Burn-in: Add memory step

Add a clean step for memory burn-in via stress-ng. Get basic
run parameters from the node's driver_info.

Story: #2007523
Task: #42383

Change-Id: I33a83968c9f87cf795ec7ec922bce98b52c5181c

6 files changed, 99 insertions, 0 deletions

diff --git a/doc/source/admin/hardware_managers.rst b/doc/source/admin/hardware_managers.rst
index 51a6e3e3..2144cf9f 100644
--- a/doc/source/admin/hardware_managers.rst
+++ b/doc/source/admin/hardware_managers.rst
@@ -77,6 +77,9 @@ Clean steps
 ``deploy.burnin_cpu``
     Stress-test the CPUs of a node via stress-ng for a configurable
     amount of time. Disabled by default.
+``deploy.burnin_memory``
+    Stress-test the memory of a node via stress-ng for a configurable
+    amount of time. Disabled by default.
 ``deploy.erase_devices``
     Securely erases all information from all recognized disk devices.
     Relatively fast when secure ATA erase is available, otherwise can take
diff --git a/ironic_python_agent/burnin.py b/ironic_python_agent/burnin.py
index 5a9275e2..bd654547 100644
--- a/ironic_python_agent/burnin.py
+++ b/ironic_python_agent/burnin.py
@@ -46,3 +46,35 @@ def stress_ng_cpu(node):
                      {'err': e})
         LOG.error(error_msg)
         raise errors.CommandExecutionError(error_msg)
+
+
+def stress_ng_vm(node):
+    """Burn-in the memory with the vm stressor in stress-ng
+
+    Run stress-ng with a configurable number of workers on
+    a configurable amount of the available memory for
+    a configurable amount of time. Without config use
+    as many workers as CPUs, 98% of the memory and stress
+    it for 24 hours.
+
+    :param node: Ironic node object
+    :raises: CommandExecutionError if the execution of stress-ng fails.
+    """
+    info = node.get('driver_info', {})
+    vm = info.get('agent_burnin_vm_vm', 0)
+    vm_bytes = info.get('agent_burnin_vm_vm-bytes', '98%')
+    timeout = info.get('agent_burnin_vm_timeout', 86400)
+
+    args = ('stress-ng', '--vm', vm, '--vm-bytes', vm_bytes,
+            '--timeout', timeout, '--metrics-brief')
+    LOG.debug('Burn-in stress_ng_vm command: %s', args)
+
+    try:
+        _, err = utils.execute(*args)
+        # stress-ng reports on stderr only
+        LOG.info(err)
+    except (processutils.ProcessExecutionError, OSError) as e:
+        error_msg = ("stress-ng (vm) failed with error %(err)s",
+                     {'err': e})
+        LOG.error(error_msg)
+        raise errors.CommandExecutionError(error_msg)
diff --git a/ironic_python_agent/hardware.py b/ironic_python_agent/hardware.py
index 97c45449..0180adae 100644
--- a/ironic_python_agent/hardware.py
+++ b/ironic_python_agent/hardware.py
@@ -1402,6 +1402,14 @@ class GenericHardwareManager(HardwareManager):
         """
         burnin.stress_ng_cpu(node)
 
+    def burnin_memory(self, node, ports):
+        """Burn-in the memory
+
+        :param node: Ironic node object
+        :param ports: list of Ironic port objects
+        """
+        burnin.stress_ng_vm(node)
+
     def _shred_block_device(self, node, block_device):
         """Erase a block device using shred.
 
@@ -1882,6 +1890,13 @@ class GenericHardwareManager(HardwareManager):
                 'reboot_requested': False,
                 'abortable': True
             },
+            {
+                'step': 'burnin_memory',
+                'priority': 0,
+                'interface': 'deploy',
+                'reboot_requested': False,
+                'abortable': True
+            },
         ]
 
     def get_deploy_steps(self, node, ports):
diff --git a/ironic_python_agent/tests/unit/test_burnin.py b/ironic_python_agent/tests/unit/test_burnin.py
index d8339b6c..7f411b9f 100644
--- a/ironic_python_agent/tests/unit/test_burnin.py
+++ b/ironic_python_agent/tests/unit/test_burnin.py
@@ -54,3 +54,38 @@ class TestBurnin(base.IronicAgentTest):
 
         self.assertRaises(errors.CommandExecutionError,
                           burnin.stress_ng_cpu, node)
+
+    def test_stress_ng_vm_default(self, mock_execute):
+
+        node = {'driver_info': {}}
+        mock_execute.return_value = (['out', 'err'])
+
+        burnin.stress_ng_vm(node)
+
+        mock_execute.assert_called_once_with(
+            'stress-ng', '--vm', 0, '--vm-bytes', '98%',
+            '--timeout', 86400, '--metrics-brief')
+
+    def test_stress_ng_vm_non_default(self, mock_execute):
+
+        node = {'driver_info': {'agent_burnin_vm_vm': 2,
+                                'agent_burnin_vm_vm-bytes': '25%',
+                                'agent_burnin_vm_timeout': 120}}
+        mock_execute.return_value = (['out', 'err'])
+
+        burnin.stress_ng_vm(node)
+
+        mock_execute.assert_called_once_with(
+            'stress-ng', '--vm', 2, '--vm-bytes', '25%',
+            '--timeout', 120, '--metrics-brief')
+
+    def test_stress_ng_vm_no_stress_ng(self, mock_execute):
+
+        node = {'driver_info': {}}
+        mock_execute.side_effect = (['out', 'err'],
+                                    processutils.ProcessExecutionError())
+
+        burnin.stress_ng_vm(node)
+
+        self.assertRaises(errors.CommandExecutionError,
+                          burnin.stress_ng_vm, node)
diff --git a/ironic_python_agent/tests/unit/test_hardware.py b/ironic_python_agent/tests/unit/test_hardware.py
index 019a14c0..5884344c 100644
--- a/ironic_python_agent/tests/unit/test_hardware.py
+++ b/ironic_python_agent/tests/unit/test_hardware.py
@@ -156,6 +156,13 @@ class TestGenericHardwareManager(base.IronicAgentTest):
                 'interface': 'deploy',
                 'reboot_requested': False,
                 'abortable': True
+            },
+            {
+                'step': 'burnin_memory',
+                'priority': 0,
+                'interface': 'deploy',
+                'reboot_requested': False,
+                'abortable': True
             }
         ]
         clean_steps = self.hardware.get_clean_steps(self.node, [])
diff --git a/releasenotes/notes/add_burnin_memory-4099ca42bd3b99db.yaml b/releasenotes/notes/add_burnin_memory-4099ca42bd3b99db.yaml
new file mode 100644
index 00000000..8aeb854e
--- /dev/null
+++ b/releasenotes/notes/add_burnin_memory-4099ca42bd3b99db.yaml
@@ -0,0 +1,7 @@
+---
+features:
+  - |
+    Adds a burn-in cleaning step 'burnin_memory' to stress test memory for a
+    configurable amount of time with stress-ng. To use this step, stress-ng
+    needs to be installed on the RAM disk.
+