Skip to content

Improve DB robustness + SR destruction + size getters #95

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Wescoeur merged 2 commits into from
Aug 26, 2025
Merged

Improve DB robustness + SR destruction + size getters #95

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
237bce4
fix(linstorvolumemanager): robustify "usable size" getters
Wescoeur Aug 18, 2025
5e61b9a
feat(LinstorSR): improve DB volume robustness
Wescoeur Aug 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion drivers/LinstorSR.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@
from linstorvolumemanager import get_controller_node_name
from linstorvolumemanager import LinstorVolumeManager
from linstorvolumemanager import LinstorVolumeManagerError
from linstorvolumemanager import DATABASE_VOLUME_NAME
from linstorvolumemanager import PERSISTENT_PREFIX

LINSTOR_AVAILABLE = True
Expand Down Expand Up @@ -618,7 +619,6 @@ def create(self, uuid, size) -> None:
ips,
self._redundancy,
thin_provisioning=self._provisioning == 'thin',
auto_quorum=self._monitor_db_quorum,
logger=util.SMlog
)
self._vhdutil = LinstorVhdUtil(self.session, self._linstor)
Expand Down Expand Up @@ -677,6 +677,8 @@ def delete(self, uuid) -> None:
)

try:
if self._monitor_db_quorum:
self._linstor.set_drbd_ha_properties(DATABASE_VOLUME_NAME, enabled=False)
self._update_drbd_reactor_on_all_hosts(
controller_node_name=node_name, enabled=False
)
Expand All @@ -692,6 +694,8 @@ def delete(self, uuid) -> None:
self._update_drbd_reactor_on_all_hosts(
controller_node_name=node_name, enabled=True
)
if self._monitor_db_quorum:
self._linstor.set_drbd_ha_properties(DATABASE_VOLUME_NAME, enabled=True)
except Exception as e2:
util.SMlog(
'Failed to restart drbd-reactor after destroy fail: {}'
Expand Down Expand Up @@ -741,6 +745,9 @@ def attach(self, uuid) -> None:
opterr='no such group: {}'.format(self._group_name)
)

if self._monitor_db_quorum and self.is_master():
self._linstor.set_drbd_ha_properties(DATABASE_VOLUME_NAME)

@override
@_locked_load
def detach(self, uuid) -> None:
Expand Down
108 changes: 55 additions & 53 deletions drivers/linstorvolumemanager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -927,6 +927,34 @@ def set_auto_promote_timeout(self, volume_uuid, timeout):
.format(volume_uuid, error_str)
)

def set_drbd_ha_properties(self, volume_name, enabled=True):
"""
Set or not HA DRBD properties required by drbd-reactor and
by specific volumes.
:param str volume_name: The volume to modify.
:param bool enabled: Enable or disable HA properties.
"""

properties = {
'DrbdOptions/auto-quorum': 'disabled',
'DrbdOptions/Resource/auto-promote': 'no',
'DrbdOptions/Resource/on-no-data-accessible': 'io-error',
'DrbdOptions/Resource/on-no-quorum': 'io-error',
'DrbdOptions/Resource/on-suspended-primary-outdated': 'force-secondary',
'DrbdOptions/Resource/quorum': 'majority'
}
if enabled:
result = self._linstor.resource_dfn_modify(volume_name, properties)
else:
result = self._linstor.resource_dfn_modify(volume_name, {}, delete_props=list(properties.keys()))

error_str = self._get_error_str(result)
if error_str:
raise LinstorVolumeManagerError(
'Could not modify HA DRBD properties on volume `{}`: {}'
.format(volume_name, error_str)
)

def get_volume_info(self, volume_uuid):
"""
Get the volume info of a particular volume.
Expand Down Expand Up @@ -1735,33 +1763,21 @@ def get_all_group_names(cls, base_name):
return [cls._build_group_name(base_name), cls._build_ha_group_name(base_name)]

@classmethod
def create_sr(
cls, group_name, ips, redundancy,
thin_provisioning, auto_quorum,
logger=default_logger.__func__
):
def create_sr(cls, group_name, ips, redundancy, thin_provisioning, logger=default_logger.__func__):
"""
Create a new SR on the given nodes.
:param str group_name: The SR group_name to use.
:param set(str) ips: Node ips.
:param int redundancy: How many copy of volumes should we store?
:param bool thin_provisioning: Use thin or thick provisioning.
:param bool auto_quorum: DB quorum is monitored by LINSTOR.
:param function logger: Function to log messages.
:return: A new LinstorSr instance.
:rtype: LinstorSr
"""

try:
cls._start_controller(start=True)
sr = cls._create_sr(
group_name,
ips,
redundancy,
thin_provisioning,
auto_quorum,
logger
)
sr = cls._create_sr(group_name, ips, redundancy, thin_provisioning, logger)
finally:
# Controller must be stopped and volume unmounted because
# it is the role of the drbd-reactor daemon to do the right
Expand All @@ -1775,11 +1791,7 @@ def create_sr(
return sr

@classmethod
def _create_sr(
cls, group_name, ips, redundancy,
thin_provisioning, auto_quorum,
logger=default_logger.__func__
):
def _create_sr(cls, group_name, ips, redundancy, thin_provisioning, logger=default_logger.__func__):
# 1. Check if SR already exists.
uri = 'linstor://localhost'

Expand Down Expand Up @@ -1921,7 +1933,7 @@ def _create_sr(
try:
logger('Creating database volume...')
volume_path = cls._create_database_volume(
lin, ha_group_name, storage_pool_name, node_names, redundancy, auto_quorum
lin, ha_group_name, storage_pool_name, node_names, redundancy
)
except LinstorVolumeManagerError as e:
if e.code != LinstorVolumeManagerError.ERR_VOLUME_EXISTS:
Expand Down Expand Up @@ -2099,7 +2111,7 @@ def _get_volumes_info(self, volume_name=None):
if not self._volume_info_cache_dirty:
return self._volume_info_cache

for resource in self._get_resource_cache().resources:
def process_resource(resource):
if resource.name not in all_volume_info:
current = all_volume_info[resource.name] = self.VolumeInfo(
resource.name
Expand Down Expand Up @@ -2131,15 +2143,21 @@ def _get_volumes_info(self, volume_name=None):
):
current.virtual_size = usable_size

if current.virtual_size <= 0:
raise LinstorVolumeManagerError(
'Failed to get usable size of `{}` on `{}`'
.format(resource.name, volume.storage_pool_name)
)
try:
for resource in self._get_resource_cache().resources:
process_resource(resource)
for volume in all_volume_info.values():
if volume.virtual_size <= 0:
raise LinstorVolumeManagerError(
'Failed to get usable size of `{}`'
.format(volume.name)
)

for current in all_volume_info.values():
current.allocated_size *= 1024
current.virtual_size *= 1024
volume.allocated_size *= 1024
volume.virtual_size *= 1024
except LinstorVolumeManagerError:
self._mark_resource_cache_as_dirty()
raise

self._volume_info_cache_dirty = False
self._volume_info_cache = all_volume_info
Expand All @@ -2158,16 +2176,14 @@ def _get_volume_node_names_and_size(self, volume_name):
node_names.add(resource.node_name)

current_size = volume.usable_size
if current_size < 0:
raise LinstorVolumeManagerError(
'Failed to get usable size of `{}` on `{}`'
.format(resource.name, volume.storage_pool_name)
)

if size < 0:
if current_size > 0 and (current_size < size or size < 0):
size = current_size
else:
size = min(size, current_size)

if size < 0:
raise LinstorVolumeManagerError(
'Failed to get usable size of `{}` on `{}`'
.format(resource.name, volume.storage_pool_name)
)

return (node_names, size * 1024)

Expand Down Expand Up @@ -2691,7 +2707,7 @@ def _request_database_path(cls, lin, activate=False):

@classmethod
def _create_database_volume(
cls, lin, group_name, storage_pool_name, node_names, redundancy, auto_quorum
cls, lin, group_name, storage_pool_name, node_names, redundancy
):
try:
dfns = lin.resource_dfn_list_raise().resource_definitions
Expand Down Expand Up @@ -2775,20 +2791,6 @@ def _create_database_volume(
)
)

# We must modify the quorum. Otherwise we can't use correctly the
# drbd-reactor daemon.
if auto_quorum:
result = lin.resource_dfn_modify(DATABASE_VOLUME_NAME, {
'DrbdOptions/auto-quorum': 'disabled',
'DrbdOptions/Resource/quorum': 'majority'
})
error_str = cls._get_error_str(result)
if error_str:
raise LinstorVolumeManagerError(
'Could not activate quorum on database volume: {}'
.format(error_str)
)

# Create database and ensure path exists locally and
# on replicated devices.
current_device_path = cls._request_database_path(lin, activate=True)
Expand Down