Open Stack

+1

I'm against this on a theoretical basis as I don't think backporting 
features/architectural changes is good practice for a stable branch.

People will say this will increase stability (no longer lose namespaces 
on container restart), in reality the chances are it will decrease 
stability on the branch at least for some time after it merges due to 
the inherent nature of new features, combined with our limited ability 
to verify new changes in the gate.

So far the evidence for needing to do this is anecdotal at best. While I 
don't doubt some operators will complain about having to docker exec to 
work with namespaces, it doesn't strike me as critical enough to shift 
effort away from the Mitaka release.

All that being said if other cores strongly agree with the change and 
commit to seeing it through, I trust them enough to make it work. Which 
is why I'm +1.

On 21/02/16 16:18, Michał Jastrzębski wrote:
> So for thin containers, as opposed to data containers, there is no
> migration script needed whatsoever. All it takes is to tear down
> neutron-agents and start thin containers.
>
> On 21 February 2016 at 06:47, Jeffrey Zhang <zhang.lei.fly at gmail.com> wrote:
>> I like the thin container idea, and I am +1 too. But the only concern is
>> that we MUST provide a robust migrate script( or Ansible role task) to do
>> the convert stuff. Doesn't we have enough time for this?
>>
>> On Sun, Feb 21, 2016 at 3:44 PM, Michal Rostecki <mrostecki at mirantis.com>
>> wrote:
>>>
>>> On 02/20/2016 05:39 PM, Steven Dake (stdake) wrote:
>>>>
>>>> Sam,
>>>>
>>>> I seem to recall Paul was not in favor, so there was not a majority of
>>>> cores there.  There were 6 core reviewers at the midcycle, and if you
>>>> only count kolla-core (which at this time I do for policy changes) that
>>>> means we had a vote of 5.  We have 11 core reviewers, so we need a vote
>>>> of 6+ for simple majority. I was also sort of –1 because it is an
>>>> exception, but I do agree the value is warranted.  I believe I expressed
>>>> at  the midcycle that I was –1 to the idea, atleast until the broader
>>>> core review team voted.  If I wasn't clear on that, I apologize.
>>>>
>>>> I'll roll with the community on this one unless I have to tie break –
>>>> then groan :)
>>>>
>>>> That is why a decision was made by the group to take this to the mailing
>>>> list.
>>>>
>>>> Regards
>>>> -steve
>>>>
>>>> From: Sam Yaple <samuel at yaple.net <mailto:samuel at yaple.net>>
>>>> Reply-To: "sam at yaple.net <mailto:sam at yaple.net>" <sam at yaple.net
>>>> <mailto:sam at yaple.net>>, "OpenStack Development Mailing List (not for
>>>> usage questions)" <openstack-dev at lists.openstack.org
>>>> <mailto:openstack-dev at lists.openstack.org>>
>>>> Date: Saturday, February 20, 2016 at 9:32 AM
>>>> To: "OpenStack Development Mailing List (not for usage questions)"
>>>> <openstack-dev at lists.openstack.org
>>>> <mailto:openstack-dev at lists.openstack.org>>
>>>> Subject: Re: [openstack-dev] [kolla][vote] port neutron thin containers
>>>> to stable/liberty
>>>>
>>>>      I was under the impression we did have a majority of cores in favor
>>>>      of the idea at the midcycle. But if this is a vote-vote, then I am a
>>>>      very strong +1 as well. This is something operators will absolutely
>>>>      want and and need.
>>>>
>>>>      Sam Yaple
>>>>
>>>>      On Sat, Feb 20, 2016 at 4:27 PM, Michał Jastrzębski
>>>>      <inc007 at gmail.com <mailto:inc007 at gmail.com>> wrote:
>>>>
>>>>          Strong +1 from me. This have multiple benefits:
>>>>          Easier (aka possible) debugging of networking in running envs
>>>> (not
>>>>          having tools like tcpdump at your disposal is a pain) - granted,
>>>>          there
>>>>          are ways to get this working without thin containers but require
>>>>          fair
>>>>          amount of docker knowledge.
>>>>          Docker daemon restart will not break routers - currently with
>>>> docker
>>>>          restart container with namespace dies and we lose our routers
>>>> (they
>>>>          will migrate using HA, but well, still a networking downtime).
>>>> This
>>>>          will no longer be the case so...
>>>>          Upgrades with no vm downtime whatsoever depends on this one.
>>>>          If we could deploy liberty code with all these nice stuff, I'd be
>>>>          happier person;)
>>>>
>>>>          Cheers,
>>>>          Michal
>>>>
>>>>          On 20 February 2016 at 07:40, Steven Dake (stdake)
>>>>          <stdake at cisco.com <mailto:stdake at cisco.com>> wrote:
>>>>          > Just clarifying, this is not a "revote" - there were not enough
>>>> core
>>>>          > reviewers in favor of this idea at the Kolla midcycle, so we
>>>> need to have a
>>>>          > vote on the mailing list to sort out this policy decision of
>>>> managing
>>>>          > stable/liberty.
>>>>          >
>>>>          > Regards,
>>>>          > -steve
>>>>          >
>>>>          >
>>>>          > From: Steven Dake <stdake at cisco.com <mailto:stdake at cisco.com>>
>>>>          > Reply-To: "OpenStack Development Mailing List (not for usage
>>>> questions)"
>>>>          > <openstack-dev at lists.openstack.org
>>>>          <mailto:openstack-dev at lists.openstack.org>>
>>>>          > Date: Saturday, February 20, 2016 at 6:28 AM
>>>>          > To: "OpenStack Development Mailing List (not for usage
>>>> questions)"
>>>>          > <openstack-dev at lists.openstack.org
>>>>          <mailto:openstack-dev at lists.openstack.org>>
>>>>          > Subject: [openstack-dev] [kolla][vote] port neutron thin
>>>> containers to
>>>>          > stable/liberty
>>>>          >
>>>>          > Folks,
>>>>          >
>>>>          > There were not enough core reviewers to pass a majority
>>>> approval of the
>>>>          > neutron thin container backport idea, so we separated it out
>>>> from fixing
>>>>          > stable/liberty itself.
>>>>          >
>>>>          > I am going to keep voting open for *2* weeks this time.  The
>>>> reason for the
>>>>          > two weeks is I would like a week of discussion before people
>>>> just blindly
>>>>          > vote ;)
>>>>          >
>>>>          > Voting begins now and concludes March 4th.  Since this is a
>>>> policy decision,
>>>>          > no veto votes are permitted, just a +1 and a  -1.  Abstaining
>>>> is the same as
>>>>          > voting –1.
>>>>          >
>>>
>>>
>>> I'm +1, but under condition that we will provide some script to migrate
>>> from supervisord-container to thin-containers (even if such a script will
>>> bring risk of downtime of the cloud).
>>>
>>>
>>> __________________________________________________________________________
>>> OpenStack Development Mailing List (not for usage questions)
>>> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>>
>>
>> --
>> Jeffrey Zhang
>> Blog: http://xcodest.me
>>
>> __________________________________________________________________________
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>