def purge(self, service_name: str) -> None:
self.mgr.check_mon_command({
'prefix': 'config rm',
'who': utils.name_to_config_section(service_name),
'name': 'rgw_realm',
})
self.mgr.check_mon_command({
'prefix': 'config rm',
'who': utils.name_to_config_section(service_name),
'name': 'rgw_zone',
})
self.mgr.check_mon_command({
'prefix': 'config-key rm',
'key': f'rgw/cert/{service_name}',
})
def post_remove(self, daemon: DaemonDescription, is_failed_deploy: bool) -> None:
super().post_remove(daemon, is_failed_deploy=is_failed_deploy)
self.mgr.check_mon_command({
'prefix': 'config rm',
'who': utils.name_to_config_section(daemon.name()),
'name': 'rgw_frontends',
})
def create(self, igw_id, host, spec) -> str:
ret, keyring, err = self.mgr.check_mon_command({
'prefix':
'auth get-or-create',
'entity':
utils.name_to_config_section('iscsi') + '.' + igw_id,
'caps': [
'mon', 'profile rbd, '
'allow command "osd blacklist", '
'allow command "config-key get" with "key" prefix "iscsi/"',
'osd', f'allow rwx pool={spec.pool}'
],
})
if spec.ssl_cert:
if isinstance(spec.ssl_cert, list):
cert_data = '\n'.join(spec.ssl_cert)
else:
cert_data = spec.ssl_cert
ret, out, err = self.mgr.mon_command({
'prefix': 'config-key set',
'key':
f'iscsi/{utils.name_to_config_section("iscsi")}.{igw_id}/iscsi-gateway.crt',
'val': cert_data,
})
if spec.ssl_key:
if isinstance(spec.ssl_key, list):
key_data = '\n'.join(spec.ssl_key)
else:
key_data = spec.ssl_key
ret, out, err = self.mgr.mon_command({
'prefix': 'config-key set',
'key':
f'iscsi/{utils.name_to_config_section("iscsi")}.{igw_id}/iscsi-gateway.key',
'val': key_data,
})
api_secure = 'false' if spec.api_secure is None else spec.api_secure
igw_conf = f"""
# generated by cephadm
[config]
cluster_client_name = {utils.name_to_config_section('iscsi')}.{igw_id}
pool = {spec.pool}
trusted_ip_list = {spec.trusted_ip_list or ''}
minimum_gateways = 1
api_port = {spec.api_port or ''}
api_user = {spec.api_user or ''}
api_password = {spec.api_password or ''}
api_secure = {api_secure}
"""
extra_config = {'iscsi-gateway.cfg': igw_conf}
return self.mgr._create_daemon('iscsi',
igw_id,
host,
keyring=keyring,
extra_config=extra_config)
def prepare_create(self, daemon_spec: CephadmDaemonDeploySpec) -> CephadmDaemonDeploySpec:
assert self.TYPE == daemon_spec.daemon_type
rgw_id, _ = daemon_spec.daemon_id, daemon_spec.host
spec = cast(RGWSpec, self.mgr.spec_store[daemon_spec.service_name].spec)
keyring = self.get_keyring(rgw_id)
if daemon_spec.ports:
port = daemon_spec.ports[0]
else:
# this is a redeploy of older instance that doesn't have an explicitly
# assigned port, in which case we can assume there is only 1 per host
# and it matches the spec.
port = spec.get_port()
# configure frontend
args = []
ftype = spec.rgw_frontend_type or "beast"
if ftype == 'beast':
if spec.ssl:
if daemon_spec.ip:
args.append(
f"ssl_endpoint={build_url(host=daemon_spec.ip, port=port).lstrip('/')}")
else:
args.append(f"ssl_port={port}")
args.append(f"ssl_certificate=config://rgw/cert/{spec.service_name()}")
else:
if daemon_spec.ip:
args.append(f"endpoint={build_url(host=daemon_spec.ip, port=port).lstrip('/')}")
else:
args.append(f"port={port}")
elif ftype == 'civetweb':
if spec.ssl:
if daemon_spec.ip:
# note the 's' suffix on port
args.append(f"port={build_url(host=daemon_spec.ip, port=port).lstrip('/')}s")
else:
args.append(f"port={port}s") # note the 's' suffix on port
args.append(f"ssl_certificate=config://rgw/cert/{spec.service_name()}")
else:
if daemon_spec.ip:
args.append(f"port={build_url(host=daemon_spec.ip, port=port).lstrip('/')}")
else:
args.append(f"port={port}")
frontend = f'{ftype} {" ".join(args)}'
ret, out, err = self.mgr.check_mon_command({
'prefix': 'config set',
'who': utils.name_to_config_section(daemon_spec.name()),
'name': 'rgw_frontends',
'value': frontend
})
daemon_spec.keyring = keyring
daemon_spec.final_config, daemon_spec.deps = self.generate_config(daemon_spec)
return daemon_spec
def generate_config(
self, daemon_spec: CephadmDaemonSpec[NFSServiceSpec]
) -> Tuple[Dict[str, Any], List[str]]:
assert self.TYPE == daemon_spec.daemon_type
assert daemon_spec.spec
daemon_type = daemon_spec.daemon_type
daemon_id = daemon_spec.daemon_id
host = daemon_spec.host
spec = daemon_spec.spec
deps: List[str] = []
# create the keyring
user = f'{daemon_type}.{daemon_id}'
entity = utils.name_to_config_section(user)
keyring = self.get_or_create_keyring(entity)
# update the caps after get-or-create, the keyring might already exist!
self.update_keyring_caps(entity, spec)
# create the rados config object
self.create_rados_config_obj(spec)
# generate the ganesha config
def get_ganesha_conf() -> str:
context = dict(user=user,
nodeid=daemon_spec.name(),
pool=spec.pool,
namespace=spec.namespace if spec.namespace else '',
url=spec.rados_config_location())
return self.mgr.template.render('services/nfs/ganesha.conf.j2',
context)
# generate the cephadm config json
def get_cephadm_config() -> Dict[str, Any]:
config: Dict[str, Any] = {}
config['pool'] = spec.pool
if spec.namespace:
config['namespace'] = spec.namespace
config['userid'] = user
config['extra_args'] = ['-N', 'NIV_EVENT']
config['files'] = {
'ganesha.conf': get_ganesha_conf(),
}
config.update(
self.mgr._get_config_and_keyring(daemon_type,
daemon_id,
keyring=keyring,
host=host))
logger.debug('Generated cephadm config-json: %s' % config)
return config
return get_cephadm_config(), deps
def _do_upgrade(self):
# type: () -> None
if not self.upgrade_state:
logger.debug('_do_upgrade no state, exiting')
return
target_image = self.target_image
target_id = self.upgrade_state.target_id
target_digests = self.upgrade_state.target_digests
target_version = self.upgrade_state.target_version
first = False
if not target_id or not target_version or not target_digests:
# need to learn the container hash
logger.info('Upgrade: First pull of %s' % target_image)
self.upgrade_info_str = 'Doing first pull of %s image' % (
target_image)
try:
target_id, target_version, target_digests = CephadmServe(
self.mgr)._get_container_image_info(target_image)
except OrchestratorError as e:
self._fail_upgrade(
'UPGRADE_FAILED_PULL', {
'severity': 'warning',
'summary': 'Upgrade: failed to pull target image',
'count': 1,
'detail': [str(e)],
})
return
if not target_version:
self._fail_upgrade(
'UPGRADE_FAILED_PULL', {
'severity':
'warning',
'summary':
'Upgrade: failed to pull target image',
'count':
1,
'detail':
['unable to extract ceph version from container'],
})
return
self.upgrade_state.target_id = target_id
# extract the version portion of 'ceph version {version} ({sha1})'
self.upgrade_state.target_version = target_version.split(' ')[2]
self.upgrade_state.target_digests = target_digests
self._save_upgrade_state()
target_image = self.target_image
first = True
if target_digests is None:
target_digests = []
if target_version.startswith('ceph version '):
# tolerate/fix upgrade state from older version
self.upgrade_state.target_version = target_version.split(' ')[2]
target_version = self.upgrade_state.target_version
target_major, target_minor, target_patch = target_version.split('.')
target_major_name = self.mgr.lookup_release_name(int(target_major))
if first:
logger.info('Upgrade: Target is version %s (%s)' %
(target_version, target_major_name))
logger.info('Upgrade: Target container is %s, digests %s' %
(target_image, target_digests))
version_error = self._check_target_version(target_version)
if version_error:
self._fail_upgrade(
'UPGRADE_BAD_TARGET_VERSION', {
'severity': 'error',
'summary':
f'Upgrade: cannot upgrade/downgrade to {target_version}',
'count': 1,
'detail': [version_error],
})
return
image_settings = self.get_distinct_container_image_settings()
daemons = [
d for d in self.mgr.cache.get_daemons()
if d.daemon_type in CEPH_UPGRADE_ORDER
]
done = 0
for daemon_type in CEPH_UPGRADE_ORDER:
logger.debug('Upgrade: Checking %s daemons' % daemon_type)
need_upgrade_self = False
need_upgrade: List[Tuple[DaemonDescription, bool]] = []
need_upgrade_deployer: List[Tuple[DaemonDescription, bool]] = []
for d in daemons:
if d.daemon_type != daemon_type:
continue
assert d.daemon_type is not None
assert d.daemon_id is not None
correct_digest = False
if (any(d in target_digests
for d in (d.container_image_digests or []))
or d.daemon_type in MONITORING_STACK_TYPES):
logger.debug('daemon %s.%s container digest correct' %
(daemon_type, d.daemon_id))
correct_digest = True
if any(d in target_digests for d in (d.deployed_by or [])):
logger.debug(
'daemon %s.%s deployed by correct version' %
(d.daemon_type, d.daemon_id))
done += 1
continue
if self.mgr.daemon_is_self(d.daemon_type, d.daemon_id):
logger.info('Upgrade: Need to upgrade myself (mgr.%s)' %
self.mgr.get_mgr_id())
need_upgrade_self = True
continue
if correct_digest:
logger.debug(
'daemon %s.%s not deployed by correct version' %
(d.daemon_type, d.daemon_id))
need_upgrade_deployer.append((d, True))
else:
logger.debug(
'daemon %s.%s not correct (%s, %s, %s)' %
(daemon_type, d.daemon_id, d.container_image_name,
d.container_image_digests, d.version))
need_upgrade.append((d, False))
if not need_upgrade_self:
# only after the mgr itself is upgraded can we expect daemons to have
# deployed_by == target_digests
need_upgrade += need_upgrade_deployer
# prepare filesystems for daemon upgrades?
if (daemon_type == 'mds' and need_upgrade
and not self._prepare_for_mds_upgrade(
target_major, [d_entry[0]
for d_entry in need_upgrade])):
return
if need_upgrade:
self.upgrade_info_str = 'Currently upgrading %s daemons' % (
daemon_type)
to_upgrade: List[Tuple[DaemonDescription, bool]] = []
known_ok_to_stop: List[str] = []
for d_entry in need_upgrade:
d = d_entry[0]
assert d.daemon_type is not None
assert d.daemon_id is not None
assert d.hostname is not None
if not d.container_image_id:
if d.container_image_name == target_image:
logger.debug(
'daemon %s has unknown container_image_id but has correct image name'
% (d.name()))
continue
if known_ok_to_stop:
if d.name() in known_ok_to_stop:
logger.info(
f'Upgrade: {d.name()} is also safe to restart')
to_upgrade.append(d_entry)
continue
if d.daemon_type in ['mon', 'osd', 'mds']:
# NOTE: known_ok_to_stop is an output argument for
# _wait_for_ok_to_stop
if not self._wait_for_ok_to_stop(d, known_ok_to_stop):
return
to_upgrade.append(d_entry)
# if we don't have a list of others to consider, stop now
if not known_ok_to_stop:
break
num = 1
for d_entry in to_upgrade:
d = d_entry[0]
assert d.daemon_type is not None
assert d.daemon_id is not None
assert d.hostname is not None
self._update_upgrade_progress(done / len(daemons))
# make sure host has latest container image
out, errs, code = CephadmServe(self.mgr)._run_cephadm(
d.hostname,
'',
'inspect-image', [],
image=target_image,
no_fsid=True,
error_ok=True)
if code or not any(d in target_digests
for d in json.loads(''.join(out)).get(
'repo_digests', [])):
logger.info('Upgrade: Pulling %s on %s' %
(target_image, d.hostname))
self.upgrade_info_str = 'Pulling %s image on host %s' % (
target_image, d.hostname)
out, errs, code = CephadmServe(self.mgr)._run_cephadm(
d.hostname,
'',
'pull', [],
image=target_image,
no_fsid=True,
error_ok=True)
if code:
self._fail_upgrade(
'UPGRADE_FAILED_PULL', {
'severity':
'warning',
'summary':
'Upgrade: failed to pull target image',
'count':
1,
'detail': [
'failed to pull %s on host %s' %
(target_image, d.hostname)
],
})
return
r = json.loads(''.join(out))
if not any(d in target_digests
for d in r.get('repo_digests', [])):
logger.info(
'Upgrade: image %s pull on %s got new digests %s (not %s), restarting'
% (target_image, d.hostname, r['repo_digests'],
target_digests))
self.upgrade_info_str = 'Image %s pull on %s got new digests %s (not %s), restarting' % (
target_image, d.hostname, r['repo_digests'],
target_digests)
self.upgrade_state.target_digests = r['repo_digests']
self._save_upgrade_state()
return
self.upgrade_info_str = 'Currently upgrading %s daemons' % (
daemon_type)
if len(to_upgrade) > 1:
logger.info(
'Upgrade: Updating %s.%s (%d/%d)' %
(d.daemon_type, d.daemon_id, num, len(to_upgrade)))
else:
logger.info('Upgrade: Updating %s.%s' %
(d.daemon_type, d.daemon_id))
action = 'Upgrading' if not d_entry[1] else 'Redeploying'
try:
daemon_spec = CephadmDaemonDeploySpec.from_daemon_description(
d)
self.mgr._daemon_action(
daemon_spec,
'redeploy',
image=target_image if not d_entry[1] else None)
except Exception as e:
self._fail_upgrade(
'UPGRADE_REDEPLOY_DAEMON', {
'severity': 'warning',
'summary':
f'{action} daemon {d.name()} on host {d.hostname} failed.',
'count': 1,
'detail': [f'Upgrade daemon: {d.name()}: {e}'],
})
return
num += 1
if to_upgrade:
return
# complete mon upgrade?
if daemon_type == 'mon':
if not self.mgr.get("have_local_config_map"):
logger.info(
'Upgrade: Restarting mgr now that mons are running pacific'
)
need_upgrade_self = True
if need_upgrade_self:
try:
self.mgr.mgr_service.fail_over()
except OrchestratorError as e:
self._fail_upgrade(
'UPGRADE_NO_STANDBY_MGR', {
'severity':
'warning',
'summary':
f'Upgrade: {e}',
'count':
1,
'detail': [
'The upgrade process needs to upgrade the mgr, '
'but it needs at least one standby to proceed.',
],
})
return
return # unreachable code, as fail_over never returns
elif daemon_type == 'mgr':
if 'UPGRADE_NO_STANDBY_MGR' in self.mgr.health_checks:
del self.mgr.health_checks['UPGRADE_NO_STANDBY_MGR']
self.mgr.set_health_checks(self.mgr.health_checks)
# make sure 'ceph versions' agrees
ret, out_ver, err = self.mgr.check_mon_command({
'prefix':
'versions',
})
j = json.loads(out_ver)
for version, count in j.get(daemon_type, {}).items():
short_version = version.split(' ')[2]
if short_version != target_version:
logger.warning(
'Upgrade: %d %s daemon(s) are %s != target %s' %
(count, daemon_type, short_version, target_version))
# push down configs
daemon_type_section = name_to_config_section(daemon_type)
if image_settings.get(daemon_type_section) != target_image:
logger.info('Upgrade: Setting container_image for all %s' %
daemon_type)
self.mgr.set_container_image(daemon_type_section, target_image)
to_clean = []
for section in image_settings.keys():
if section.startswith(
name_to_config_section(daemon_type) + '.'):
to_clean.append(section)
if to_clean:
logger.debug('Upgrade: Cleaning up container_image for %s' %
to_clean)
for section in to_clean:
ret, image, err = self.mgr.check_mon_command({
'prefix':
'config rm',
'name':
'container_image',
'who':
section,
})
logger.debug('Upgrade: All %s daemons are up to date.' %
daemon_type)
# complete osd upgrade?
if daemon_type == 'osd':
osdmap = self.mgr.get("osd_map")
osd_min_name = osdmap.get("require_osd_release", "argonaut")
osd_min = ceph_release_to_major(osd_min_name)
if osd_min < int(target_major):
logger.info(
f'Upgrade: Setting require_osd_release to {target_major} {target_major_name}'
)
ret, _, err = self.mgr.check_mon_command({
'prefix':
'osd require-osd-release',
'release':
target_major_name,
})
# complete mds upgrade?
if daemon_type == 'mds' and self.upgrade_state.fs_original_max_mds:
for i in self.mgr.get("fs_map")['filesystems']:
fs_id = i["id"]
fs_name = i['mdsmap']['fs_name']
new_max = self.upgrade_state.fs_original_max_mds.get(fs_id)
if new_max:
self.mgr.log.info(
'Upgrade: Scaling up filesystem %s max_mds to %d' %
(fs_name, new_max))
ret, _, err = self.mgr.check_mon_command({
'prefix':
'fs set',
'fs_name':
fs_name,
'var':
'max_mds',
'val':
str(new_max),
})
self.upgrade_state.fs_original_max_mds = {}
self._save_upgrade_state()
# clean up
logger.info('Upgrade: Finalizing container_image settings')
self.mgr.set_container_image('global', target_image)
for daemon_type in CEPH_UPGRADE_ORDER:
ret, image, err = self.mgr.check_mon_command({
'prefix':
'config rm',
'name':
'container_image',
'who':
name_to_config_section(daemon_type),
})
logger.info('Upgrade: Complete!')
if self.upgrade_state.progress_id:
self.mgr.remote('progress', 'complete',
self.upgrade_state.progress_id)
self.upgrade_state = None
self._save_upgrade_state()
return
def _do_upgrade(self):
# type: () -> None
if not self.upgrade_state:
logger.debug('_do_upgrade no state, exiting')
return
target_name = self.upgrade_state.target_name
target_id = self.upgrade_state.target_id
if not target_id:
# need to learn the container hash
logger.info('Upgrade: First pull of %s' % target_name)
try:
target_id, target_version = self.mgr._get_container_image_id(
target_name)
except OrchestratorError as e:
self._fail_upgrade(
'UPGRADE_FAILED_PULL', {
'severity': 'warning',
'summary': 'Upgrade: failed to pull target image',
'count': 1,
'detail': [str(e)],
})
return
self.upgrade_state.target_id = target_id
self.upgrade_state.target_version = target_version
self._save_upgrade_state()
target_version = self.upgrade_state.target_version
logger.info('Upgrade: Target is %s with id %s' %
(target_name, target_id))
# get all distinct container_image settings
image_settings = {}
ret, out, err = self.mgr.check_mon_command({
'prefix': 'config dump',
'format': 'json',
})
config = json.loads(out)
for opt in config:
if opt['name'] == 'container_image':
image_settings[opt['section']] = opt['value']
daemons = self.mgr.cache.get_daemons()
done = 0
for daemon_type in CEPH_UPGRADE_ORDER:
logger.info('Upgrade: Checking %s daemons...' % daemon_type)
need_upgrade_self = False
for d in daemons:
if d.daemon_type != daemon_type:
continue
if d.container_image_id == target_id:
logger.debug('daemon %s.%s version correct' %
(daemon_type, d.daemon_id))
done += 1
continue
logger.debug('daemon %s.%s not correct (%s, %s, %s)' %
(daemon_type, d.daemon_id, d.container_image_name,
d.container_image_id, d.version))
if daemon_type == 'mgr' and \
d.daemon_id == self.mgr.get_mgr_id():
logger.info('Upgrade: Need to upgrade myself (mgr.%s)' %
self.mgr.get_mgr_id())
need_upgrade_self = True
continue
# make sure host has latest container image
out, err, code = self.mgr._run_cephadm(d.hostname,
'',
'inspect-image', [],
image=target_name,
no_fsid=True,
error_ok=True)
if code or json.loads(
''.join(out)).get('image_id') != target_id:
logger.info('Upgrade: Pulling %s on %s' %
(target_name, d.hostname))
out, err, code = self.mgr._run_cephadm(d.hostname,
'',
'pull', [],
image=target_name,
no_fsid=True,
error_ok=True)
if code:
self._fail_upgrade(
'UPGRADE_FAILED_PULL', {
'severity':
'warning',
'summary':
'Upgrade: failed to pull target image',
'count':
1,
'detail': [
'failed to pull %s on host %s' %
(target_name, d.hostname)
],
})
return
r = json.loads(''.join(out))
if r.get('image_id') != target_id:
logger.info(
'Upgrade: image %s pull on %s got new image %s (not %s), restarting'
% (target_name, d.hostname, r['image_id'],
target_id))
self.upgrade_state.target_id = r['image_id']
self._save_upgrade_state()
return
self._update_upgrade_progress(done / len(daemons))
if not d.container_image_id:
if d.container_image_name == target_name:
logger.debug(
'daemon %s has unknown container_image_id but has correct image name'
% (d.name()))
continue
if not self._wait_for_ok_to_stop(d):
return
logger.info('Upgrade: Redeploying %s.%s' %
(d.daemon_type, d.daemon_id))
ret, out, err = self.mgr.check_mon_command({
'prefix':
'config set',
'name':
'container_image',
'value':
target_name,
'who':
name_to_config_section(daemon_type + '.' + d.daemon_id),
})
self.mgr._daemon_action(d.daemon_type, d.daemon_id, d.hostname,
'redeploy')
return
if need_upgrade_self:
mgr_map = self.mgr.get('mgr_map')
num = len(mgr_map.get('standbys'))
if not num:
self._fail_upgrade(
'UPGRADE_NO_STANDBY_MGR', {
'severity':
'warning',
'summary':
'Upgrade: Need standby mgr daemon',
'count':
1,
'detail': [
'The upgrade process needs to upgrade the mgr, '
'but it needs at least one standby to proceed.',
],
})
return
logger.info('Upgrade: there are %d other already-upgraded '
'standby mgrs, failing over' % num)
self._update_upgrade_progress(done / len(daemons))
# fail over
ret, out, err = self.mgr.check_mon_command({
'prefix':
'mgr fail',
'who':
self.mgr.get_mgr_id(),
})
return
elif daemon_type == 'mgr':
if 'UPGRADE_NO_STANDBY_MGR' in self.mgr.health_checks:
del self.mgr.health_checks['UPGRADE_NO_STANDBY_MGR']
self.mgr.set_health_checks(self.mgr.health_checks)
# make sure 'ceph versions' agrees
ret, out_ver, err = self.mgr.check_mon_command({
'prefix':
'versions',
})
j = json.loads(out_ver)
for version, count in j.get(daemon_type, {}).items():
if version != target_version:
logger.warning(
'Upgrade: %d %s daemon(s) are %s != target %s' %
(count, daemon_type, version, target_version))
# push down configs
if image_settings.get(daemon_type) != target_name:
logger.info('Upgrade: Setting container_image for all %s...' %
daemon_type)
ret, out, err = self.mgr.check_mon_command({
'prefix':
'config set',
'name':
'container_image',
'value':
target_name,
'who':
name_to_config_section(daemon_type),
})
to_clean = []
for section in image_settings.keys():
if section.startswith(
name_to_config_section(daemon_type) + '.'):
to_clean.append(section)
if to_clean:
logger.debug('Upgrade: Cleaning up container_image for %s...' %
to_clean)
for section in to_clean:
ret, image, err = self.mgr.check_mon_command({
'prefix':
'config rm',
'name':
'container_image',
'who':
section,
})
logger.info('Upgrade: All %s daemons are up to date.' %
daemon_type)
# clean up
logger.info('Upgrade: Finalizing container_image settings')
ret, out, err = self.mgr.check_mon_command({
'prefix': 'config set',
'name': 'container_image',
'value': target_name,
'who': 'global',
})
for daemon_type in CEPH_UPGRADE_ORDER:
ret, image, err = self.mgr.check_mon_command({
'prefix':
'config rm',
'name':
'container_image',
'who':
name_to_config_section(daemon_type),
})
logger.info('Upgrade: Complete!')
if self.upgrade_state.progress_id:
self.mgr.remote('progress', 'complete',
self.upgrade_state.progress_id)
self.upgrade_state = None
self._save_upgrade_state()
return
def get_keyring_entity(self) -> str:
return utils.name_to_config_section(self.get_rados_user())
def _do_upgrade(self):
# type: () -> None
if not self.upgrade_state:
logger.debug('_do_upgrade no state, exiting')
return
target_image = self.target_image
target_id = self.upgrade_state.target_id
if not target_id or (self.mgr.use_repo_digest
and not self.upgrade_state.repo_digest):
# need to learn the container hash
logger.info('Upgrade: First pull of %s' % target_image)
try:
target_id, target_version, repo_digest = CephadmServe(
self.mgr)._get_container_image_info(target_image)
except OrchestratorError as e:
self._fail_upgrade(
'UPGRADE_FAILED_PULL', {
'severity': 'warning',
'summary': 'Upgrade: failed to pull target image',
'count': 1,
'detail': [str(e)],
})
return
self.upgrade_state.target_id = target_id
self.upgrade_state.target_version = target_version
self.upgrade_state.repo_digest = repo_digest
self._save_upgrade_state()
target_image = self.target_image
target_version = self.upgrade_state.target_version
logger.info('Upgrade: Target is %s with id %s' %
(target_image, target_id))
image_settings = self.get_distinct_container_image_settings()
daemons = self.mgr.cache.get_daemons()
done = 0
for daemon_type in CEPH_UPGRADE_ORDER:
logger.info('Upgrade: Checking %s daemons...' % daemon_type)
need_upgrade_self = False
for d in daemons:
if d.daemon_type != daemon_type:
continue
if d.container_image_id == target_id:
logger.debug('daemon %s.%s version correct' %
(daemon_type, d.daemon_id))
done += 1
continue
logger.debug('daemon %s.%s not correct (%s, %s, %s)' %
(daemon_type, d.daemon_id, d.container_image_name,
d.container_image_id, d.version))
assert d.daemon_type is not None
assert d.daemon_id is not None
assert d.hostname is not None
if self.mgr.daemon_is_self(d.daemon_type, d.daemon_id):
logger.info('Upgrade: Need to upgrade myself (mgr.%s)' %
self.mgr.get_mgr_id())
need_upgrade_self = True
continue
# make sure host has latest container image
out, err, code = CephadmServe(self.mgr)._run_cephadm(
d.hostname,
'',
'inspect-image', [],
image=target_image,
no_fsid=True,
error_ok=True)
if code or json.loads(
''.join(out)).get('image_id') != target_id:
logger.info('Upgrade: Pulling %s on %s' %
(target_image, d.hostname))
out, err, code = CephadmServe(self.mgr)._run_cephadm(
d.hostname,
'',
'pull', [],
image=target_image,
no_fsid=True,
error_ok=True)
if code:
self._fail_upgrade(
'UPGRADE_FAILED_PULL', {
'severity':
'warning',
'summary':
'Upgrade: failed to pull target image',
'count':
1,
'detail': [
'failed to pull %s on host %s' %
(target_image, d.hostname)
],
})
return
r = json.loads(''.join(out))
if r.get('image_id') != target_id:
logger.info(
'Upgrade: image %s pull on %s got new image %s (not %s), restarting'
% (target_image, d.hostname, r['image_id'],
target_id))
self.upgrade_state.target_id = r['image_id']
self._save_upgrade_state()
return
self._update_upgrade_progress(done / len(daemons))
if not d.container_image_id:
if d.container_image_name == target_image:
logger.debug(
'daemon %s has unknown container_image_id but has correct image name'
% (d.name()))
continue
if not self._wait_for_ok_to_stop(d):
return
logger.info('Upgrade: Redeploying %s.%s' %
(d.daemon_type, d.daemon_id))
try:
self.mgr._daemon_action(d.daemon_type,
d.daemon_id,
d.hostname,
'redeploy',
image=target_image)
except Exception as e:
self._fail_upgrade(
'UPGRADE_REDEPLOY_DAEMON', {
'severity': 'warning',
'summary':
f'Upgrading daemon {d.name()} on host {d.hostname} failed.',
'count': 1,
'detail': [f'Upgrade daemon: {d.name()}: {e}'],
})
return
if need_upgrade_self:
try:
self.mgr.mgr_service.fail_over()
except OrchestratorError as e:
self._fail_upgrade(
'UPGRADE_NO_STANDBY_MGR', {
'severity':
'warning',
'summary':
f'Upgrade: {e}',
'count':
1,
'detail': [
'The upgrade process needs to upgrade the mgr, '
'but it needs at least one standby to proceed.',
],
})
return
return # unreachable code, as fail_over never returns
elif daemon_type == 'mgr':
if 'UPGRADE_NO_STANDBY_MGR' in self.mgr.health_checks:
del self.mgr.health_checks['UPGRADE_NO_STANDBY_MGR']
self.mgr.set_health_checks(self.mgr.health_checks)
# make sure 'ceph versions' agrees
ret, out_ver, err = self.mgr.check_mon_command({
'prefix':
'versions',
})
j = json.loads(out_ver)
for version, count in j.get(daemon_type, {}).items():
if version != target_version:
logger.warning(
'Upgrade: %d %s daemon(s) are %s != target %s' %
(count, daemon_type, version, target_version))
# push down configs
if image_settings.get(daemon_type) != target_image:
logger.info('Upgrade: Setting container_image for all %s...' %
daemon_type)
self.mgr.set_container_image(
name_to_config_section(daemon_type), target_image)
to_clean = []
for section in image_settings.keys():
if section.startswith(
name_to_config_section(daemon_type) + '.'):
to_clean.append(section)
if to_clean:
logger.debug('Upgrade: Cleaning up container_image for %s...' %
to_clean)
for section in to_clean:
ret, image, err = self.mgr.check_mon_command({
'prefix':
'config rm',
'name':
'container_image',
'who':
section,
})
logger.info('Upgrade: All %s daemons are up to date.' %
daemon_type)
# clean up
logger.info('Upgrade: Finalizing container_image settings')
self.mgr.set_container_image('global', target_image)
for daemon_type in CEPH_UPGRADE_ORDER:
ret, image, err = self.mgr.check_mon_command({
'prefix':
'config rm',
'name':
'container_image',
'who':
name_to_config_section(daemon_type),
})
logger.info('Upgrade: Complete!')
if self.upgrade_state.progress_id:
self.mgr.remote('progress', 'complete',
self.upgrade_state.progress_id)
self.upgrade_state = None
self._save_upgrade_state()
return
