Azure: fix reformatting of ephemeral disks on resize to large types.

Large instance types have a different disk format on the newly
partitioned ephemeral drive. So we have to adjust the logic in the
Azure datasource to recognize that a disk with 2 partitions and
an empty ntfs filesystem on the second one is acceptable.

This also adjusts the datasources's builtin fs_setup config to remove
the 'replace_fs' entry. This entry was previously ignored, and confusing.
I've clarified the doc on that also.

LP: #1686514

2 files changed, 63 insertions, 40 deletions

diff --git a/cloudinit/config/cc_disk_setup.py b/cloudinit/config/cc_disk_setup.py
index 6f827ddc..29eb5dd8 100644
--- a/cloudinit/config/cc_disk_setup.py
+++ b/cloudinit/config/cc_disk_setup.py
@@ -68,6 +68,9 @@ specified using ``filesystem``.
     Using ``overwrite: true`` for filesystems is dangerous and can lead to data
     loss, so double check the entry in ``fs_setup``.
 
+.. note::
+    ``replace_fs`` is ignored unless ``partition`` is ``auto`` or ``any``.
+
 **Internal name:** ``cc_disk_setup``
 
 **Module frequency:** per instance
@@ -127,7 +130,7 @@ def handle(_name, cfg, cloud, log, _args):
         log.debug("Partitioning disks: %s", str(disk_setup))
         for disk, definition in disk_setup.items():
             if not isinstance(definition, dict):
-                log.warn("Invalid disk definition for %s" % disk)
+                log.warning("Invalid disk definition for %s" % disk)
                 continue
 
             try:
@@ -144,7 +147,7 @@ def handle(_name, cfg, cloud, log, _args):
         update_fs_setup_devices(fs_setup, cloud.device_name_to_device)
         for definition in fs_setup:
             if not isinstance(definition, dict):
-                log.warn("Invalid file system definition: %s" % definition)
+                log.warning("Invalid file system definition: %s" % definition)
                 continue
 
             try:
@@ -199,8 +202,13 @@ def update_fs_setup_devices(disk_setup, tformer):
             definition['_origname'] = origname
             definition['device'] = tformed
 
-        if part and 'partition' in definition:
-            definition['_partition'] = definition['partition']
+        if part:
+            # In origname with <dev>.N, N overrides 'partition' key.
+            if 'partition' in definition:
+                LOG.warning("Partition '%s' from dotted device name '%s' "
+                            "overrides 'partition' key in %s", part, origname,
+                            definition)
+                definition['_partition'] = definition['partition']
             definition['partition'] = part
 
 
@@ -849,7 +857,8 @@ def mkfs(fs_cfg):
         # Check to see if the fs already exists
         LOG.debug("Checking device %s", device)
         check_label, check_fstype, _ = check_fs(device)
-        LOG.debug("Device %s has %s %s", device, check_label, check_fstype)
+        LOG.debug("Device '%s' has check_label='%s' check_fstype=%s",
+                  device, check_label, check_fstype)
 
         if check_label == label and check_fstype == fs_type:
             LOG.debug("Existing file system found at %s", device)
diff --git a/cloudinit/sources/DataSourceAzure.py b/cloudinit/sources/DataSourceAzure.py
index 5254e18a..44857c09 100644
--- a/cloudinit/sources/DataSourceAzure.py
+++ b/cloudinit/sources/DataSourceAzure.py
@@ -196,8 +196,7 @@ BUILTIN_CLOUD_CONFIG = {
                        'overwrite': True},
     },
     'fs_setup': [{'filesystem': DEFAULT_FS,
-                  'device': 'ephemeral0.1',
-                  'replace_fs': 'ntfs'}],
+                  'device': 'ephemeral0.1'}],
 }
 
 DS_CFG_PATH = ['datasource', DS_NAME]
@@ -413,56 +412,71 @@ class DataSourceAzureNet(sources.DataSource):
         return
 
 
+def _partitions_on_device(devpath, maxnum=16):
+    # return a list of tuples (ptnum, path) for each part on devpath
+    for suff in ("-part", "p", ""):
+        found = []
+        for pnum in range(1, maxnum):
+            ppath = devpath + suff + str(pnum)
+            if os.path.exists(ppath):
+                found.append((pnum, os.path.realpath(ppath)))
+        if found:
+            return found
+    return []
+
+
+def _has_ntfs_filesystem(devpath):
+    ntfs_devices = util.find_devs_with("TYPE=ntfs", no_cache=True)
+    LOG.debug('ntfs_devices found = %s', ntfs_devices)
+    return os.path.realpath(devpath) in ntfs_devices
+
+
 def can_dev_be_reformatted(devpath):
-    # determine if the ephemeral block device path devpath
-    # is newly formatted after a resize.
+    """Determine if block device devpath is newly formatted ephemeral.
+
+    A newly formatted disk will:
+      a.) have a partition table (dos or gpt)
+      b.) have 1 partition that is ntfs formatted, or
+          have 2 partitions with the second partition ntfs formatted.
+          (larger instances with >2TB ephemeral disk have gpt, and will
+           have a microsoft reserved partition as part 1.  LP: #1686514)
+      c.) the ntfs partition will have no files other than possibly
+          'dataloss_warning_readme.txt'"""
     if not os.path.exists(devpath):
         return False, 'device %s does not exist' % devpath
 
-    realpath = os.path.realpath(devpath)
-    LOG.debug('Resolving realpath of %s -> %s', devpath, realpath)
-
-    # it is possible that the block device might exist, but the kernel
-    # have not yet read the partition table and sent events.  we udevadm settle
-    # to hope to resolve that.  Better here would probably be to test and see,
-    # and then settle if we didn't find anything and try again.
-    if util.which("udevadm"):
-        util.subp(["udevadm", "settle"])
+    LOG.debug('Resolving realpath of %s -> %s', devpath,
+              os.path.realpath(devpath))
 
     # devpath of /dev/sd[a-z] or /dev/disk/cloud/azure_resource
     # where partitions are "<devpath>1" or "<devpath>-part1" or "<devpath>p1"
-    part1path = None
-    for suff in ("-part", "p", ""):
-        cand = devpath + suff + "1"
-        if os.path.exists(cand):
-            if os.path.exists(devpath + suff + "2"):
-                msg = ('device %s had more than 1 partition: %s, %s' %
-                       devpath, cand, devpath + suff + "2")
-                return False, msg
-            part1path = cand
-            break
-
-    if part1path is None:
+    partitions = _partitions_on_device(devpath)
+    if len(partitions) == 0:
         return False, 'device %s was not partitioned' % devpath
+    elif len(partitions) > 2:
+        msg = ('device %s had 3 or more partitions: %s' %
+               (devpath, ' '.join([p[1] for p in partitions])))
+        return False, msg
+    elif len(partitions) == 2:
+        cand_part, cand_path = partitions[1]
+    else:
+        cand_part, cand_path = partitions[0]
 
-    real_part1path = os.path.realpath(part1path)
-    ntfs_devices = util.find_devs_with("TYPE=ntfs", no_cache=True)
-    LOG.debug('ntfs_devices found = %s', ntfs_devices)
-    if real_part1path not in ntfs_devices:
-        msg = ('partition 1 (%s -> %s) on device %s was not ntfs formatted' %
-               (part1path, real_part1path, devpath))
+    if not _has_ntfs_filesystem(cand_path):
+        msg = ('partition %s (%s) on device %s was not ntfs formatted' %
+               (cand_part, cand_path, devpath))
         return False, msg
 
     def count_files(mp):
         ignored = set(['dataloss_warning_readme.txt'])
         return len([f for f in os.listdir(mp) if f.lower() not in ignored])
 
-    bmsg = ('partition 1 (%s -> %s) on device %s was ntfs formatted' %
-            (part1path, real_part1path, devpath))
+    bmsg = ('partition %s (%s) on device %s was ntfs formatted' %
+            (cand_part, cand_path, devpath))
     try:
-        file_count = util.mount_cb(part1path, count_files)
+        file_count = util.mount_cb(cand_path, count_files)
     except util.MountFailedError as e:
-        return False, bmsg + ' but mount of %s failed: %s' % (part1path, e)
+        return False, bmsg + ' but mount of %s failed: %s' % (cand_part, e)
 
     if file_count != 0:
         return False, bmsg + ' but had %d files on it.' % file_count