Current Series Release Notes

21.0.0-13

Upgrade Notes

  • Support for Python 3.10 has been removed. Now Python 3.11 is the minimum version supported.

Deprecation Notes

  • The [DEFAULT] use_forwarded_for option has been deprecated. Use the [oslo_middleware] enable_proxy_headers_parsing option instead.

Bug Fixes

  • Fixes an issue where the OpenStack-API-Version response header was incorrectly included in 406 Not Acceptable error responses when a client requested an unsupported API microversion. The header is now properly suppressed to comply with API guidelines. LP#1646357

  • Corrects the service type string returned in the OpenStack-API-Version header during successful API responses. Previously, it incorrectly echoed the internal project name (masakari). It now correctly returns the official API service type (instance-ha).

  • If coordination is NOT configured, synchronized becomes effectively a no-op across multiple API workers, which can lead to duplicate notifications due to race conditions.

    When [coordination] backend_url is not set, Masakari can now apply a small random delay before creating notifications. This significantly reduces, but does not fully eliminate, the probability of duplicate notifications being created.

    The workaround is controlled by the [DEFAULT] notification_delay_max option. Default is 5, set it to 0 to disable the workaround.

    This workaround is skipped entirely when coordination is configured. Operators should treat coordination as the proper fix.

    LP#2028450

  • Fixes an issue during COMPUTE_HOST recovery where transient failures from Nova or Keystone API calls could leave instance evacuation records in PENDING while the notification was still marked finished. Masakari now reports the host recovery as failed when any instance evacuation cannot be accounted for, so operators can retry or investigate the recovery instead of seeing a successful notification with missing instance results.

    Unlock failures during evacuation cleanup are now logged without escaping the instance recovery worker.

    LP#2158101