sma: initial support for volume connection via discovery

This patch adds support for connecting volumes via discovery service.
The user specifies a volume UUID/GUID and a list of discovery endpoints,
which are then used to start the discovery service on and attach the
volume to a device.

SMA will keep track of the attached volumes and will also refcount the
connections to discovery services.  So if two volumes are attached using
the same discovery endpoint, it'll be disconnected only after both of
them are detached.

Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ie8ea50a2a784cf0db8a5953234c6bb2b68685d7c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12413
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This commit is contained in:
Konrad Sztyber 2022-04-27 14:30:00 +02:00 committed by Tomasz Zawadzki
parent 20e3fb3687
commit 0c2b10f250
4 changed files with 256 additions and 5 deletions

View File

@ -4,16 +4,18 @@ from multiprocessing import Lock
import grpc
import logging
from .device import DeviceException
from .volume import VolumeException, VolumeManager
from .proto import sma_pb2 as pb2
from .proto import sma_pb2_grpc as pb2_grpc
class StorageManagementAgent(pb2_grpc.StorageManagementAgentServicer):
def __init__(self, config):
def __init__(self, config, client):
addr, port = config['address'], config['port']
self._devices = {}
self._server = grpc.server(futures.ThreadPoolExecutor(max_workers=1))
self._server.add_insecure_port(f'{addr}:{port}')
self._volume_mgr = VolumeManager(client, config['discovery_timeout'])
pb2_grpc.add_StorageManagementAgentServicer_to_server(self, self._server)
def _grpc_method(f):
@ -43,19 +45,37 @@ class StorageManagementAgent(pb2_grpc.StorageManagementAgentServicer):
pass
return None
def _cleanup_volume(self, volume_id, existing):
if volume_id is None or existing:
return
try:
self._volume_mgr.disconnect_volume(volume_id)
except VolumeException:
logging.warning('Failed to cleanup volume {volume_id}')
@_grpc_method
def CreateDevice(self, request, context):
response = pb2.CreateDeviceResponse()
volume_id, existing = None, False
try:
if request.HasField('volume'):
volume_id, existing = self._volume_mgr.connect_volume(request.volume)
manager = self._find_device_by_name(request.WhichOneof('params'))
if manager is None:
raise DeviceException(grpc.StatusCode.INVALID_ARGUMENT,
'Unsupported device type')
response = manager.create_device(request)
except DeviceException as ex:
# Now that we know the device handle, mark the volume as attached to
# that device
if volume_id is not None:
self._volume_mgr.set_device(volume_id, response.handle)
except (DeviceException, VolumeException) as ex:
self._cleanup_volume(volume_id, existing)
context.set_details(ex.message)
context.set_code(ex.code)
except NotImplementedError:
self._cleanup_volume(volume_id, existing)
context.set_details('Method is not implemented by selected device type')
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
return response
@ -69,6 +89,8 @@ class StorageManagementAgent(pb2_grpc.StorageManagementAgentServicer):
raise DeviceException(grpc.StatusCode.NOT_FOUND,
'Invalid device handle')
device.delete_device(request)
# Remove all volumes attached to that device
self._volume_mgr.disconnect_device_volumes(request.handle)
except DeviceException as ex:
context.set_details(ex.message)
context.set_code(ex.code)
@ -80,15 +102,23 @@ class StorageManagementAgent(pb2_grpc.StorageManagementAgentServicer):
@_grpc_method
def AttachVolume(self, request, context):
response = pb2.AttachVolumeResponse()
volume_id, existing = None, False
try:
if not request.HasField('volume'):
raise VolumeException(grpc.StatusCode.INVALID_ARGUMENT,
'Missing required field: volume')
volume_id, existing = self._volume_mgr.connect_volume(request.volume,
request.device_handle)
device = self._find_device_by_handle(request.device_handle)
if device is None:
raise DeviceException(grpc.StatusCode.NOT_FOUND, 'Invalid device handle')
device.attach_volume(request)
except DeviceException as ex:
except (DeviceException, VolumeException) as ex:
self._cleanup_volume(volume_id, existing)
context.set_details(ex.message)
context.set_code(ex.code)
except NotImplementedError:
self._cleanup_volume(volume_id, existing)
context.set_details('Method is not implemented by selected device type')
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
return response
@ -100,6 +130,7 @@ class StorageManagementAgent(pb2_grpc.StorageManagementAgentServicer):
device = self._find_device_by_handle(request.device_handle)
if device is not None:
device.detach_volume(request)
self._volume_mgr.disconnect_volume(request.volume_id)
except DeviceException as ex:
context.set_details(ex.message)
context.set_code(ex.code)

View File

@ -0,0 +1,2 @@
from .volume import VolumeException
from .volume import VolumeManager

View File

@ -0,0 +1,217 @@
import grpc
import ipaddress
import logging
import uuid
from dataclasses import dataclass
from spdk.rpc.client import JSONRPCException
from ..common import format_volume_id
from ..proto import sma_pb2
log = logging.getLogger(__name__)
class VolumeException(Exception):
def __init__(self, code, message):
self.code = code
self.message = message
class Volume:
def __init__(self, volume_id, device_handle, discovery_services):
self.volume_id = volume_id
self.discovery_services = discovery_services
self.device_handle = device_handle
class VolumeManager:
def __init__(self, client, discovery_timeout):
self._client = client
# Discovery service map (name -> refcnt)
self._discovery = {}
# Volume map (volume_id -> Volume)
self._volumes = {}
self._discovery_timeout = int(discovery_timeout * 1000)
def _get_discovery_info(self):
try:
with self._client() as client:
return client.call('bdev_nvme_get_discovery_info')
except JSONRPCException:
raise VolumeException(grpc.StatusCode.INTERNAL,
'Failed to retrieve discovery service status')
def _compare_trid(self, trid1, trid2):
return (trid1['trtype'].lower() == trid2['trtype'].lower() and
trid1['traddr'].lower() == trid2['traddr'].lower() and
trid1['trsvcid'].lower() == trid2['trsvcid'].lower() and
trid1['adrfam'].lower() == trid2['adrfam'].lower())
def _get_adrfam(self, traddr):
try:
return 'ipv{}'.format(ipaddress.ip_address(traddr).version)
except ValueError:
raise VolumeException(grpc.StatusCode.INVALID_ARGUMENT,
'Invalid traddr')
def _get_volume_bdev(self, volume_id, timeout):
try:
with self._client() as client:
return client.call('bdev_get_bdevs',
{'name': volume_id,
'timeout': timeout})[0]
except JSONRPCException:
return None
def _start_discovery(self, trid, hostnqn):
try:
# Use random UUID as name
name = str(uuid.uuid4())
log.debug(f'Starting discovery service {name}')
with self._client() as client:
client.call('bdev_nvme_start_discovery',
{'name': name,
'wait_for_attach': True,
'attach_timeout_ms': self._discovery_timeout,
'hostnqn': hostnqn,
**trid})
self._discovery[name] = 1
return name
except JSONRPCException:
raise VolumeException(grpc.StatusCode.INTERNAL,
'Failed to start discovery')
def _stop_discovery(self, name):
refcnt = self._discovery.get(name)
log.debug(f'Stopping discovery service {name}, refcnt={refcnt}')
if refcnt is None:
# Should never happen
log.warning('Tried to stop discovery using non-existing name')
return
# Check the refcount to leave the service running if there are more volumes using it
if refcnt > 1:
self._discovery[name] = refcnt - 1
return
del self._discovery[name]
try:
with self._client() as client:
client.call('bdev_nvme_stop_discovery',
{'name': name})
log.debug(f'Stopped discovery service {name}')
except JSONRPCException:
raise VolumeException(grpc.StatusCode.INTERNAL,
'Failed to stop discovery')
def connect_volume(self, params, device_handle=None):
""" Connects a volume through a discovery service. Returns a tuple (volume_id, existing):
the first item is a volume_id as str, while the second denotes whether the selected volume
existed prior to calling this method.
"""
volume_id = format_volume_id(params.volume_id)
if volume_id is None:
raise VolumeException(grpc.StatusCode.INVALID_ARGUMENT,
'Invalid volume ID')
if volume_id in self._volumes:
volume = self._volumes[volume_id]
if device_handle is not None and volume.device_handle != device_handle:
raise VolumeException(grpc.StatusCode.ALREADY_EXISTS,
'Volume is already attached to a different device')
return volume_id, True
discovery_services = set()
try:
# First start discovery connecting to specified endpoints
for req_ep in params.nvmf.discovery.discovery_endpoints:
info = self._get_discovery_info()
trid = {'trtype': req_ep.trtype,
'traddr': req_ep.traddr,
'trsvcid': req_ep.trsvcid,
'adrfam': self._get_adrfam(req_ep.traddr)}
name = None
for discovery in info:
if self._compare_trid(discovery['trid'], trid):
name = discovery['name']
break
if next(filter(lambda r: self._compare_trid(r['trid'], trid),
discovery['referrals']), None):
name = discovery['name']
break
if name is not None:
# If we've already attached a discovery service, it probably means that the user
# specified a referred address
if name not in discovery_services:
refcnt = self._discovery.get(name)
if refcnt is None:
log.warning('Found a discovery service missing from internal map')
refcnt = 0
self._discovery[name] = refcnt + 1
else:
name = self._start_discovery(trid, params.nvmf.hostnqn)
discovery_services.add(name)
# Now check if a bdev with specified volume_id exists, give it 1s to appear
bdev = self._get_volume_bdev(volume_id, timeout=1000)
if bdev is None:
raise VolumeException(grpc.StatusCode.NOT_FOUND,
'Volume could not be found')
# Check subsystem's NQN if it's specified
if params.nvmf.subnqn:
nvme = bdev.get('driver_specific', {}).get('nvme', [])
# The NVMe bdev can report multiple subnqns, but they all should be the same, so
# don't bother checking more than the first one
subnqn = next(iter(nvme), {}).get('trid', {}).get('subnqn')
if subnqn != params.nvmf.subnqn:
raise VolumeException(grpc.StatusCode.INVALID_ARGUMENT,
'Unexpected subsystem NQN')
# Finally remember that volume
self._volumes[volume_id] = Volume(volume_id, device_handle, discovery_services)
except Exception as ex:
for name in discovery_services:
try:
self._stop_discovery(name)
except Exception:
log.warning(f'Failed to cleanup discovery service: {name}')
raise ex
return volume_id, False
def disconnect_volume(self, volume_id):
"""Disconnects a volume connected through discovery service"""
id = format_volume_id(volume_id)
if id is None:
raise VolumeException(grpc.StatusCode.INVALID_ARGUMENT,
'Invalid volume ID')
# Return immediately if the volume is not on our map
volume = self._volumes.get(id)
if volume is None:
return
# Delete the volume from the map and stop the services it uses
for name in volume.discovery_services:
try:
self._stop_discovery(name)
except Exception:
# There's no good way to handle this, so just print an error message and
# continue
log.error(f'Failed to stop discovery service: {name}')
del self._volumes[id]
def set_device(self, volume_id, device_handle):
"""Marks a previously connected volume as being attached to specified device. This is only
necessary if the device handle is not known at a time a volume is connected.
"""
id = format_volume_id(volume_id)
if id is None:
raise VolumeException(grpc.StatusCode.INVALID_ARGUMENT,
'Invalid volume ID')
volume = self._volumes.get(id)
if volume is None:
raise VolumeException(grpc.StatusCode.NOT_FOUND,
'Volume could not be found')
if volume.device_handle is not None and volume.device_handle != device_handle:
raise VolumeException(grpc.StatusCode.ALREADY_EXISTS,
'Volume is already attached to a different device')
volume.device_handle = device_handle
def disconnect_device_volumes(self, device_handle):
"""Disconnects all volumes attached to a specific device"""
volumes = [i for i, v in self._volumes.items() if v.device_handle == device_handle]
for volume_id in volumes:
self.disconnect_volume(volume_id)

View File

@ -32,7 +32,8 @@ def parse_argv():
parser.add_argument('--config', '-c', help='Path to config file')
defaults = {'address': 'localhost',
'socket': '/var/tmp/spdk.sock',
'port': 8080}
'port': 8080,
'discovery_timeout': 10.0}
# Merge the default values, config file, and the command-line
args = vars(parser.parse_args())
config = parse_config(args.get('config'))
@ -116,7 +117,7 @@ if __name__ == '__main__':
# Wait until the SPDK process starts responding to RPCs
wait_for_listen(client, timeout=60.0)
agent = sma.StorageManagementAgent(config)
agent = sma.StorageManagementAgent(config, client)
devices = [sma.NvmfTcpDeviceManager(client)]
devices += load_plugins(config.get('plugins') or [], client)