Open Stack

Hi all,

Thanks for the info.  I would be very keen to meet with the Senlin
team again this summit and continue to discuss VM HA.  I think there
are different sides to the problem, since we need to allow policies
regarding failure detection and automatic recovery to be specified by
both users and operators:

  - Recovery from failures at the compute node level is mainly the
    concern of the operators (typically requires fencing which could
    affect multiple projects), but we still need to notify the users
    somehow.

  - Recovery from failures of a VM's virtual hardware might be a
    shared responsibility between both operators and users.

  - Failures within the VM are purely the concern of the user, and the
    operator does not need to know about them.

The specs which we are currently working on should clarify this in
much more detail.  It would be great to have input from the Senlin
team, since the goal is a modular, converged upstream design which
should allow all implementations to integrate together in a clean and
future-proof way.

I will try to attend at least one of the Senlin presentations, but
hopefully we can also find a timeslot to meet and discuss in more
detail.  Looking forward to seeing you all soon!

Regards,
Adam

Ethan Lynn <xuanlangjian at gmail.com> wrote:
> Hi,
> We will share our solution through this event "On Building an Auto-healing Resource Cluster using Senlin <https://www.openstack.org/summit/barcelona-2016/summit-schedule/events/15037/on-building-an-auto-healing-resource-cluster-using-senlin>” on Barcelona Summit.
> If the timeslot is available for us, we are definitely willing to discuss more details about how we(senlin) cooperate with the PWG for achieving the HA solution.
> 
> Best Regards,
> Ethan Lynn
> xuanlangjian at gmail.com
> 
> > On Oct 19, 2016, at 16:36, Sam P <sam47priya at gmail.com> wrote:
> > Hi,
> > I was there when we had this discussion with Senlin team in last
> > summit at Austin.
> > If you are attending to this summit, hope we can discuss more in details.
> > PWG have some sessions on this topic.
> > --- Regards,
> > Sampath
> > 
> > On Sat, Oct 15, 2016 at 11:31 AM, Qiming Teng
> > <tengqim at linux.vnet.ibm.com> wrote:
> >> Hi,
> >> 
> >> Team had discussions with Adam on previous summits, but senlin is
> >> curently not deeply involved into the effort due to 1) not well educated
> >> about the direction of VM-HA in that context; 2) lack of resource from
> >> senlin team.
> >> 
> >> However, we do feel VM-HA a strong requirement from many enterprise
> >> customers. Senlin provides a health policy for users to specify how VM
> >> failures should be detected and how they should be recovered. For
> >> failure detection, we support polling and listeners to VM lifecycle
> >> events today. We planned to support LBaaS based detection but haven't
> >> landed due to bugs in Octavia. For recovery, the default action is
> >> 'rebuild', we are also adding 'reboot', 'recreate', 'evacuate' as other
> >> options. We are also planning to add fencing support before doing
> >> recovery.
> >> 
> >> As for Masakari, we heard of that name from Adam but haven't looked
> >> deeply into the details.
> >> 
> >> So ... that is the high-availability use case senlin is trying to
> >> support. Other use cases we want to support include auto-scaling,
> >> load-balancing, cross-region deployment, cross-AZ deployment, blue-green
> >> deployment, rolling-upgrade, instanteneous provisioning, container
> >> clustering, bare-metal clustering etc.
> >> 
> >> Please feel free to raise any questions on Senlin. We are open to any
> >> suggestions/comments on how to improve the service so that it is useful,
> >> usable and eventually widely used.
> >> 
> >> Thanks.
> >> 
> >> Regards,
> >>  Qiming
> >> 
> >> On Fri, Oct 14, 2016 at 02:44:56PM +0000, Peter Chadwick wrote:
> >>> Qiming - One area that we have a question about is how Senlin interacts
> >>> with other projects such as Masakari. We have created a user story
> >>> around HA: https://review.openstack.org/#/c/289469/8/user-stories/propo
> >>> sed/ha_vm.rst that a cross-project team has been using to define specs
> >>> for HA capabilities. It would be great if the Senlin team can take part
> >>> in those discussions.
> >>> 
> >>> I am copying Adam Spiers on this email as he has been one of the
> >>> developers working on the cross product HA team.
> >>> 
> >>> Pete
> >>> 
> >>> 
> >>> On Fri, 2016-10-14 at 21:34 +0800, Qiming Teng wrote:
> >>>> Hi,
> >>>> 
> >>>> My name is Qiming Teng, PTL for the senlin project [1] during Mitaka
> >>>> and
> >>>> Newton cycle. I was the liaison for product working group [2] for
> >>>> senlin
> >>>> project. Until recently, I wasn't contacted by the product working
> >>>> group
> >>>> for the multi-release roadmap [3] or things like that.
> >>>> 
> >>>> On seeing the latest roadmap slides, senlin was again not listed in
> >>>> the
> >>>> whole picture. Team was surprised by this and we are wondering if
> >>>> there
> >>>> are things we missed to get the message out. I'm new to this group.
> >>>> But
> >>>> since I'm stepping down as PTL for Ocata cycle, I will have more
> >>>> cycles
> >>>> working with people from other teams to make senlin a good citizen in
> >>>> the community.
> >>>> 
> >>>> Can anyone help guide me on next steps, how can we get senlin
> >>>> included?
> >>>> Senlin has released its 2.0.0 version with official OpenStack
> >>>> release.
> >>>> 
> >>>> Thanks.
> >>>> 
> >>>> Regards,
> >>>>  Qiming
> >>>> 
> >>>> 
> >>>> [1] https://wiki.openstack.org/wiki/Senlin
> >>>> [2] https://wiki.openstack.org/wiki/CrossProjectLiaisons
> >>>> [3] https://wiki.openstack.org/wiki/ProductTeam/MultiRelease_Roadmap
> >>>> [4]
> >>>> https://www.openstack.org/assets/software/newton/Community-Generated-
> >>>> OpenStack-Roadmap-Newton-Release-v3.pdf
> >>>> 
> >>>> 
> >>>> _______________________________________________
> >>>> Product-wg mailing list
> >>>> Product-wg at lists.openstack.org
> >>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/product-wg
> >>>> 
> >>> --
> >>> Pete Chadwick
> >>> Director of Product Management
> >>> Cloud and Systems Management
> >>> SUSE
> >>> pchadwick at suse.com
> >>> (M) +1.617.281.2847
> >>> 
> >>> www.suse.com
> >> 
> >> 
> >> _______________________________________________
> >> Product-wg mailing list
> >> Product-wg at lists.openstack.org
> >> http://lists.openstack.org/cgi-bin/mailman/listinfo/product-wg
> > 
> > _______________________________________________
> > Product-wg mailing list
> > Product-wg at lists.openstack.org
> > http://lists.openstack.org/cgi-bin/mailman/listinfo/product-wg
>