summaryrefslogtreecommitdiff
path: root/cloudinit
diff options
context:
space:
mode:
Diffstat (limited to 'cloudinit')
-rw-r--r--cloudinit/cmd/main.py3
-rw-r--r--cloudinit/config/cc_mounts.py12
-rw-r--r--cloudinit/sources/DataSourceAzure.py199
-rw-r--r--cloudinit/sources/__init__.py12
-rw-r--r--cloudinit/stages.py7
5 files changed, 135 insertions, 98 deletions
diff --git a/cloudinit/cmd/main.py b/cloudinit/cmd/main.py
index 83eb02c9..fe37075e 100644
--- a/cloudinit/cmd/main.py
+++ b/cloudinit/cmd/main.py
@@ -326,6 +326,9 @@ def main_init(name, args):
util.logexc(LOG, "Failed to re-adjust output redirection!")
logging.setupLogging(mods.cfg)
+ # give the activated datasource a chance to adjust
+ init.activate_datasource()
+
# Stage 10
return (init.datasource, run_module_section(mods, name, name))
diff --git a/cloudinit/config/cc_mounts.py b/cloudinit/config/cc_mounts.py
index dfc4b598..452c9e83 100644
--- a/cloudinit/config/cc_mounts.py
+++ b/cloudinit/config/cc_mounts.py
@@ -312,7 +312,8 @@ def handle_swapcfg(swapcfg):
def handle(_name, cfg, cloud, log, _args):
# fs_spec, fs_file, fs_vfstype, fs_mntops, fs-freq, fs_passno
def_mnt_opts = "defaults,nobootwait"
- if cloud.distro.uses_systemd():
+ uses_systemd = cloud.distro.uses_systemd()
+ if uses_systemd:
def_mnt_opts = "defaults,nofail,x-systemd.requires=cloud-init.service"
defvals = [None, None, "auto", def_mnt_opts, "0", "2"]
@@ -447,7 +448,12 @@ def handle(_name, cfg, cloud, log, _args):
except Exception:
util.logexc(log, "Failed to make '%s' config-mount", d)
+ activate_cmd = ["mount", "-a"]
+ if uses_systemd:
+ activate_cmd = ["systemctl", "daemon-reload"]
+ fmt = "Activate mounts: %s:" + ' '.join(activate_cmd)
try:
- util.subp(("mount", "-a"))
+ util.subp(activate_cmd)
+ LOG.debug(fmt, "PASS")
except util.ProcessExecutionError:
- util.logexc(log, "Activating mounts via 'mount -a' failed")
+ util.logexc(log, fmt, "FAIL")
diff --git a/cloudinit/sources/DataSourceAzure.py b/cloudinit/sources/DataSourceAzure.py
index b802b03e..22f9004a 100644
--- a/cloudinit/sources/DataSourceAzure.py
+++ b/cloudinit/sources/DataSourceAzure.py
@@ -19,7 +19,6 @@
import base64
import contextlib
import crypt
-import fnmatch
from functools import partial
import os
import os.path
@@ -28,7 +27,6 @@ from xml.dom import minidom
import xml.etree.ElementTree as ET
from cloudinit import log as logging
-from cloudinit.settings import PER_ALWAYS
from cloudinit import sources
from cloudinit.sources.helpers.azure import get_metadata_from_fabric
from cloudinit import util
@@ -42,6 +40,9 @@ BOUNCE_COMMAND = [
'sh', '-xc',
"i=$interface; x=0; ifdown $i || x=$?; ifup $i || x=$?; exit $x"
]
+# azure systems will always have a resource disk, and 66-azure-ephemeral.rules
+# ensures that it gets linked to this path.
+RESOURCE_DISK_PATH = '/dev/disk/cloud/azure_resource'
BUILTIN_DS_CONFIG = {
'agent_command': AGENT_START,
@@ -53,7 +54,7 @@ BUILTIN_DS_CONFIG = {
'command': BOUNCE_COMMAND,
'hostname_command': 'hostname',
},
- 'disk_aliases': {'ephemeral0': '/dev/sdb'},
+ 'disk_aliases': {'ephemeral0': RESOURCE_DISK_PATH},
'dhclient_lease_file': '/var/lib/dhcp/dhclient.eth0.leases',
}
@@ -245,15 +246,6 @@ class DataSourceAzureNet(sources.DataSource):
self.metadata['instance-id'] = util.read_dmi_data('system-uuid')
self.metadata.update(fabric_data)
- found_ephemeral = find_fabric_formatted_ephemeral_disk()
- if found_ephemeral:
- self.ds_cfg['disk_aliases']['ephemeral0'] = found_ephemeral
- LOG.debug("using detected ephemeral0 of %s", found_ephemeral)
-
- cc_modules_override = support_new_ephemeral(self.sys_cfg)
- if cc_modules_override:
- self.cfg['cloud_init_modules'] = cc_modules_override
-
return True
def device_name_to_device(self, name):
@@ -266,97 +258,104 @@ class DataSourceAzureNet(sources.DataSource):
# quickly (local check only) if self.instance_id is still valid
return sources.instance_id_matches_system_uuid(self.get_instance_id())
-
-def count_files(mp):
- return len(fnmatch.filter(os.listdir(mp), '*[!cdrom]*'))
+ def activate(self, cfg, is_new_instance):
+ address_ephemeral_resize(is_new_instance=is_new_instance)
+ return
-def find_fabric_formatted_ephemeral_part():
- """
- Locate the first fabric formatted ephemeral device.
- """
- potential_locations = ['/dev/disk/cloud/azure_resource-part1',
- '/dev/disk/azure/resource-part1']
- device_location = None
- for potential_location in potential_locations:
- if os.path.exists(potential_location):
- device_location = potential_location
+def can_dev_be_reformatted(devpath):
+ # determine if the ephemeral block device path devpath
+ # is newly formatted after a resize.
+ 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"])
+
+ # 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 device_location is None:
- LOG.debug("no azure resource disk partition path found")
- return None
- ntfs_devices = util.find_devs_with("TYPE=ntfs")
- real_device = os.path.realpath(device_location)
- if real_device in ntfs_devices:
- return device_location
- LOG.debug("'%s' existed (%s) but was not ntfs formated",
- device_location, real_device)
- return None
-
-
-def find_fabric_formatted_ephemeral_disk():
- """
- Get the ephemeral disk.
- """
- part_dev = find_fabric_formatted_ephemeral_part()
- if part_dev:
- return part_dev.split('-')[0]
- return None
+ if part1path is None:
+ return False, 'device %s was not partitioned' % devpath
-def support_new_ephemeral(cfg):
- """
- Windows Azure makes ephemeral devices ephemeral to boot; a ephemeral device
- may be presented as a fresh device, or not.
+ 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))
+ return False, msg
- Since the knowledge of when a disk is supposed to be plowed under is
- specific to Windows Azure, the logic resides here in the datasource. When a
- new ephemeral device is detected, cloud-init overrides the default
- frequency for both disk-setup and mounts for the current boot only.
- """
- device = find_fabric_formatted_ephemeral_part()
- if not device:
- LOG.debug("no default fabric formated ephemeral0.1 found")
- return None
- LOG.debug("fabric formated ephemeral0.1 device at %s", device)
+ def count_files(mp):
+ ignored = {'dataloss_warning_readme.txt'}
+ return len([f for f in os.listdir(mp) if f.lower() not in ignored])
- file_count = 0
+ bmsg = ('partition 1 (%s -> %s) on device %s was ntfs formatted' %
+ (part1path, real_part1path, devpath))
try:
- file_count = util.mount_cb(device, count_files)
- except Exception:
- return None
- LOG.debug("fabric prepared ephmeral0.1 has %s files on it", file_count)
-
- if file_count >= 1:
- LOG.debug("fabric prepared ephemeral0.1 will be preserved")
- return None
+ file_count = util.mount_cb(part1path, count_files)
+ except util.MountFailedError as e:
+ return False, bmsg + ' but mount of %s failed: %s' % (part1path, e)
+
+ if file_count != 0:
+ return False, bmsg + ' but had %d files on it.' % file_count
+
+ return True, bmsg + ' and had no important files. Safe for reformatting.'
+
+
+def address_ephemeral_resize(devpath=RESOURCE_DISK_PATH, maxwait=120,
+ is_new_instance=False):
+ # wait for ephemeral disk to come up
+ naplen = .2
+ missing = wait_for_files([devpath], maxwait=maxwait, naplen=naplen,
+ log_pre="Azure ephemeral disk: ")
+
+ if missing:
+ LOG.warn("ephemeral device '%s' did not appear after %d seconds.",
+ devpath, maxwait)
+ return
+
+ result = False
+ msg = None
+ if is_new_instance:
+ result, msg = (True, "First instance boot.")
else:
- # if device was already mounted, then we need to unmount it
- # race conditions could allow for a check-then-unmount
- # to have a false positive. so just unmount and then check.
- try:
- util.subp(['umount', device])
- except util.ProcessExecutionError as e:
- if device in util.mounts():
- LOG.warn("Failed to unmount %s, will not reformat.", device)
- LOG.debug("Failed umount: %s", e)
- return None
-
- LOG.debug("cloud-init will format ephemeral0.1 this boot.")
- LOG.debug("setting disk_setup and mounts modules 'always' for this boot")
-
- cc_modules = cfg.get('cloud_init_modules')
- if not cc_modules:
- return None
-
- mod_list = []
- for mod in cc_modules:
- if mod in ("disk_setup", "mounts"):
- mod_list.append([mod, PER_ALWAYS])
- LOG.debug("set module '%s' to 'always' for this boot", mod)
+ result, msg = can_dev_be_reformatted(devpath)
+
+ LOG.debug("reformattable=%s: %s" % (result, msg))
+ if not result:
+ return
+
+ for mod in ['disk_setup', 'mounts']:
+ sempath = '/var/lib/cloud/instance/sem/config_' + mod
+ bmsg = 'Marker "%s" for module "%s"' % (sempath, mod)
+ if os.path.exists(sempath):
+ try:
+ os.unlink(sempath)
+ LOG.debug(bmsg + " removed.")
+ except Exception as e:
+ # python3 throws FileNotFoundError, python2 throws OSError
+ LOG.warn(bmsg + ": remove failed! (%s)" % e)
else:
- mod_list.append(mod)
- return mod_list
+ LOG.debug(bmsg + " did not exist.")
+ return
def perform_hostname_bounce(hostname, cfg, prev_hostname):
@@ -408,15 +407,25 @@ def pubkeys_from_crt_files(flist):
return pubkeys
-def wait_for_files(flist, maxwait=60, naplen=.5):
+def wait_for_files(flist, maxwait=60, naplen=.5, log_pre=""):
need = set(flist)
waited = 0
- while waited < maxwait:
+ while True:
need -= set([f for f in need if os.path.exists(f)])
if len(need) == 0:
+ LOG.debug("%sAll files appeared after %s seconds: %s",
+ log_pre, waited, flist)
return []
+ if waited == 0:
+ LOG.info("%sWaiting up to %s seconds for the following files: %s",
+ log_pre, maxwait, flist)
+ if waited + naplen > maxwait:
+ break
time.sleep(naplen)
waited += naplen
+
+ LOG.warn("%sStill missing files after %s seconds: %s",
+ log_pre, maxwait, need)
return need
diff --git a/cloudinit/sources/__init__.py b/cloudinit/sources/__init__.py
index d1395270..13fb7c62 100644
--- a/cloudinit/sources/__init__.py
+++ b/cloudinit/sources/__init__.py
@@ -261,6 +261,18 @@ class DataSource(object):
def first_instance_boot(self):
return
+ def activate(self, cfg, is_new_instance):
+ """activate(cfg, is_new_instance)
+
+ This is called before the init_modules will be called.
+ The cfg is fully up to date config, it contains a merged view of
+ system config, datasource config, user config, vendor config.
+ It should be used rather than the sys_cfg passed to __init__.
+
+ is_new_instance is a boolean indicating if this is a new instance.
+ """
+ return
+
def normalize_pubkey_data(pubkey_data):
keys = []
diff --git a/cloudinit/stages.py b/cloudinit/stages.py
index 47deac6e..86a13785 100644
--- a/cloudinit/stages.py
+++ b/cloudinit/stages.py
@@ -371,6 +371,13 @@ class Init(object):
self._store_userdata()
self._store_vendordata()
+ def activate_datasource(self):
+ if self.datasource is None:
+ raise RuntimeError("Datasource is None, cannot activate.")
+ self.datasource.activate(cfg=self.cfg,
+ is_new_instance=self.is_new_instance())
+ self._write_to_cache()
+
def _store_userdata(self):
raw_ud = self.datasource.get_userdata_raw()
if raw_ud is None: