From 33a84d94eb4d495312e3ba750b194213737a8933 Mon Sep 17 00:00:00 2001
From: Vladyslav Drok <vdrok@mirantis.com>
Date: Tue, 26 Nov 2019 12:51:40 +0100
Subject: [PATCH] Add timeout when querying agent's command statuses

Otherwise the node can get locked up for a significant amount of time.
It conforms to the behaviour of all other (POST) HTTP requests done to
agent.

Story: 2006946
Task: 37626
Change-Id: I968606a300bc43675d7bd07f73de37967ef80c26
---
 ironic/drivers/modules/agent_client.py                 |  2 +-
 ironic/tests/unit/drivers/modules/test_agent_client.py | 10 ++++++++++
 .../get-commands-status-timeout-ecbac91ea149e755.yaml  |  6 ++++++
 3 files changed, 17 insertions(+), 1 deletion(-)
 create mode 100644 releasenotes/notes/get-commands-status-timeout-ecbac91ea149e755.yaml

diff --git a/ironic/drivers/modules/agent_client.py b/ironic/drivers/modules/agent_client.py
index 545a6aea49..0e3fcb57b8 100644
--- a/ironic/drivers/modules/agent_client.py
+++ b/ironic/drivers/modules/agent_client.py
@@ -162,7 +162,7 @@ class AgentClient(object):
         """
         url = self._get_command_url(node)
         LOG.debug('Fetching status of agent commands for node %s', node.uuid)
-        resp = self.session.get(url)
+        resp = self.session.get(url, timeout=CONF.agent.command_timeout)
         result = resp.json()['commands']
         status = '; '.join('%(cmd)s: result "%(res)s", error "%(err)s"' %
                            {'cmd': r.get('command_name'),
diff --git a/ironic/tests/unit/drivers/modules/test_agent_client.py b/ironic/tests/unit/drivers/modules/test_agent_client.py
index 1c563f7fd3..4a5d6f89f3 100644
--- a/ironic/tests/unit/drivers/modules/test_agent_client.py
+++ b/ironic/tests/unit/drivers/modules/test_agent_client.py
@@ -21,10 +21,14 @@ import six
 from six.moves import http_client
 
 from ironic.common import exception
+from ironic import conf
 from ironic.drivers.modules import agent_client
 from ironic.tests import base
 
 
+CONF = conf.CONF
+
+
 class MockResponse(object):
     def __init__(self, text, status_code=http_client.OK):
         assert isinstance(text, six.string_types)
@@ -181,6 +185,12 @@ class TestAgentClient(base.TestCase):
             res.json.return_value = {'commands': []}
             mock_get.return_value = res
             self.assertEqual([], self.client.get_commands_status(self.node))
+            agent_url = self.node.driver_internal_info.get('agent_url')
+            mock_get.assert_called_once_with(
+                '%(agent_url)s/%(api_version)s/commands' % {
+                    'agent_url': agent_url,
+                    'api_version': CONF.agent.agent_api_version},
+                timeout=CONF.agent.command_timeout)
 
     def test_prepare_image(self):
         self.client._command = mock.MagicMock(spec_set=[])
diff --git a/releasenotes/notes/get-commands-status-timeout-ecbac91ea149e755.yaml b/releasenotes/notes/get-commands-status-timeout-ecbac91ea149e755.yaml
new file mode 100644
index 0000000000..4d77dfd92c
--- /dev/null
+++ b/releasenotes/notes/get-commands-status-timeout-ecbac91ea149e755.yaml
@@ -0,0 +1,6 @@
+---
+fixes:
+  - |
+    Add timeout when querying agent for commands status. Without it,
+    node can lock up for a quite long time and ironic will not allow
+    to perform any operations with it.