2023.1 Series Release Notes

2023.1-eom

Bug Fixes

  • PowerFlex driver bug #2004630: Fixed regression which caused the PowerFlex (ScaleIO) connector to login multiple times, even with an existing valid token. The ScaleIO connector now caches and reuses the token.

  • Bug #2067949: Fixed issue where we try to write into a multipath device and fail since it is not ready for I/O. Now we wait until the I/O is likely to succeed. We introduced 2 new config options to make the wait time configurable:

    • wait_mpath_device_attempts

    • wait_mpath_device_interval

    These options defaults to 4 attempts and 1 second interval respectively. See help text of the config options for more information.

6.2.4

Bug Fixes

  • Dell PowerFlex driver Bug #2046810: Added new error code for PowerFlex 4.x which was causing this error. Added handling for the new error code returned from PowerFlex v4.x and allow the driver to ignore and skip a disconnect operation when the volume is not mapped.

  • FC connector bug #2051237: Fix issue with fibre channel connector scanning partial targets. We search for target information in sysfs, first in fc_transport and then in fc_report_ports.

  • Fixes an warning seen when running in a container using overlayfs.

6.2.3

New Features

  • NVMe-oF connector: Improved speed of creation of the /etc/nvme/hostnqn file.

  • NVMe-oF connector: Always write the same value in the same system for the /etc/nvme/hostnqn file in older nvme-cli versions when system UUID can be read from DMI.

  • iSCSI and FCP: Support for different SCSI addressing modes: SAM, SAM-2, and SAM-3 flat addressing. Defaults to SAM/transparent, but cinder drivers can set key addressing_mode in the connection properties to indicate other addressing modes using one of the constants from os_brick.constants.SCSI_ADDRESSING_* as the value.

Bug Fixes

  • PowerFlex driver Bug #2013749: Added new error code for powerflex 4.x which was causing this error. Added handling for the new error code returned from Powerflex 4.x when a we try to map a volume which is already mapped.

  • NVMe-oF connector bug #2035695: Fixed attaching volumes when all portals of a subsystem are reconnecting at the time of the request.

  • NVMe-oF bug #2016029: The NVMe-oF connector now creates /etc/nvme/hostid when it’s missing from the system. That way the Host ID will persist and always be the same, instead of being randomly generated.

  • NVMe-oF connector bug #2026257: Fixes _get_host_uuid for nvmeof failing when using a btrfs root fs.

  • NVMe-oF connector bug #2035811: Fixed attaching volumes on systems using newer NVMe kernel modules that present additional information in /sys/class/nvme/nvme#/address.

  • NVMe-oF connector bug #2035606: Fixed generation of hostnqn file on old nvme-cli versions.

  • NVMe-oF connector bug #1961222: Fixed support of newer NVMe CLI v2.

6.2.2

New Features

  • FC connector: Added support to force and ignore_errors parameters on disconnect_volume method.

6.2.0

New Features

  • Support separate privsep logging levels. Now services using os-brick can have different log levels for the service and os-brick’s privsep calls. The service usually uses the default “oslo_privsep.daemon” name while os-brick now uses “os_brick.privileged”.

Bug Fixes

  • Bug #1922052: Fixed missing os-brick DEBUG logs on nova compute service.