Azure: Retry dhcp on timeouts when polling reprovisiondata (#998)

In the nic attach path, we skip doing dhcp since we already did it
when bringing the interface up. However when polling for
reprovisiondata, it is possible for the request to timeout due to
platform issues. In those cases we still need to do dhcp and try again
since we tear down the context. We can only skip the first dhcp
attempt.

2 files changed, 38 insertions, 0 deletions

diff --git a/cloudinit/sources/DataSourceAzure.py b/cloudinit/sources/DataSourceAzure.py
index fddfe363..caffa944 100755
--- a/cloudinit/sources/DataSourceAzure.py
+++ b/cloudinit/sources/DataSourceAzure.py
@@ -1317,6 +1317,10 @@ class DataSourceAzure(sources.DataSource):
             except UrlError:
                 # Teardown our EphemeralDHCPv4 context on failure as we retry
                 self._ephemeral_dhcp_ctx.clean_network()
+
+                # Also reset this flag which determines if we should do dhcp
+                # during retries.
+                is_ephemeral_ctx_present = False
             finally:
                 if nl_sock:
                     nl_sock.close()
diff --git a/tests/unittests/test_datasource/test_azure.py b/tests/unittests/test_datasource/test_azure.py
index 851cf82e..a4296bf6 100644
--- a/tests/unittests/test_datasource/test_azure.py
+++ b/tests/unittests/test_datasource/test_azure.py
@@ -3055,6 +3055,40 @@ class TestPreprovisioningPollIMDS(CiTestCase):
         self.assertEqual(0, m_dhcp.call_count)
         self.assertEqual(0, m_media_switch.call_count)
 
+    @mock.patch('os.path.isfile')
+    @mock.patch(MOCKPATH + 'EphemeralDHCPv4')
+    def test_poll_imds_does_dhcp_on_retries_if_ctx_present(
+            self, m_ephemeral_dhcpv4, m_isfile, report_ready_func, m_request,
+            m_media_switch, m_dhcp, m_net):
+        """The poll_imds function should reuse the dhcp ctx if it is already
+           present. This happens when we wait for nic to be hot-attached before
+           polling for reprovisiondata. Note that if this ctx is set when
+           _poll_imds is called, then it is not expected to be waiting for
+           media_disconnect_connect either."""
+
+        tries = 0
+
+        def fake_timeout_once(**kwargs):
+            nonlocal tries
+            tries += 1
+            if tries == 1:
+                raise requests.Timeout('Fake connection timeout')
+            return mock.MagicMock(status_code=200, text="good", content="good")
+
+        m_request.side_effect = fake_timeout_once
+        report_file = self.tmp_path('report_marker', self.tmp)
+        m_isfile.return_value = True
+        dsa = dsaz.DataSourceAzure({}, distro=None, paths=self.paths)
+        with mock.patch(MOCKPATH + 'REPORTED_READY_MARKER_FILE', report_file),\
+                mock.patch.object(dsa, '_ephemeral_dhcp_ctx') as m_dhcp_ctx:
+            m_dhcp_ctx.obtain_lease.return_value = "Dummy lease"
+            dsa._ephemeral_dhcp_ctx = m_dhcp_ctx
+            dsa._poll_imds()
+            self.assertEqual(1, m_dhcp_ctx.clean_network.call_count)
+        self.assertEqual(1, m_ephemeral_dhcpv4.call_count)
+        self.assertEqual(0, m_media_switch.call_count)
+        self.assertEqual(2, m_request.call_count)
+
     def test_does_not_poll_imds_report_ready_when_marker_file_exists(
             self, m_report_ready, m_request, m_media_switch, m_dhcp, m_net):
         """poll_imds should not call report ready when the reported ready