.disk_utils.fix_gpt_partition', autospec=True)
     @mock.patch('builtins.open', autospec=True)
     @mock.patch('ironic_python_agent.utils.execute', autospec=True)
-    def test_write_image(self, execute_mock, open_mock):
+    def test_write_image(self, execute_mock, open_mock, fix_gpt_mock):
         image_info = _build_fake_image_info()
         device = '/dev/sda'
         location = standby._image_location(image_info)
@@ -169,6 +171,10 @@ class TestStandbyExtension(base.IronicAgentTest):
 
         standby._write_image(image_info, device)
         execute_mock.assert_called_once_with(*command, check_exit_code=[0])
+        fix_gpt_mock.assert_called_once_with(device, node_uuid=None)
+
+        fix_gpt_mock.side_effect = exception.InstanceDeployFailure
+        standby._write_image(image_info, device)
 
         execute_mock.reset_mock()
         execute_mock.return_value = ('', '')
@@ -1085,11 +1091,12 @@ class TestStandbyExtension(base.IronicAgentTest):
         download_mock.assert_called_once_with(image_info)
         write_mock.assert_called_once_with(image_info, device)
 
+    @mock.patch('ironic_lib.disk_utils.fix_gpt_partition', autospec=True)
     @mock.patch('hashlib.md5', autospec=True)
     @mock.patch('builtins.open', autospec=True)
     @mock.patch('requests.get', autospec=True)
     def test_stream_raw_image_onto_device(self, requests_mock, open_mock,
-                                          md5_mock):
+                                          md5_mock, fix_gpt_mock):
         image_info = _build_fake_image_info()
         response = requests_mock.return_value
         response.status_code = 200
@@ -1107,6 +1114,7 @@ class TestStandbyExtension(base.IronicAgentTest):
                                               stream=True, proxies={})
         expected_calls = [mock.call('some'), mock.call('content')]
         file_mock.write.assert_has_calls(expected_calls)
+        fix_gpt_mock.assert_called_once_with('/dev/foo', node_uuid=None)
 
     @mock.patch('hashlib.md5', autospec=True)
     @mock.patch('builtins.open', autospec=True)

releasenotes/notes/executes-gpt-partition-fix-b6156cc16da00dfc.yaml

@@ -0,0 +1,7 @@
+---
+fixes:
+  - |
+    Fixes an issue where secondary GPT partition tables were not being updated
+    after the ``ironic-python-agent`` wrote the disk image to the target disk.
+    The agent now unconditionally attempts to repair the secondary partition
+    table. Previously, software RAID volumes would report errors upon restart.