When I implemented support for reserved store options, I somehow
overlooked the fact that some of the registered options are not really
used. This is mainly because glance does not use 'add' API of
the filesystem store but just write data directly into the store
This deprecates the ineffective options so that we can remove these
in a future release.
This introduces support for the new cinder_do_extend_attached option
which was added in the 2023.1 release. See the dependent change to find
The option has been managed by the underlying puppet-oslo module but
has not been configurable. This introduces the parameter to customize
The rbd backend implementation of glance_store requires not only
python3-rbd but also python3-rados. In CentOS the python3-rados package
is installed as a dependency of python3-rbd, but in Debian and Ubuntu
that dependency is not defined and we should install python3-rados
This fixes the missing python3-rados package in Ubuntu, by using
the meta python3-ceph package which installs all python libraries.
This parameter was deprecated during the past cycle because it is not
used by glance-api actually, and has had no effect since then.
These parameters were deprecated during the previous cycle by  and
have had no effect. These parameters are not really used by recent
... service validation was deprecated during Yoga cycle by .
The parameters were left for interface compatibility but have had no
effect since then.
The python-ceph/rbd package is required so that glance services can
access to its ceph backend. This change ensures the package resource
is tagged with the glance-support-package tag so that the package is
always installed before starting glance services.
When running glance-api behind a proxy or load balancer that hides
glance behind an extended base URL, it is necessary to provide
information about the base URL to glance-api to ensure that clients
know where to pass subsequent requests.
It turned out defining dependency for openstacklib::wsgi::apache
doesn't properly enforce resource order and the default vhost file
is not purged properly.
This change adds the more explicit dependency to enforce the order
Currently parameter description of the <module>::wsgi::apache classes
are formatted differently in individual modules, and this is making
the maintenance effort quite difficult.
This change updates the description format following the standard one
we are globally using in our modules to reduce undesired differences
The apache class is already included by the openstacklib::wsgi::apache
defined types and the apache::wsgi class and the apache::ssl class is
automatically included by the apache::vhost defined type.
This removes the redundant inclusion of these classes.
This change fixes the left over of  and makes sure the priority
parameter is an integer value, according to the type validation in
puppetlabs-apache 8.0.0 .
The puppetlabs-apache module is enforcing more strict data type
This change updates the default values to adapt to that change.
This change introduces a basic implementation to run glance-api as
an application managed by httpd+mod_wsgi, instead of a standalone
... so that users can customize that option using the native interface.
In TripleO we append additional arguments(to load glance-cache.conf)
and we need to set the option explicitly to ensure the correct file
path is used.
Also, the pipeline parameter is confusing because of its naming
different from the actual parameter in Glance, so this change renames
the parameter as well.
A recent change in os-brick , which is used by cinder backends,
requires external file locks. This patch adds support for configuring
the lock path, with a default value that matches the corresponding
lock paths used by other services that use os-brick (namely cinder
This change fixes the duplicate resources between glance::api and
glance::backend::file, which were caused by the new parameters added
to the glance::backend::multistore::file defined type in .
... so that users can avoid executing the same cron command at the same
time in multiple nodes.
Technically this can be done without this change by setting a different
cron job schedule in each node but it requires complicated parameter