def test_parse_placement_specs_raises(test_input):
with pytest.raises(ServiceSpecValidationError):
PlacementSpec.from_string(test_input)
def describe_service(
self,
service_type: Optional[str] = None,
service_name: Optional[str] = None,
refresh: bool = False) -> List[orchestrator.ServiceDescription]:
now = datetime.datetime.utcnow()
# CephCluster
cl = self.rook_cluster.rook_api_get("cephclusters/{0}".format(
self.rook_cluster.rook_env.cluster_name))
self.log.debug('CephCluster %s' % cl)
image_name = cl['spec'].get('cephVersion', {}).get('image', None)
num_nodes = len(self.rook_cluster.get_node_names())
spec = {}
if service_type == 'mon' or service_type is None:
spec['mon'] = orchestrator.ServiceDescription(
spec=ServiceSpec(
'mon',
placement=PlacementSpec(count=cl['spec'].get(
'mon', {}).get('count', 1), ),
),
size=cl['spec'].get('mon', {}).get('count', 1),
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'mgr' or service_type is None:
spec['mgr'] = orchestrator.ServiceDescription(
spec=ServiceSpec(
'mgr',
placement=PlacementSpec.from_string('count:1'),
),
size=1,
container_image_name=image_name,
last_refresh=now,
)
if not cl['spec'].get('crashCollector', {}).get('disable', False):
spec['crash'] = orchestrator.ServiceDescription(
spec=ServiceSpec(
'crash',
placement=PlacementSpec.from_string('*'),
),
size=num_nodes,
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'mds' or service_type is None:
# CephFilesystems
all_fs = self.rook_cluster.rook_api_get("cephfilesystems/")
self.log.debug('CephFilesystems %s' % all_fs)
for fs in all_fs.get('items', []):
svc = 'mds.' + fs['metadata']['name']
if svc in spec:
continue
# FIXME: we are conflating active (+ standby) with count
active = fs['spec'].get('metadataServer',
{}).get('activeCount', 1)
total_mds = active
if fs['spec'].get('metadataServer',
{}).get('activeStandby', False):
total_mds = active * 2
spec[svc] = orchestrator.ServiceDescription(
spec=ServiceSpec(
service_type='mds',
service_id=fs['metadata']['name'],
placement=PlacementSpec(count=active),
),
size=total_mds,
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'rgw' or service_type is None:
# CephObjectstores
all_zones = self.rook_cluster.rook_api_get("cephobjectstores/")
self.log.debug('CephObjectstores %s' % all_zones)
for zone in all_zones.get('items', []):
rgw_realm = zone['metadata']['name']
rgw_zone = rgw_realm
svc = 'rgw.' + rgw_realm + '.' + rgw_zone
if svc in spec:
continue
active = zone['spec']['gateway']['instances']
if 'securePort' in zone['spec']['gateway']:
ssl = True
port = zone['spec']['gateway']['securePort']
else:
ssl = False
port = zone['spec']['gateway']['port'] or 80
spec[svc] = orchestrator.ServiceDescription(
spec=RGWSpec(
service_id=rgw_realm + '.' + rgw_zone,
rgw_realm=rgw_realm,
rgw_zone=rgw_zone,
ssl=ssl,
rgw_frontend_port=port,
placement=PlacementSpec(count=active),
),
size=active,
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'nfs' or service_type is None:
# CephNFSes
all_nfs = self.rook_cluster.rook_api_get("cephnfses/")
self.log.warning('CephNFS %s' % all_nfs)
for nfs in all_nfs.get('items', []):
nfs_name = nfs['metadata']['name']
svc = 'nfs.' + nfs_name
if svc in spec:
continue
active = nfs['spec'].get('server', {}).get('active')
spec[svc] = orchestrator.ServiceDescription(
spec=NFSServiceSpec(
service_id=nfs_name,
pool=nfs['spec']['rados']['pool'],
namespace=nfs['spec']['rados'].get('namespace', None),
placement=PlacementSpec(count=active),
),
size=active,
last_refresh=now,
)
for dd in self._list_daemons():
if dd.service_name() not in spec:
continue
service = spec[dd.service_name()]
service.running += 1
if not service.container_image_id:
service.container_image_id = dd.container_image_id
if not service.container_image_name:
service.container_image_name = dd.container_image_name
if service.last_refresh is None or not dd.last_refresh or dd.last_refresh < service.last_refresh:
service.last_refresh = dd.last_refresh
if service.created is None or dd.created is None or dd.created < service.created:
service.created = dd.created
return [v for k, v in spec.items()]
def test_parse_placement_specs(test_input, expected):
ret = PlacementSpec.from_string(test_input)
assert str(ret) == expected
def describe_service(
self,
service_type: Optional[str] = None,
service_name: Optional[str] = None,
refresh: bool = False) -> List[orchestrator.ServiceDescription]:
now = datetime_now()
# CephCluster
cl = self.rook_cluster.rook_api_get("cephclusters/{0}".format(
self.rook_cluster.rook_env.cluster_name))
self.log.debug('CephCluster %s' % cl)
image_name = cl['spec'].get('cephVersion', {}).get('image', None)
num_nodes = len(self.rook_cluster.get_node_names())
spec = {}
if service_type == 'mon' or service_type is None:
spec['mon'] = orchestrator.ServiceDescription(
spec=ServiceSpec(
'mon',
placement=PlacementSpec(count=cl['spec'].get(
'mon', {}).get('count', 1), ),
),
size=cl['spec'].get('mon', {}).get('count', 1),
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'mgr' or service_type is None:
spec['mgr'] = orchestrator.ServiceDescription(
spec=ServiceSpec(
'mgr',
placement=PlacementSpec.from_string('count:1'),
),
size=1,
container_image_name=image_name,
last_refresh=now,
)
if (service_type == 'crash'
or service_type is None and not cl['spec'].get(
'crashCollector', {}).get('disable', False)):
spec['crash'] = orchestrator.ServiceDescription(
spec=ServiceSpec(
'crash',
placement=PlacementSpec.from_string('*'),
),
size=num_nodes,
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'mds' or service_type is None:
# CephFilesystems
all_fs = self.rook_cluster.get_resource("cephfilesystems")
for fs in all_fs:
svc = 'mds.' + fs['metadata']['name']
if svc in spec:
continue
# FIXME: we are conflating active (+ standby) with count
active = fs['spec'].get('metadataServer',
{}).get('activeCount', 1)
total_mds = active
if fs['spec'].get('metadataServer',
{}).get('activeStandby', False):
total_mds = active * 2
spec[svc] = orchestrator.ServiceDescription(
spec=ServiceSpec(
service_type='mds',
service_id=fs['metadata']['name'],
placement=PlacementSpec(count=active),
),
size=total_mds,
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'rgw' or service_type is None:
# CephObjectstores
all_zones = self.rook_cluster.get_resource("cephobjectstores")
for zone in all_zones:
svc = 'rgw.' + zone['metadata']['name']
if svc in spec:
continue
active = zone['spec']['gateway']['instances']
if 'securePort' in zone['spec']['gateway']:
ssl = True
port = zone['spec']['gateway']['securePort']
else:
ssl = False
port = zone['spec']['gateway']['port'] or 80
rgw_zone = zone['spec'].get('zone', {}).get('name') or None
spec[svc] = orchestrator.ServiceDescription(
spec=RGWSpec(
service_id=zone['metadata']['name'],
rgw_zone=rgw_zone,
ssl=ssl,
rgw_frontend_port=port,
placement=PlacementSpec(count=active),
),
size=active,
container_image_name=image_name,
last_refresh=now,
)
if service_type == 'nfs' or service_type is None:
# CephNFSes
all_nfs = self.rook_cluster.get_resource("cephnfses")
nfs_pods = self.rook_cluster.describe_pods('nfs', None, None)
for nfs in all_nfs:
if nfs['spec']['rados']['pool'] != NFS_POOL_NAME:
continue
nfs_name = nfs['metadata']['name']
svc = 'nfs.' + nfs_name
if svc in spec:
continue
active = nfs['spec'].get('server', {}).get('active')
creation_timestamp = datetime.datetime.strptime(
nfs['metadata']['creationTimestamp'], '%Y-%m-%dT%H:%M:%SZ')
spec[svc] = orchestrator.ServiceDescription(
spec=NFSServiceSpec(
service_id=nfs_name,
placement=PlacementSpec(count=active),
),
size=active,
last_refresh=now,
running=len([
1 for pod in nfs_pods
if pod['labels']['ceph_nfs'] == nfs_name
]),
created=creation_timestamp.astimezone(
tz=datetime.timezone.utc))
if service_type == 'osd' or service_type is None:
# OSDs
# FIXME: map running OSDs back to their respective services...
# the catch-all unmanaged
all_osds = self.rook_cluster.get_osds()
svc = 'osd'
spec[svc] = orchestrator.ServiceDescription(
spec=DriveGroupSpec(
unmanaged=True,
service_type='osd',
),
size=len(all_osds),
last_refresh=now,
running=sum(osd.status.phase == 'Running' for osd in all_osds))
# drivegroups
for name, dg in self._drive_group_map.items():
spec[f'osd.{name}'] = orchestrator.ServiceDescription(
spec=dg,
last_refresh=now,
size=0,
running=0,
)
if service_type == 'rbd-mirror' or service_type is None:
# rbd-mirrors
all_mirrors = self.rook_cluster.get_resource("cephrbdmirrors")
for mirror in all_mirrors:
logging.warn(mirror)
mirror_name = mirror['metadata']['name']
svc = 'rbd-mirror.' + mirror_name
if svc in spec:
continue
spec[svc] = orchestrator.ServiceDescription(
spec=ServiceSpec(
service_id=mirror_name,
service_type="rbd-mirror",
placement=PlacementSpec(count=1),
),
size=1,
last_refresh=now,
)
for dd in self._list_daemons():
if dd.service_name() not in spec:
continue
service = spec[dd.service_name()]
service.running += 1
if not service.container_image_id:
service.container_image_id = dd.container_image_id
if not service.container_image_name:
service.container_image_name = dd.container_image_name
if service.last_refresh is None or not dd.last_refresh or dd.last_refresh < service.last_refresh:
service.last_refresh = dd.last_refresh
if service.created is None or dd.created is None or dd.created < service.created:
service.created = dd.created
return [v for k, v in spec.items()]
def test_parse_placement_specs(test_input, expected):
ret = PlacementSpec.from_string(test_input)
assert str(ret) == expected
assert PlacementSpec.from_string(
ret.pretty_str()) == ret, f'"{ret.pretty_str()}" != "{test_input}"'
def test_bad_placements(placement):
try:
s = PlacementSpec.from_string(placement.split(' '))
assert False
except ServiceSpecValidationError as e:
pass
def describe_service(self, service_type=None, service_name=None,
refresh=False):
now = datetime.datetime.utcnow()
# CephCluster
cl = self.rook_cluster.rook_api_get(
"cephclusters/{0}".format(self.rook_cluster.rook_env.cluster_name))
self.log.debug('CephCluster %s' % cl)
image_name = cl['spec'].get('cephVersion', {}).get('image', None)
num_nodes = len(self.rook_cluster.get_node_names())
spec = {}
spec['mon'] = orchestrator.ServiceDescription(
service_name='mon',
spec=ServiceSpec(
'mon',
placement=PlacementSpec(
count=cl['spec'].get('mon', {}).get('count', 1),
),
),
size=cl['spec'].get('mon', {}).get('count', 1),
container_image_name=image_name,
last_refresh=now,
)
spec['mgr'] = orchestrator.ServiceDescription(
service_name='mgr',
spec=ServiceSpec(
'mgr',
placement=PlacementSpec.from_string('count:1'),
),
size=1,
container_image_name=image_name,
last_refresh=now,
)
if not cl['spec'].get('crashCollector', {}).get('disable', False):
spec['crash'] = orchestrator.ServiceDescription(
service_name='crash',
spec=ServiceSpec(
'crash',
placement=PlacementSpec.from_string('all:true'),
),
size=num_nodes,
container_image_name=image_name,
last_refresh=now,
)
# CephFilesystems
all_fs = self.rook_cluster.rook_api_get(
"cephfilesystems/")
self.log.debug('CephFilesystems %s' % all_fs)
for fs in all_fs.get('items', []):
svc = 'mds.' + fs['metadata']['name']
if svc in spec:
continue
# FIXME: we are conflating active (+ standby) with count
active = fs['spec'].get('metadataServer', {}).get('activeCount', 1)
total_mds = active
if fs['spec'].get('metadataServer', {}).get('activeStandby', False):
total_mds = active * 2
spec[svc] = orchestrator.ServiceDescription(
service_name=svc,
spec=ServiceSpec(
svc,
placement=PlacementSpec(count=active),
),
size=total_mds,
container_image_name=image_name,
last_refresh=now,
)
# FIXME: CephObjectstores
for dd in self._list_daemons():
if dd.service_name() not in spec:
continue
spec[dd.service_name()].running += 1
return [v for k, v in spec.items()]
