Merge "Fix bootloader install issue with MDRAID"

ironic_python_agent/hardware.py

@@ -380,12 +380,24 @@ def list_all_block_devices(block_type='disk',
         # Other possible type values, which we skip recording:
         #   lvm, part, rom, loop
         if devtype != block_type:
-            if (devtype is not None
+            if (devtype is None or ignore_raid):
-                    and any(x in devtype for x in ['raid', 'md'])
+                LOG.debug("Skipping: {!r}".format(line))
-                    and not ignore_raid):
+                continue
+            elif ('raid' in devtype
+                  and block_type in ['raid', 'disk']):
                 LOG.debug(
-                    "TYPE detected to contain 'raid' or 'md', signifying a "
+                    "TYPE detected to contain 'raid', signifying a "
                     "RAID volume. Found: {!r}".format(line))
+            elif (devtype == 'md'
+                  and (block_type == 'part'
+                       or block_type == 'md')):
+                # NOTE(dszumski): Partitions on software RAID devices have type
+                # 'md'. This may also contain RAID devices in a broken state in
+                # rare occasions. See https://review.opendev.org/#/c/670807 for
+                # more detail.
+                LOG.debug(
+                    "TYPE detected to contain 'md', signifying a "
+                    "RAID partition. Found: {!r}".format(line))
             else:
                 LOG.debug(
                     "TYPE did not match. Wanted: {!r} but found: {!r}".format(
@@ -1790,6 +1802,16 @@ class GenericHardwareManager(HardwareManager):
             raid_devices = list_all_block_devices(block_type='raid',
                                                   ignore_raid=False,
                                                   ignore_empty=False)
+            # NOTE(dszumski): Fetch all devices of type 'md'. This
+            # will generally contain partitions on a software RAID
+            # device, but crucially may also contain devices in a
+            # broken state. See https://review.opendev.org/#/c/670807/
+            # for more detail.
+            raid_devices.extend(
+                list_all_block_devices(block_type='md',
+                                       ignore_raid=False,
+                                       ignore_empty=False)
+            )
             return raid_devices
 
         raid_devices = _scan_raids()

ironic_python_agent/tests/unit/test_hardware.py

@@ -159,13 +159,24 @@ BLK_DEVICE_TEMPLATE_SMALL_DEVICES = [
 
 # NOTE(TheJulia): This list intentionally contains duplicates
 # as the code filters them out by kernel device name.
+# NOTE(dszumski): We include some partitions here to verify that
+# they are filtered out when not requested. It is assumed that
+# ROTA has been set to 0 on some software RAID devices for testing
+# purposes. In practice is appears to inherit from the underyling
+# devices, so in this example it would normally be 1.
 RAID_BLK_DEVICE_TEMPLATE = (
     'KNAME="sda" MODEL="DRIVE 0" SIZE="1765517033472" '
     'ROTA="1" TYPE="disk"\n'
+    'KNAME="sda1" MODEL="DRIVE 0" SIZE="107373133824" '
+    'ROTA="1" TYPE="part"\n'
     'KNAME="sdb" MODEL="DRIVE 1" SIZE="1765517033472" '
     'ROTA="1" TYPE="disk"\n'
     'KNAME="sdb" MODEL="DRIVE 1" SIZE="1765517033472" '
     'ROTA="1" TYPE="disk"\n'
+    'KNAME="sdb1" MODEL="DRIVE 1" SIZE="107373133824" '
+    'ROTA="1" TYPE="part"\n'
+    'KNAME="md0p1" MODEL="RAID" SIZE="107236818944" '
+    'ROTA="0" TYPE="md"\n'
     'KNAME="md0" MODEL="RAID" SIZE="1765517033470" '
     'ROTA="0" TYPE="raid1"\n'
     'KNAME="md0" MODEL="RAID" SIZE="1765517033470" '
@@ -3843,9 +3854,11 @@ class TestGenericHardwareManager(base.IronicAgentTest):
 
         hardware.list_all_block_devices.side_effect = [
             [raid_device1, raid_device2],  # list_all_block_devices raid
+            [],  # list_all_block_devices raid (md)
             [sda, sdb, sdc],  # list_all_block_devices disks
             [],  # list_all_block_devices parts
             [],  # list_all_block_devices raid
+            [],  # list_all_block_devices raid (md)
         ]
         mocked_get_component.side_effect = [
             ["/dev/sda1", "/dev/sdb1"],
@@ -3925,13 +3938,14 @@ class TestGenericHardwareManager(base.IronicAgentTest):
         raid_device1_part1 = hardware.BlockDevice('/dev/md0p1', 'RAID-1',
                                                   1073741824, True)
         hardware.list_all_block_devices.side_effect = [
-            [raid_device1_part1],  # list_all_block_devices raid
+            [],  # list_all_block_devices raid
+            [raid_device1_part1],  # list_all_block_devices raid (md)
             [],  # list_all_block_devices disks
             [],  # list_all_block_devices parts
             [],  # list_all_block_devices raid
+            [],  # list_all_block_devices raid (md)
         ]
         mocked_get_component.return_value = []
-
         self.assertIsNone(self.hardware.delete_configuration(self.node, []))
         mocked_execute.assert_has_calls([
             mock.call('mdadm', '--assemble', '--scan', check_exit_code=False),
@@ -3952,12 +3966,15 @@ class TestGenericHardwareManager(base.IronicAgentTest):
 
         hardware.list_all_block_devices.side_effect = [
             [raid_device1],  # list_all_block_devices raid
+            [],  # list_all_block_devices raid (type md)
             [],  # list_all_block_devices disks
             [],  # list_all_block_devices parts
             [raid_device1],  # list_all_block_devices raid
+            [],  # list_all_block_devices raid (type md)
             [],  # list_all_block_devices disks
             [],  # list_all_block_devices parts
             [raid_device1],  # list_all_block_devices raid
+            [],  # list_all_block_devices raid (type md)
         ]
         mocked_get_component.return_value = []
 

releasenotes/notes/fix-bootloader-install-with-mdraid-0a254035df9d0bed.yaml

@@ -0,0 +1,7 @@
+---
+fixes:
+  - |
+    Fixes an issue where the bootloader installation can fail on a software
+    RAID volume when no root_device hint is set.
+    See `Story 2007905
+    <https://storyboard.openstack.org/#!/story/2007905>`_