Train Series Release Notes¶
Services using os-brick need to set the
lock_pathconfiguration option in their
[oslo_concurrency]section since it doesn’t have a valid default (related bug #1947370).
Bug #1865754: the
RBDConnectorclass generates a temporary configuration file to connect to Ceph. Previously, os-brick did not include a
[global]section to contain the options it sets, but with the Octopus release (15.2.0+), Ceph has begun enforcing the presence of this section marker, which dates back at least to the Hammer release of Ceph. With this release, os-brick includes the
[global]section in the generated configuration file, which should be backward-compatible at least to Ceph Hammer.
Bug #1915678: Fix unhandled exception during iscsi volume attachment with multipath enabled that resulted in the cinder-volume service becoming stuck and requiring a restart.
Bug #1944474: Fixed missing retries to reinitiate iSCSI connections with high concurrency of connections and with multipath enabled.
Bug #1947370: Fixed race conditions on iSCSI with shared targets and NVMe
Bug #1924652: Fix issue with newer multipathd implementations where path devices are kept in multipathd even after volume detachment completes, preventing it from creating a multipath device when a new device attachment is made shortly with the same volume device or the same device path.
This release fixes an issue that could cause data loss when the configuration enabling/disabling multipathing is changed on a compute when volumes are currently attached.
[bug 1862433] Fix an issue where platform id is needed to determine name of scsi disk.
Bug #1921381: Fix disconnecting volumes when the use_multipath value is changed from the connect_volume call to the disconnect_volume call.
Bug #1897787: Fix Fibre Channel not flushing volumes on detach when a multipath connection was requested on their attach, but one was not found.
Bug #1823200: Prior fixes for this bug changed the connection properties but did not take into account an upgrade scenario in which currently attached volumes had the old format connection properties and could fail on detatch with “KeyError: ‘config_group’”. This release updates the ‘scaleio’ connector to handle this situation. It is only applicable to deployments using a Dell EMC PowerFlex/VxFlex OS/ScaleIO backend.
Fix an incompatibility with ceph 13.2.0 (Mimic) or later, caused by a change in the output of
Improve WWN detection for arrays with multiple designators. (bug 1881608).
Improve iSCSI multipath detection to work even if we cannot find the volume’s WWN in sysfs. (bug 1881619).
This release corrects a problem with the fix for Bug #1823200 in the previous release. If you are running OpenStack on Python 3, you should not be impacted by this bug. If you are not using Dell EMC VxFlex OS (ScaleIO) Storage as a backend for Cinder, you do not need to upgrade to this release.
Dell EMC VxFlex OS driver: This release corrects an issue where the fix for Bug #1823200 did not run correctly when using Python 2.7. If you are interested in this release because of OSSN-0086, be aware that you must also deploy a configuration file on compute nodes, cinder nodes, and anywhere you would perform a volume attachment in your deployment. Please see the release notes for the previous release for details.
The fix for Bug #1823200 requires that a configuration file be deployed on compute nodes, cinder nodes, and anywhere you would perform a volume attachment in your deployment, when using Cinder with a Dell EMC VxFlex OS backend. See the Dell EMC VxFlex OS (ScaleIO) Storage driver documentation for details about this configuration file.
Bug #1823200: This release contains an updated connector for use with the Dell EMC VxFlex OS backend. It requires that a configuration file be deployed on compute nodes, cinder nodes, and anywhere you would perform a volume attachment in your deployment. See the Dell EMC VxFlex OS (ScaleIO) Storage driver documentation for details about the configuration file, and see OSSN-0086 for more information about the security vulnerability.
Fix an issue where SCSI LUN scans for FC were unnecessarily too broad. Now OS-Brick will not use wildcards unless it doesn’t find any target ports in sysfs and the Cinder driver doesn’t disable them.
A LUKS2 encryptor has been introduced providing support for this latest version of the Linux Unified Key Setup disk encryption format. This requires
cryptsetupversion 2.0.0 or greater.
The plain CryptsetupEncryptor is deprecated and will be removed in a future release. Existing users are encouraged to retype any existing volumes using this encryptor to the luks LuksEncryptor or luks2 Luks2Encryptor encryptors as soon as possible
Always check if we are dealing with a single WWNN Fibre Channel target, even when we receive an initiator_target_map. This allows us to exclude unconnected HBAs from our scan for storage arrays that automatically connect all target ports (due to their architecture and design) even if the Cinder driver returns the initiator_target_map, provided the target has a single WWNN. Excluding these HBAs prevents undesired volumes from being connected.