charm-swift-storage/lib/swift_storage_utils.py
James Page 591c084532 Add support for block device encryption
Add new secrets-storage relation to vault, supporting the
use of block device encryption using dm-crypt/vaultlocker.

Prepared devices are now recorded in the local unit kv
store; this information is used to provide a list of
configured block devices to the swift-proxy charm,
rather than the previous best guess provided by
determine_block_devices.  This allows us to use the
dm-crypt device name, rather than the underlying
block device.

Encrypted block devices are unlocked on boot using
vaultlocker-decrypt systemd units (enabled by vaultlocker);
/etc/fstab entries for such devices make use of a
x-systemd.requires option to ensure that the block device
is unlocked prior to attempting to mount it.

Add new storage binding to allow charm to be used with
Juju storage.

Add new ephemeral-unmount configuration option to allow
cloud ephemeral storage to be used for testing purposes;
update functional testing to use this option.  The behaviour
of 'overwrite' was changed to accomodate the use of
encrypted block devices.

Change-Id: I9b3f8cd2de412ee96e0139dba4d4abdf998ecaf2
2018-05-08 12:52:37 +01:00

710 lines
23 KiB
Python

import json
import os
import re
import subprocess
import shutil
import tempfile
import uuid
from subprocess import check_call, call, CalledProcessError, check_output
# Stuff copied from cinder py charm, needs to go somewhere
# common.
from misc_utils import (
ensure_block_device,
clean_storage,
is_paused
)
from swift_storage_context import (
SwiftStorageContext,
SwiftStorageServerContext,
RsyncContext,
)
from charmhelpers.fetch import (
apt_upgrade,
apt_update
)
from charmhelpers.core.unitdata import (
Storage as KVStore,
)
import charmhelpers.core.fstab
from charmhelpers.core.host import (
mkdir,
mount,
fstab_add,
service_restart,
lsb_release,
CompareHostReleases,
)
from charmhelpers.core.hookenv import (
config,
log,
DEBUG,
INFO,
WARNING,
ERROR,
unit_private_ip,
local_unit,
relation_get,
relation_ids,
iter_units_for_relation_name,
ingress_address,
storage_list,
storage_get,
)
from charmhelpers.contrib.network import ufw
from charmhelpers.contrib.network.ip import get_host_ip
from charmhelpers.contrib.storage.linux.utils import (
is_block_device,
is_device_mounted,
mkfs_xfs,
)
from charmhelpers.contrib.storage.linux.loopback import (
is_mapped_loopback_device,
)
from charmhelpers.contrib.openstack.utils import (
configure_installation_source,
get_os_codename_install_source,
get_os_codename_package,
save_script_rc as _save_script_rc,
)
from charmhelpers.contrib.openstack import (
templating,
context
)
from charmhelpers.core.decorators import (
retry_on_exception,
)
import charmhelpers.contrib.openstack.vaultlocker as vaultlocker
from charmhelpers.core.unitdata import kv
PACKAGES = [
'swift', 'swift-account', 'swift-container', 'swift-object',
'xfsprogs', 'gdisk', 'lvm2', 'python-jinja2', 'python-psutil',
'ufw',
]
VERSION_PACKAGE = 'swift-account'
TEMPLATES = 'templates/'
REQUIRED_INTERFACES = {
'proxy': ['swift-storage'],
}
ACCOUNT_SVCS = [
'swift-account', 'swift-account-auditor',
'swift-account-reaper', 'swift-account-replicator'
]
CONTAINER_SVCS = [
'swift-container', 'swift-container-auditor',
'swift-container-updater', 'swift-container-replicator',
'swift-container-sync'
]
OBJECT_SVCS = [
'swift-object', 'swift-object-auditor',
'swift-object-updater', 'swift-object-replicator'
]
SWIFT_SVCS = ACCOUNT_SVCS + CONTAINER_SVCS + OBJECT_SVCS
RESTART_MAP = {
'/etc/rsync-juju.d/050-swift-storage.conf': ['rsync'],
'/etc/swift/account-server.conf': ACCOUNT_SVCS,
'/etc/swift/container-server.conf': CONTAINER_SVCS,
'/etc/swift/object-server.conf': OBJECT_SVCS,
'/etc/swift/swift.conf': ACCOUNT_SVCS + CONTAINER_SVCS + OBJECT_SVCS
}
SWIFT_CONF_DIR = '/etc/swift'
SWIFT_RING_EXT = 'ring.gz'
FIRST = 1
# NOTE(hopem): we intentionally place this database outside of unit context so
# that if the unit, service or even entire environment is
# destroyed, there will still be a record of what devices were in
# use so that when the swift charm next executes, already used
# devices will not be unintentionally reformatted. If devices are
# to be recycled, they will need to be manually removed from this
# database.
# FIXME: add charm support for removing devices (see LP: #1448190)
KV_DB_PATH = '/var/lib/juju/swift_storage/charm_kvdata.db'
def ensure_swift_directories():
'''
Ensure all directories required for a swift storage node exist with
correct permissions.
'''
dirs = [
SWIFT_CONF_DIR,
'/var/cache/swift',
'/srv/node',
]
[mkdir(d, owner='swift', group='swift') for d in dirs
if not os.path.isdir(d)]
def register_configs():
release = get_os_codename_package('python-swift', fatal=False) or 'essex'
configs = templating.OSConfigRenderer(templates_dir=TEMPLATES,
openstack_release=release)
configs.register('/etc/swift/swift.conf',
[SwiftStorageContext()])
configs.register('/etc/rsync-juju.d/050-swift-storage.conf',
[RsyncContext(), SwiftStorageServerContext()])
# NOTE: add VaultKVContext so interface status can be assessed
for server in ['account', 'object', 'container']:
configs.register('/etc/swift/%s-server.conf' % server,
[SwiftStorageServerContext(),
context.BindHostContext(),
context.WorkerConfigContext(),
vaultlocker.VaultKVContext(
vaultlocker.VAULTLOCKER_BACKEND)]),
return configs
def swift_init(target, action, fatal=False):
'''
Call swift-init on a specific target with given action, potentially
raising exception.
'''
cmd = ['swift-init', target, action]
if fatal:
return check_call(cmd)
return call(cmd)
def do_openstack_upgrade(configs):
new_src = config('openstack-origin')
new_os_rel = get_os_codename_install_source(new_src)
log('Performing OpenStack upgrade to %s.' % (new_os_rel))
configure_installation_source(new_src)
dpkg_opts = [
'--option', 'Dpkg::Options::=--force-confnew',
'--option', 'Dpkg::Options::=--force-confdef',
]
apt_update()
apt_upgrade(options=dpkg_opts, fatal=True, dist=True)
configs.set_release(openstack_release=new_os_rel)
configs.write_all()
if not is_paused():
for service in SWIFT_SVCS:
service_restart(service)
def _is_storage_ready(partition):
"""
A small helper to determine if a given device is suitabe to be used as
a storage device.
"""
return is_block_device(partition) and not is_device_mounted(partition)
def get_mount_point(device):
mnt_point = None
try:
out = check_output(['findmnt', device])
mnt_points = []
for line in out.split('\n'):
if line and not line.startswith('TARGET'):
mnt_points.append(line.split()[0])
if len(mnt_points) > 1:
log('Device {} mounted in multiple times, ignoring'.format(device))
else:
mnt_point = mnt_points[0]
except CalledProcessError:
# findmnt returns non-zero rc if dev not mounted
pass
return mnt_point
def find_block_devices(include_mounted=False):
found = []
incl = ['sd[a-z]', 'vd[a-z]', 'cciss\/c[0-9]d[0-9]']
with open('/proc/partitions') as proc:
partitions = [p.split() for p in proc.readlines()[2:]]
for partition in [p[3] for p in partitions if p]:
for inc in incl:
_re = re.compile(r'^(%s)$' % inc)
if _re.match(partition):
found.append(os.path.join('/dev', partition))
if include_mounted:
devs = [f for f in found if is_block_device(f)]
else:
devs = [f for f in found if _is_storage_ready(f)]
return devs
def guess_block_devices():
bdevs = find_block_devices(include_mounted=True)
gdevs = []
for dev in bdevs:
if is_device_mounted(dev):
mnt_point = get_mount_point(dev)
if mnt_point and mnt_point.startswith('/srv/node'):
gdevs.append(dev)
else:
gdevs.append(dev)
return gdevs
def determine_block_devices():
block_device = config('block-device')
if not block_device or block_device.lower() == 'none':
log("No storage devices specified in 'block_device' config",
level=ERROR)
return None
if block_device == 'guess':
bdevs = guess_block_devices()
else:
bdevs = block_device.split(' ')
# List storage instances for the 'block-devices'
# store declared for this charm too, and add
# their block device paths to the list.
storage_ids = storage_list('block-devices')
bdevs.extend((storage_get('location', s) for s in storage_ids))
bdevs = list(set(bdevs))
# attempt to ensure block devices, but filter out missing devs
_none = ['None', 'none']
valid_bdevs = \
[x for x in map(ensure_block_device, bdevs) if str(x).lower() not in
_none]
log('Valid ensured block devices: %s' % valid_bdevs)
return valid_bdevs
def devstore_safe_load(devstore):
"""Attempt to decode json data and return None if an error occurs while
also printing a log.
"""
if not devstore:
return None
try:
return json.loads(devstore)
except ValueError:
log("Unable to decode JSON devstore", level=DEBUG)
return None
def is_device_in_ring(dev, skip_rel_check=False, ignore_deactivated=True):
"""Check if device has been added to the ring.
First check local KV store then check storage rel with proxy.
"""
d = os.path.dirname(KV_DB_PATH)
if not os.path.isdir(d):
mkdir(d)
log("Device '%s' does not appear to be in use by Swift" % (dev),
level=INFO)
return False
# First check local KV store
kvstore = KVStore(KV_DB_PATH)
devstore = devstore_safe_load(kvstore.get(key='devices'))
kvstore.close()
deactivated = []
if devstore:
blk_uuid = get_device_blkid("/dev/%s" % (dev))
env_uuid = os.environ.get('JUJU_ENV_UUID',
os.environ.get('JUJU_MODEL_UUID'))
masterkey = "%s@%s" % (dev, env_uuid)
if (masterkey in devstore and
devstore[masterkey].get('blkid') == blk_uuid and
devstore[masterkey].get('status') == 'active'):
log("Device '%s' appears to be in use by Swift (found in local "
"devstore)" % (dev), level=INFO)
return True
for key, val in devstore.iteritems():
if key != masterkey and val.get('blkid') == blk_uuid:
log("Device '%s' appears to be in use by Swift (found in "
"local devstore) but has a different "
"JUJU_[ENV|MODEL]_UUID (current=%s, expected=%s). "
"This could indicate that the device was added as part of "
"a previous deployment and will require manual removal or "
"updating if it needs to be reformatted."
% (dev, key, masterkey), level=INFO)
return True
if ignore_deactivated:
deactivated = [k == masterkey and v.get('blkid') == blk_uuid and
v.get('status') != 'active'
for k, v in devstore.iteritems()]
if skip_rel_check:
log("Device '%s' does not appear to be in use by swift (searched "
"local devstore only)" % (dev), level=INFO)
return False
# Then check swift-storage relation with proxy
for rid in relation_ids('swift-storage'):
devstore = relation_get(attribute='device', rid=rid, unit=local_unit())
if devstore and dev in devstore.split(':'):
if not ignore_deactivated or dev not in deactivated:
log("Device '%s' appears to be in use by swift (found on "
"proxy relation) but was not found in local devstore so "
"will be added to the cache" % (dev), level=INFO)
remember_devices([dev])
return True
log("Device '%s' does not appear to be in use by swift (searched local "
"devstore and proxy relation)" % (dev), level=INFO)
return False
def get_device_blkid(dev):
"""Try to get the fs uuid of the provided device.
If this is called for a new unformatted device we expect blkid to fail
hence return None to indicate the device is not in use.
:param dev: block device path
:returns: UUID of device if found else None
"""
try:
blk_uuid = subprocess.check_output(['blkid', '-s', 'UUID', dev])
except CalledProcessError:
# If the device has not be used or formatted yet we expect this to
# fail.
return None
blk_uuid = re.match(r'^%s:\s+UUID="(.+)"$' % (dev), blk_uuid.strip())
if blk_uuid:
return blk_uuid.group(1)
else:
log("Failed to obtain device UUID for device '%s' - returning None" %
dev, level=WARNING)
return None
def remember_devices(devs):
"""Add device to local store of ringed devices."""
d = os.path.dirname(KV_DB_PATH)
if not os.path.isdir(d):
mkdir(d)
kvstore = KVStore(KV_DB_PATH)
devstore = devstore_safe_load(kvstore.get(key='devices')) or {}
env_uuid = os.environ.get('JUJU_ENV_UUID',
os.environ.get('JUJU_MODEL_UUID'))
for dev in devs:
blk_uuid = get_device_blkid("/dev/%s" % (dev))
key = "%s@%s" % (dev, env_uuid)
if key in devstore and devstore[key].get('blkid') == blk_uuid:
log("Device '%s' already in devstore (status:%s)" %
(dev, devstore[key].get('status')), level=DEBUG)
else:
existing = [(k, v) for k, v in devstore.iteritems()
if v.get('blkid') == blk_uuid and
re.match("^(.+)@(.+)$", k).group(1) == dev]
if existing:
log("Device '%s' already in devstore but has a different "
"JUJU_[ENV|MODEL]_UUID (%s)" %
(dev, re.match(".+@(.+)$", existing[0][0]).group(1)),
level=WARNING)
else:
log("Adding device '%s' with blkid='%s' to devstore" %
(dev, blk_uuid),
level=DEBUG)
devstore[key] = {'blkid': blk_uuid, 'status': 'active'}
if devstore:
kvstore.set(key='devices', value=json.dumps(devstore))
kvstore.flush()
kvstore.close()
def ensure_devs_tracked():
for rid in relation_ids('swift-storage'):
devs = relation_get(attribute='device', rid=rid, unit=local_unit())
if devs:
for dev in devs.split(':'):
# this will migrate if not already in the local store
is_device_in_ring(dev, skip_rel_check=True)
def setup_storage(encrypt=False):
# Preflight check vault relation if encryption is enabled
vault_kv = vaultlocker.VaultKVContext(vaultlocker.VAULTLOCKER_BACKEND)
context = vault_kv()
if encrypt and not vault_kv.complete:
log("Encryption requested but vault relation not complete",
level=DEBUG)
return
elif encrypt and vault_kv.complete:
# NOTE: only write vaultlocker configuration once relation is complete
# otherwise we run the chance of an empty configuration file
# being installed on a machine with other vaultlocker based
# services
vaultlocker.write_vaultlocker_conf(context, priority=90)
# Ensure /srv/node exists just in case no disks
# are detected and used.
mkdir(os.path.join('/srv', 'node'),
owner='swift', group='swift',
perms=0o755)
reformat = str(config('overwrite')).lower() == "true"
db = kv()
prepared_devices = db.get('prepared-devices', [])
for dev in determine_block_devices():
if dev in prepared_devices:
log('Device {} already processed by charm,'
' skipping'.format(dev))
continue
if is_device_in_ring(os.path.basename(dev)):
log("Device '%s' already in the ring - ignoring" % (dev))
# NOTE: record existing use of device dealing with
# upgrades from older versions of charms without
# this feature
prepared_devices.append(dev)
db.set('prepared-devices', prepared_devices)
db.flush()
continue
# NOTE: this deals with a dm-crypt'ed block device already in
# use
if is_device_mounted(dev):
log("Device '{}' is already mounted, ignoring".format(dev))
continue
if reformat:
clean_storage(dev)
loopback_device = is_mapped_loopback_device(dev)
options = None
if encrypt and not loopback_device:
dev_uuid = str(uuid.uuid4())
check_call(['vaultlocker', 'encrypt',
'--uuid', dev_uuid,
dev])
dev = '/dev/mapper/crypt-{}'.format(dev_uuid)
options = ','.join([
"defaults",
"nofail",
("x-systemd.requires="
"vaultlocker-decrypt@{uuid}.service".format(uuid=dev_uuid)),
"comment=vaultlocker",
])
try:
# If not cleaned and in use, mkfs should fail.
mkfs_xfs(dev, force=reformat)
except subprocess.CalledProcessError as exc:
# This is expected is a formatted device is provided and we are
# forcing the format.
log("Format device '%s' failed (%s) - continuing to next device" %
(dev, exc), level=WARNING)
continue
basename = os.path.basename(dev)
_mp = os.path.join('/srv', 'node', basename)
mkdir(_mp, owner='swift', group='swift')
mountpoint = '/srv/node/%s' % basename
if loopback_device:
# If an exiting fstab entry exists using the image file as the
# source then preserve it, otherwise use the loopback device
# directly to avoid a secound implicit loopback device being
# created on mount. Bug #1762390
fstab = charmhelpers.core.fstab.Fstab()
fstab_entry = fstab.get_entry_by_attr('mountpoint', mountpoint)
if fstab_entry and loopback_device == fstab_entry.device:
dev = loopback_device
options = "loop,defaults"
filesystem = "xfs"
mount(dev, mountpoint, filesystem=filesystem)
fstab_add(dev, mountpoint, filesystem, options=options)
check_call(['chown', '-R', 'swift:swift', mountpoint])
check_call(['chmod', '-R', '0755', mountpoint])
# NOTE: record preparation of device - this will be used when
# providing block device configuration for ring builders.
prepared_devices.append(dev)
db.set('prepared-devices', prepared_devices)
db.flush()
@retry_on_exception(3, base_delay=2, exc_type=CalledProcessError)
def fetch_swift_rings(rings_url):
"""Fetch rings from leader proxy unit.
Note that we support a number of retries if a fetch fails since we may
have hit the very small update window on the proxy side.
"""
log('Fetching swift rings from proxy @ %s.' % rings_url, level=INFO)
target = SWIFT_CONF_DIR
tmpdir = tempfile.mkdtemp(prefix='swiftrings')
try:
synced = []
for server in ['account', 'object', 'container']:
url = '%s/%s.%s' % (rings_url, server, SWIFT_RING_EXT)
log('Fetching %s.' % url, level=DEBUG)
ring = '%s.%s' % (server, SWIFT_RING_EXT)
cmd = ['wget', url, '--retry-connrefused', '-t', '10', '-O',
os.path.join(tmpdir, ring)]
check_call(cmd)
synced.append(ring)
# Once all have been successfully downloaded, move them to actual
# location.
for f in synced:
os.rename(os.path.join(tmpdir, f), os.path.join(target, f))
finally:
shutil.rmtree(tmpdir)
def save_script_rc():
env_vars = {}
ip = unit_private_ip()
for server in ['account', 'container', 'object']:
port = config('%s-server-port' % server)
url = 'http://%s:%s/recon/diskusage|"mounted":true' % (ip, port)
svc = server.upper()
env_vars.update({
'OPENSTACK_PORT_%s' % svc: port,
'OPENSTACK_SWIFT_SERVICE_%s' % svc: '%s-server' % server,
'OPENSTACK_URL_%s' % svc: url,
})
_save_script_rc(**env_vars)
def assert_charm_supports_ipv6():
"""Check whether we are able to support charms ipv6."""
_release = lsb_release()['DISTRIB_CODENAME'].lower()
if CompareHostReleases(_release) < "trusty":
raise Exception("IPv6 is not supported in the charms for Ubuntu "
"versions less than Trusty 14.04")
def concat_rsync_fragments():
log('Concatenating rsyncd.d fragments')
rsyncd_dir = '/etc/rsyncd.d'
rsyncd_conf = ""
for filename in sorted(os.listdir(rsyncd_dir)):
with open(os.path.join(rsyncd_dir, filename), 'r') as fragment:
rsyncd_conf += fragment.read()
with open('/etc/rsyncd.conf', 'w') as f:
f.write(rsyncd_conf)
def setup_rsync():
'''
Ensure all directories required for rsync exist with correct permissions.
'''
root_dirs = [
'/etc/rsync-juju.d',
]
[mkdir(d, owner='root', group='root') for d in root_dirs
if not os.path.isdir(d)]
rsyncd_base = """uid = nobody
gid = nogroup
pid file = /var/run/rsyncd.pid
syslog facility = daemon
socket options = SO_KEEPALIVE
&include /etc/rsync-juju.d
"""
f = open('/etc/rsyncd.conf', 'w')
f.write(rsyncd_base)
f.close()
def assess_status(configs):
"""Assess status of current unit"""
if is_paused():
return ("maintenance",
"Paused. Use 'resume' action to resume normal service.")
else:
return ("active", "Unit is ready")
def grant_access(address, port):
"""Grant TCP access to address and port via UFW
:side effect: calls ufw.grant_access
:return: None
"""
log('granting access: {}:{}'.format(address, port), level='DEBUG')
ufw.grant_access(address, port=str(port), proto='tcp',
index=FIRST)
def revoke_access(address, port):
"""Revoke TCP access to address and port via UFW
:side effect: calls ufw.revoke_access
:return: None
"""
log('revoking access: {}'.format(address), level='DEBUG')
ufw.revoke_access(address, port=port, proto='tcp')
def setup_ufw():
"""Setup UFW firewall to ensure only swift-storage clients and storage
peers have access to the swift daemons.
:side effect: calls several external functions
:return: None
"""
if not config('enable-firewall'):
log("Firewall has been administratively disabled", "DEBUG")
return
ports = [config('object-server-port'),
config('container-server-port'),
config('account-server-port')]
# Storage peers
allowed_hosts = RsyncContext()().get('allowed_hosts', '').split(' ')
# Storage clients (swift-proxy)
allowed_hosts += [get_host_ip(ingress_address(rid=u.rid, unit=u.unit))
for u in iter_units_for_relation_name('swift-storage')]
# Grant access for peers and clients
for host in allowed_hosts:
for port in ports:
grant_access(host, port)
# Default deny for storage ports
for port in ports:
ufw.modify_access(src=None, dst='any', port=port,
proto='tcp', action='reject')