def assess_status():
'''Assess status of current unit'''
application_version_set(get_upstream_version(VERSION_PACKAGE))
# Check that the no-bootstrap config option is set in conjunction with
# having the bootstrap-source relation established
if not config('no-bootstrap') and is_relation_made('bootstrap-source'):
status_set(
'blocked', 'Cannot join the bootstrap-source relation when '
'no-bootstrap is False')
return
moncount = int(config('monitor-count'))
units = get_peer_units()
# not enough peers and mon_count > 1
if len(units.keys()) < moncount:
status_set(
'blocked', 'Insufficient peer units to bootstrap'
' cluster (require {})'.format(moncount))
return
# mon_count > 1, peers, but no ceph-public-address
ready = sum(1 for unit_ready in units.values() if unit_ready)
if ready < moncount:
status_set('waiting', 'Peer units detected, waiting for addresses')
return
# active - bootstrapped + quorum status check
if ceph.is_bootstrapped() and ceph.is_quorum():
status_set('active', 'Unit is ready and clustered')
else:
# Unit should be running and clustered, but no quorum
# TODO: should this be blocked or waiting?
status_set('blocked', 'Unit not clustered (no quorum)')
def config_changed():
# Get the cfg object so we can see if the no-bootstrap value has changed
# and triggered this hook invocation
cfg = config()
if config('prefer-ipv6'):
assert_charm_supports_ipv6()
check_for_upgrade()
log('Monitor hosts are ' + repr(get_mon_hosts()))
sysctl_dict = config('sysctl')
if sysctl_dict:
create_sysctl(sysctl_dict, '/etc/sysctl.d/50-ceph-charm.conf')
if relations_of_type('nrpe-external-master'):
update_nrpe_config()
if is_leader():
if not config('no-bootstrap'):
if not leader_get('fsid') or not leader_get('monitor-secret'):
if config('fsid'):
fsid = config('fsid')
else:
fsid = "{}".format(uuid.uuid1())
if config('monitor-secret'):
mon_secret = config('monitor-secret')
else:
mon_secret = "{}".format(ceph.generate_monitor_secret())
status_set('maintenance', 'Creating FSID and Monitor Secret')
opts = {
'fsid': fsid,
'monitor-secret': mon_secret,
}
log("Settings for the cluster are: {}".format(opts))
leader_set(opts)
elif cfg.changed('no-bootstrap') and \
is_relation_made('bootstrap-source'):
# User changed the no-bootstrap config option, we're the leader,
# and the bootstrap-source relation has been made. The charm should
# be in a blocked state indicating that the no-bootstrap option
# must be set. This block is invoked when the user is trying to
# get out of that scenario by enabling no-bootstrap.
bootstrap_source_relation_changed()
elif leader_get('fsid') is None or leader_get('monitor-secret') is None:
log('still waiting for leader to setup keys')
status_set('waiting', 'Waiting for leader to setup keys')
sys.exit(0)
emit_cephconf()
# Support use of single node ceph
if (not ceph.is_bootstrapped() and int(config('monitor-count')) == 1
and is_leader()):
status_set('maintenance', 'Bootstrapping single Ceph MON')
ceph.bootstrap_monitor_cluster(leader_get('monitor-secret'))
ceph.wait_for_bootstrap()
if cmp_pkgrevno('ceph', '12.0.0') >= 0:
status_set('maintenance', 'Bootstrapping single Ceph MGR')
ceph.bootstrap_manager()
def assess_status():
'''Assess status of current unit'''
application_version_set(get_upstream_version(VERSION_PACKAGE))
if is_unit_upgrading_set():
status_set(
"blocked", "Ready for do-release-upgrade and reboot. "
"Set complete when finished.")
return
# Check that the no-bootstrap config option is set in conjunction with
# having the bootstrap-source relation established
if not config('no-bootstrap') and is_relation_made('bootstrap-source'):
status_set(
'blocked', 'Cannot join the bootstrap-source relation when '
'no-bootstrap is False')
return
moncount = int(config('monitor-count'))
units = get_peer_units()
# not enough peers and mon_count > 1
if len(units.keys()) < moncount:
status_set(
'blocked', 'Insufficient peer units to bootstrap'
' cluster (require {})'.format(moncount))
return
# mon_count > 1, peers, but no ceph-public-address
ready = sum(1 for unit_ready in units.values() if unit_ready)
if ready < moncount:
status_set('waiting', 'Peer units detected, waiting for addresses')
return
configured_rbd_features = config('default-rbd-features')
if has_rbd_mirrors() and configured_rbd_features:
if add_rbd_mirror_features(
configured_rbd_features) != configured_rbd_features:
# The configured RBD features bitmap does not contain the features
# required for RBD Mirroring
status_set(
'blocked', 'Configuration mismatch: RBD Mirroring '
'enabled but incorrect value set for '
'``default-rbd-features``')
return
# active - bootstrapped + quorum status check
if ceph.is_bootstrapped() and ceph.is_quorum():
expected_osd_count = config('expected-osd-count') or 3
if sufficient_osds(expected_osd_count):
status_set('active', 'Unit is ready and clustered')
else:
status_set(
'waiting', 'Monitor bootstrapped but waiting for number of'
' OSDs to reach expected-osd-count ({})'.format(
expected_osd_count))
else:
# Unit should be running and clustered, but no quorum
# TODO: should this be blocked or waiting?
status_set('blocked', 'Unit not clustered (no quorum)')
def prepare_disks_and_activate():
# NOTE: vault/vaultlocker preflight check
vault_kv = vaultlocker.VaultKVContext(vaultlocker.VAULTLOCKER_BACKEND)
context = vault_kv()
if use_vaultlocker() and not vault_kv.complete:
log('Deferring OSD preparation as vault not ready', level=DEBUG)
return
elif use_vaultlocker() and vault_kv.complete:
log('Vault ready, writing vaultlocker configuration', level=DEBUG)
vaultlocker.write_vaultlocker_conf(context)
osd_journal = get_journal_devices()
if not osd_journal.isdisjoint(set(get_devices())):
raise ValueError('`osd-journal` and `osd-devices` options must not'
'overlap.')
log("got journal devs: {}".format(osd_journal), level=DEBUG)
# pre-flight check of eligible device pristinity
devices = get_devices()
# if a device has been previously touched we need to consider it as
# non-pristine. If it needs to be re-processed it has to be zapped
# via the respective action which also clears the unitdata entry.
db = kv()
touched_devices = db.get('osd-devices', [])
devices = [dev for dev in devices if dev not in touched_devices]
log('Skipping osd devices previously processed by this unit: {}'.format(
touched_devices))
# filter osd-devices that are file system paths
devices = [dev for dev in devices if dev.startswith('/dev')]
# filter osd-devices that does not exist on this unit
devices = [dev for dev in devices if os.path.exists(dev)]
# filter osd-devices that are already mounted
devices = [dev for dev in devices if not is_device_mounted(dev)]
# filter osd-devices that are active bluestore devices
devices = [
dev for dev in devices if not ceph.is_active_bluestore_device(dev)
]
log('Checking for pristine devices: "{}"'.format(devices), level=DEBUG)
if not all(ceph.is_pristine_disk(dev) for dev in devices):
status_set(
'blocked', 'Non-pristine devices detected, consult '
'`list-disks`, `zap-disk` and `blacklist-*` actions.')
return
if ceph.is_bootstrapped():
log('ceph bootstrapped, rescanning disks')
emit_cephconf()
for dev in get_devices():
ceph.osdize(dev, config('osd-format'), osd_journal,
config('ignore-device-errors'), config('osd-encrypt'),
config('bluestore'), config('osd-encrypt-keymanager'))
# Make it fast!
if config('autotune'):
ceph.tune_dev(dev)
ceph.start_osds(get_devices())
def mon_relation():
if leader_get('monitor-secret') is None:
log('still waiting for leader to setup keys')
status_set('waiting', 'Waiting for leader to setup keys')
return
emit_cephconf()
moncount = int(config('monitor-count'))
if len(get_mon_hosts()) >= moncount:
if not ceph.is_bootstrapped():
status_set('maintenance', 'Bootstrapping MON cluster')
# the following call raises an exception
# if it can't add the keyring
try:
ceph.bootstrap_monitor_cluster(leader_get('monitor-secret'))
except FileNotFoundError as e: # NOQA -- PEP8 is still PY2
log("Couldn't bootstrap the monitor yet: {}".format(str(e)))
exit(0)
ceph.wait_for_bootstrap()
ceph.wait_for_quorum()
if cmp_pkgrevno('ceph', '12.0.0') >= 0:
status_set('maintenance', 'Bootstrapping Ceph MGR')
ceph.bootstrap_manager()
# If we can and want to
if is_leader() and config('customize-failure-domain'):
# But only if the environment supports it
if os.environ.get('JUJU_AVAILABILITY_ZONE'):
cmds = [
"ceph osd getcrushmap -o /tmp/crush.map",
"crushtool -d /tmp/crush.map| "
"sed 's/step chooseleaf firstn 0 type host/step "
"chooseleaf firstn 0 type rack/' > "
"/tmp/crush.decompiled",
"crushtool -c /tmp/crush.decompiled -o /tmp/crush.map",
"crushtool -i /tmp/crush.map --test",
"ceph osd setcrushmap -i /tmp/crush.map"
]
for cmd in cmds:
try:
subprocess.check_call(cmd, shell=True)
except subprocess.CalledProcessError as e:
log("Failed to modify crush map:", level='error')
log("Cmd: {}".format(cmd), level='error')
log("Error: {}".format(e.output), level='error')
break
else:
log("Your Juju environment doesn't"
"have support for Availability Zones")
notify_osds()
notify_radosgws()
notify_client()
else:
log('Not enough mons ({}), punting.'.format(len(get_mon_hosts())))
def prepare_disks_and_activate():
osd_journal = get_journal_devices()
check_overlap(osd_journal, set(get_devices()))
log("got journal devs: {}".format(osd_journal), level=DEBUG)
already_zapped = read_zapped_journals()
non_zapped = osd_journal - already_zapped
for journ in non_zapped:
ceph.maybe_zap_journal(journ)
write_zapped_journals(osd_journal)
if ceph.is_bootstrapped():
log('ceph bootstrapped, rescanning disks')
emit_cephconf()
for dev in get_devices():
ceph.osdize(dev, config('osd-format'), osd_journal,
config('osd-reformat'), config('ignore-device-errors'),
config('osd-encrypt'), config('bluestore'))
# Make it fast!
if config('autotune'):
ceph.tune_dev(dev)
ceph.start_osds(get_devices())
def check_for_upgrade():
if not ceph.is_bootstrapped():
log("Ceph is not bootstrapped, skipping upgrade checks.")
return
c = hookenv.config()
old_version = ceph.resolve_ceph_version(c.previous('source') or
'distro')
log('old_version: {}'.format(old_version))
new_version = ceph.resolve_ceph_version(hookenv.config('source') or
'distro')
log('new_version: {}'.format(new_version))
# May be in a previous upgrade that was failed if the directories
# still need an ownership update. Check this condition.
resuming_upgrade = ceph.dirs_need_ownership_update('osd')
if old_version == new_version and not resuming_upgrade:
log("No new ceph version detected, skipping upgrade.", DEBUG)
return
if (ceph.UPGRADE_PATHS.get(old_version) == new_version) or\
resuming_upgrade:
if old_version == new_version:
log('Attempting to resume possibly failed upgrade.',
INFO)
else:
log("{} to {} is a valid upgrade path. Proceeding.".format(
old_version, new_version))
emit_cephconf(upgrading=True)
ceph.roll_osd_cluster(new_version=new_version,
upgrade_key='osd-upgrade')
emit_cephconf(upgrading=False)
else:
# Log a helpful error message
log("Invalid upgrade path from {} to {}. "
"Valid paths are: {}".format(old_version,
new_version,
ceph.pretty_print_upgrade_paths()))
def check_for_upgrade():
if not ceph.is_bootstrapped():
log("Ceph is not bootstrapped, skipping upgrade checks.")
return
c = hookenv.config()
old_version = ceph.resolve_ceph_version(c.previous('source') or 'distro')
log('old_version: {}'.format(old_version))
# Strip all whitespace
new_version = ceph.resolve_ceph_version(hookenv.config('source'))
log('new_version: {}'.format(new_version))
if (old_version in ceph.UPGRADE_PATHS
and new_version == ceph.UPGRADE_PATHS[old_version]):
log("{} to {} is a valid upgrade path. Proceeding.".format(
old_version, new_version))
ceph.roll_monitor_cluster(new_version=new_version, upgrade_key='admin')
else:
# Log a helpful error message
log("Invalid upgrade path from {} to {}. "
"Valid paths are: {}".format(old_version, new_version,
ceph.pretty_print_upgrade_paths()))
def check_for_upgrade():
if not ceph.is_bootstrapped():
log("Ceph is not bootstrapped, skipping upgrade checks.")
return
c = hookenv.config()
old_version = ceph.resolve_ceph_version(c.previous('source') or 'distro')
log('old_version: {}'.format(old_version))
# Strip all whitespace
new_version = ceph.resolve_ceph_version(hookenv.config('source'))
old_version_os = get_os_codename_install_source(
c.previous('source') or 'distro')
new_version_os = get_os_codename_install_source(hookenv.config('source'))
log('new_version: {}'.format(new_version))
if (old_version in ceph.UPGRADE_PATHS
and new_version == ceph.UPGRADE_PATHS[old_version]):
log("{} to {} is a valid upgrade path. Proceeding.".format(
old_version, new_version))
ceph.roll_monitor_cluster(new_version=new_version, upgrade_key='admin')
elif (old_version == new_version and old_version_os < new_version_os):
# See LP: #1778823
add_source(hookenv.config('source'), hookenv.config('key'))
log(("The installation source has changed yet there is no new major "
"version of Ceph in this new source. As a result no package "
"upgrade will take effect. Please upgrade manually if you need "
"to."),
level=INFO)
else:
# Log a helpful error message
log("Invalid upgrade path from {} to {}. "
"Valid paths are: {}".format(old_version, new_version,
ceph.pretty_print_upgrade_paths()),
level=ERROR)
def prepare_disks_and_activate():
# NOTE: vault/vaultlocker preflight check
vault_kv = vaultlocker.VaultKVContext(vaultlocker.VAULTLOCKER_BACKEND)
context = vault_kv()
if use_vaultlocker() and not vault_kv.complete:
log('Deferring OSD preparation as vault not ready',
level=DEBUG)
return
elif use_vaultlocker() and vault_kv.complete:
log('Vault ready, writing vaultlocker configuration',
level=DEBUG)
vaultlocker.write_vaultlocker_conf(context)
osd_journal = get_journal_devices()
if not osd_journal.isdisjoint(set(get_devices())):
raise ValueError('`osd-journal` and `osd-devices` options must not'
'overlap.')
log("got journal devs: {}".format(osd_journal), level=DEBUG)
# pre-flight check of eligible device pristinity
devices = get_devices()
# if a device has been previously touched we need to consider it as
# non-pristine. If it needs to be re-processed it has to be zapped
# via the respective action which also clears the unitdata entry.
db = kv()
touched_devices = db.get('osd-devices', [])
devices = [dev for dev in devices if dev not in touched_devices]
log('Skipping osd devices previously processed by this unit: {}'
.format(touched_devices))
# filter osd-devices that are file system paths
devices = [dev for dev in devices if dev.startswith('/dev')]
# filter osd-devices that does not exist on this unit
devices = [dev for dev in devices if os.path.exists(dev)]
# filter osd-devices that are already mounted
devices = [dev for dev in devices if not is_device_mounted(dev)]
# filter osd-devices that are active bluestore devices
devices = [dev for dev in devices
if not ceph.is_active_bluestore_device(dev)]
log('Checking for pristine devices: "{}"'.format(devices), level=DEBUG)
if not all(ceph.is_pristine_disk(dev) for dev in devices):
status_set('blocked',
'Non-pristine devices detected, consult '
'`list-disks`, `zap-disk` and `blacklist-*` actions.')
return
if ceph.is_bootstrapped():
log('ceph bootstrapped, rescanning disks')
emit_cephconf()
bluestore = use_bluestore()
ceph.udevadm_settle()
for dev in get_devices():
ceph.osdize(dev, config('osd-format'),
osd_journal,
config('ignore-device-errors'),
config('osd-encrypt'),
bluestore,
config('osd-encrypt-keymanager'))
# Make it fast!
if config('autotune'):
ceph.tune_dev(dev)
ceph.start_osds(get_devices())
# Notify MON cluster as to how many OSD's this unit bootstrapped
# into the cluster
for r_id in relation_ids('mon'):
relation_set(
relation_id=r_id,
relation_settings={
'bootstrapped-osds': len(db.get('osd-devices', []))
}
)
def config_changed():
# Get the cfg object so we can see if the no-bootstrap value has changed
# and triggered this hook invocation
cfg = config()
if config('prefer-ipv6'):
assert_charm_supports_ipv6()
check_for_upgrade()
log('Monitor hosts are ' + repr(get_mon_hosts()))
sysctl_dict = config('sysctl')
if sysctl_dict:
create_sysctl(sysctl_dict, '/etc/sysctl.d/50-ceph-charm.conf')
if relations_of_type('nrpe-external-master'):
update_nrpe_config()
if is_leader():
if not config('no-bootstrap'):
if not leader_get('fsid') or not leader_get('monitor-secret'):
fsid = "{}".format(uuid.uuid1())
if config('monitor-secret'):
mon_secret = config('monitor-secret')
else:
mon_secret = "{}".format(ceph.generate_monitor_secret())
opts = {
'fsid': fsid,
'monitor-secret': mon_secret,
}
try:
leader_set(opts)
status_set('maintenance',
'Created FSID and Monitor Secret')
log("Settings for the cluster are: {}".format(opts))
except Exception as e:
# we're probably not the leader an exception occured
# let's log it anyway.
log("leader_set failed: {}".format(str(e)))
elif (cfg.changed('no-bootstrap')
and is_relation_made('bootstrap-source')):
# User changed the no-bootstrap config option, we're the leader,
# and the bootstrap-source relation has been made. The charm should
# be in a blocked state indicating that the no-bootstrap option
# must be set. This block is invoked when the user is trying to
# get out of that scenario by enabling no-bootstrap.
bootstrap_source_relation_changed()
# unconditionally verify that the fsid and monitor-secret are set now
# otherwise we exit until a leader does this.
if leader_get('fsid') is None or leader_get('monitor-secret') is None:
log('still waiting for leader to setup keys')
status_set('waiting', 'Waiting for leader to setup keys')
return
emit_cephconf()
# Support use of single node ceph
if (not ceph.is_bootstrapped() and int(config('monitor-count')) == 1
and is_leader()):
status_set('maintenance', 'Bootstrapping single Ceph MON')
# the following call raises an exception if it can't add the keyring
try:
ceph.bootstrap_monitor_cluster(leader_get('monitor-secret'))
except FileNotFoundError as e: # NOQA -- PEP8 is still PY2
log("Couldn't bootstrap the monitor yet: {}".format(str(e)))
return
ceph.wait_for_bootstrap()
if cmp_pkgrevno('ceph', '12.0.0') >= 0:
status_set('maintenance', 'Bootstrapping single Ceph MGR')
ceph.bootstrap_manager()
# Update client relations
notify_client()