For the Windows feature other than os-win dependencies, it is up to
the projects, and if they can still
support and test those without 3rd party CI, then it is okay to keep
it. This applies to any other distro-specific
features also where they might be supported by a few projects but not
all. But they should go through the
deprecation phase warning even they are not tested so that users get
the notification.
I'm worried about "might be supported by a few projects" approach here, because that would eventually block deprecating/removing Windows support from oslo. We now have some features like eventlog[1] or some logics to support Windows[2]. Should we keep these until all projects declare deprecation and removal of Windows support ? Based on my past experiences, some of our projects have been inactive for a while and haven't been aligned with this kind of global changes being made. So I feel like this would eventually mean we can't remove these implementations from oslo really. I'd rather prefer seeing common agreement across all projects, and set the expected timeline so that we can drop unmaintained codes.
[1] https://review.opendev.org/c/openstack/oslo.log/+/894235 [2] https://github.com/openstack/oslo.concurrency/blob/master/oslo_concurrency/p...
my goal in this regard would be to land the removal of both the hyperv
and vmware driver before milestone one
and perhaps even before the ptg if there is no object to it in our irc
team meeting. (this is off topic) I'm wondering if removal of vmware drivers means that we should consider deprecating and removing vmware drivers in glance and cinder, as these drivers are meant to be used with vmware virt drivers.
What you just said. IIRC, we kinda agreed on the PTG to try to avoid as
much as possible any deprecations during
the 2023.2 Bobcat release, which is a non-SLURP release, as it would be
skipped by operators fast-jumping to 2024.1,
unless someone would forward-port the deprecation note to Caracal, hence
putting the burden on someone's shoulder. My understanding was that we can deprecate features at 2023.2, as long as we don't remove these at 2024.1, (removal should be done at 2024.2 or later), though I agree that deprecating features at 2023.2 are not much useful because the removal timeline can't be changed even if we deprecate features "early".
On Mon, Sep 11, 2023 at 6:35 PM Sylvain Bauza sylvain.bauza@gmail.com wrote:
Le lun. 11 sept. 2023 à 11:03, smooney@redhat.com a écrit :
On Sun, 2023-09-10 at 20:03 -0700, Ghanshyam Mann wrote:
---- On Fri, 08 Sep 2023 05:19:13 -0700 Takashi Kajinami wrote ---
Let me bump this old thread because we still need some follow-up
about the retirement of os-win.
I noticed that some projects have not yet deprecated the
implementations dependent on os-win.I submitted a few
patches to make these implementations deprecated so that we can
smoothly remove thesein the
future. https://review.opendev.org/c/openstack/cinder/+/894237 https://review.opendev.org/c/openstack/glance/+/894236
https://review.opendev.org/c/openstack/ceilometer/+/894296It%27d
be nice if the relevant teams can review these.
My remaining question is whether we should mark all implementations
for Windows support, which are not
directlydependent on os-win[1]. Technically we can go through
individual projects and add warning logs and release
notesabout the deprecation. However I'm not too sure if that's worth
the effort. If we can agree that we remove
supportfor running OpenStack on Windows Operating System at a specific
release, then I tend to leave the ones
independentfrom os-win, unless it has any impact on user-facing items
like config options[1].
I'd like to hear any thoughts about this plan, as well as any target
release to remove Windows "host" support
globallyfrom OpenStack (maybe after 2024.1 release ?).
Thanks for marking a few of the Windows support things as deprecated.
This is the right direction for at least where
it depends on os-win. I have started completing the os-win retirement and
deps[1]. But we need to add a deprecation
warning in one cycle and then remove it in a later one (like you are
doing in the mentioned changes). We did the
same in the Nova Hyper-V driver, which was marked deprecated in the
2023.1 cycles, and I am proposing it to be
removed in the next cycle, 2024.1[2].
you bet me too it there are already two other nova cores (myself and one other) that also planned to do this after confirming with the wider team at the next ptg so this is highly likely to proceed early in the 2024.1 cycle. my goal in this regard would be to land the removal of both the hyperv and vmware driver before milestone one and perhaps even before the ptg if there is no object to it in our irc team meeting.
i was waiting for RC-1 to be cut and the dust to settle before brign this up to discuss but it seams at least 3 fo the nova core team feel this is thr correct direction to take now.
What you just said. IIRC, we kinda agreed on the PTG to try to avoid as much as possible any deprecations during the 2023.2 Bobcat release, which is a non-SLURP release, as it would be skipped by operators fast-jumping to 2024.1, unless someone would forward-port the deprecation note to Caracal, hence putting the burden on someone's shoulder.
Reinstanting my personal take then, which is, as a Nova PTL, I'm not 100% happy with taking my burden. Please, please, let's wait for 4 days and nothing or nooone will then get hurt :-)
For the Windows feature other than os-win dependencies, it is up to the
projects, and if they can still
support and test those without 3rd party CI, then it is okay to keep
it. This applies to any other distro-specific
features also where they might be supported by a few projects but not
all. But they should go through the
deprecation phase warning even they are not tested so that users get
the notification.
[1] https://review.opendev.org/q/+topic:retire-winstackers [2] https://review.opendev.org/c/openstack/nova/+/894466
-gmann
[1] Some
examples
https://github.com/openstack/ceilometer/blob/d31d4ed3574a5d19fe4b09ab2c227db... .
py#L95-L96
https://github.com/openstack/nova/blob/master/nova/virt/disk/api.py#L624-L62...
[2] event_log option in oslo.log is one good example
https://review.opendev.org/c/openstack/oslo.log/+/894235
On Sat, Jun 24, 2023 at 7:50 AM Ghanshyam Mann
gmann@ghanshyammann.com> wrote:
As there is no volunteer to maintain this project, I have proposed
the retirement
-gmann
---- On Thu, 13 Apr 2023 07:54:12 -0700 James Page wrote ---
Hi All
As announced by Lucian last November (see [0]) Cloudbase
Solutions are no longer in a position to maintain
support for running OpenStack on Windows and have also ceased operation
of their 3rd party CI for the windows support
across a number of OpenStack projects.
This situation has resulted in the Winstackers project becoming
PTL-less for the 2023.2 cycle with no volunteers
responding to the TC's call to fill this role and take this feature in
OpenStack forward (see [1]).
This is the final call for any maintainers to step forward if
this feature is important to them in OpenStack.
The last user survey in 2022 indicated that 2% of respondents
were running on Hyper-V so this might be important
enough to warrant a commitment from someone operating OpenStack on
Windows to maintain these features going forward.
Here is a reminder from Lucian's original email on the full list
of projects which are impacted in some way: *
nova hyper-v driver - in-tree plus out-of-tree compute-hyperv driver*
os-win - common Windows library for Openstack*
neutron hyperv ml2 plugin and agent* ovs on Windows and neutron ovs
agent support* cinder drivers - SMB and Windows
iSCSI* os-brick Windows connectors - iSCSI, FC, SMB, RBD* ceilometer
Windows poller* manila Windows driver* glance
Windows support* freerdp gateway
The lack of 3rd party CI for testing all of this really needs to
be addressed as well.
If no maintainers are forthcoming between now and the next PTG in
June the TC will need to officially retire the
project and start the process of removing support for Windows across
the various projects that support this operating
system in some way - either directly or through the use of os-win.
For clarity this call refers to the use of the Hyper-V
virtualisation driver and associated Windows server
components to provide WIndows based OpenStack Hypervisors and does not
relate to the ability to run Windows images as
guests on OpenStack.
Regards James
[0]
https://lists.openstack.org/pipermail/openstack-discuss/2022-November/031044...] https://lists.openstack.org/p
ipermail/openstack-discuss/2023-March/032888.html