I don't think we have any real data, but I would expect that most
attendees will be more familiar with Ceph. But I don't think that should
limit us... I like the idea of a talk that digs into the dissoance between
something stateful like ceph and kuberetes' declarative model.
To your earlier questions, Blaine: I'll saw something brief about rook
being the way to do ceph on kubernetes, but won't go much detail beyond
what it is. I think there's also a question of how we want to schedule
things: we could do these talks first and the tutorial to follow, or the
reverse. Or do one talk before and one after. I don't have any intuition
as to which would be preferable...
sage
On Mon, 13 Jan 2020, Blaine Gardner wrote:
> It sounds like Sagy and I have different expectations for the audience. I expect to see mostly existing Ceph users for whom Kubernetes is a new and unknown thing, and Sagy expects existing Kubernetes users for whom Ceph is a more unknown thing. Do Sage or
Mike Perez have any data to suggest whether either or both of these demographics will be represented enough to merit talking to one or both?
>
> Blaine
> ________________________________
> From: Sagy Volkov <svolkov@redhat.com>
> Sent: Monday, January 13, 2020 11:17
> To: Blaine Gardner <BlGardner@suse.com>; Sebastien Han <shan@redhat.com>
> Cc: Sage Weil <sweil@redhat.com>; cephalocon-seoul-2020-pc@ceph.io <cephalocon-seoul-2020-pc@ceph.io>
> Subject: Re: Rook talks for Cephalocon
>
> Yap, it would be good to understand what each of us submitted.
> I think a phone call will serve this purpose.
> In general, I believe that Ceph have a lot of new audience looking at it because of Rook. A lot of these folks are experienced k8s users that have been "living" in the k8s stateless world (where k8s excel) and are now feeling comfortable moving or writing
stateful applications for k8s (where k8s does not do very well). My goal was bringing the storage knowledge, and how to look at storage, how to measure storage, for these people and explaining all this with Rook/Ceph is perfect as you got a very solid, stable,
long running storage system with an operator (Rook) to manage it in the k8s domain.
>
> -sv
>
> On Mon, Jan 13, 2020 at 10:55 AM Blaine Gardner <BlGardner@suse.com<mailto:BlGardner@suse.com>> wrote:
> Correction: "I [don't] consider the talk I submitted intro-y or advanced but somewhere in between."
> ________________________________
> From: Blaine Gardner <BlGardner@suse.com<mailto:BlGardner@suse.com>>
> Sent: Monday, January 13, 2020 10:51
> To: Sebastien Han <shan@redhat.com<mailto:shan@redhat.com>>; Sagy Volkov <svolkov@redhat.com<mailto:svolkov@redhat.com>>
> Cc: Sage Weil <sweil@redhat.com<mailto:sweil@redhat.com>>; cephalocon-seoul-2020-pc@ceph.io<mailto:cephalocon-seoul-2020-pc@ceph.io> <cephalocon-seoul-2020-pc@ceph.io<mailto:cephalocon-seoul-2020-pc@ceph.io>>
> Subject: Re: Rook talks for Cephalocon
>
> Seb,
>
> I don't mind collaborating. I do want to clarify some things though. It would be helpful to understand the topics each of us have submitted so that we have an idea of what we all had in mind. And it would be helpful to have Sage or Mike talk about what their
vision is for Rook topics overall.
>
> I consider the talk I submitted intro-y or advanced but somewhere in between. I anticipated having a Rook tutorial and expected that to be the base intro to Rook for most people. The topic I submitted was to develop and present a set of best-practices for
running Rook (especially for Ceph users who may not be thoroughly familiar with K8s) as a next step following a topic like "what is Rook, and how do I get started". (Things like setting resource limits, labeling nodes, and some basic performance considerations).
I'm not sure if a Rook intro is needed before the tutorial session given that a basic "what is Rook, and what is Ceph doing with it" is probably going to be a topic of the all-attendee talks at the beginning of the con.
>
> It does sound like the topic Sagy submitted may cover a little bit of what I submitted in the realm of performance, which might leave more time for non-performance-related best practices.
>
> In my estimation, the community would be served by:
>
> * Basic Rook introduction in the all-attendee headline talks
> * This will certainly happen, but I don't know to what degree
> * Do we want a session with more intro to Rook before the tutorial?
> * i.e, was the intro in the headline talks sufficient to get users to the tutorial?
> * Rook interactive tutorial
> * Best practices for running a more enterprise-ready Rook cluster
> * Mid-level guide for starting to think about Rook in production
> * Do we want a more advanced performance and failure-recovery session?
> * Will there be enough users who are probably already running/evaluating Rook and want specific performance info?
> * Or will it be better to go a little more mid-level and combine this with a best-practices session?
> * OR ... is this misinformed, misunderstanding Sage's explanation of Sagy's topic correctly and we need to think about this all very differently?
>
> Blaine
> ________________________________
> From: Sebastien Han <shan@redhat.com<mailto:shan@redhat.com>>
> Sent: Monday, January 13, 2020 10:13
> To: Sagy Volkov <svolkov@redhat.com<mailto:svolkov@redhat.com>>
> Cc: Sage Weil <sweil@redhat.com<mailto:sweil@redhat.com>>; Blaine Gardner <BlGardner@suse.com<mailto:BlGardner@suse.com>>; cephalocon-seoul-2020-pc@ceph.io<mailto:cephalocon-seoul-2020-pc@ceph.io> <cephalocon-seoul-2020-pc@ceph.io<mailto:cephalocon-seoul-2020-pc@ceph.io>>
> Subject: Re: Rook talks for Cephalocon
>
> Blaine, wanna do a joined talk?
>
> Thanks!
>
> S้bastien Han
> Senior Principal Software Engineer, Storage Architect
>
> "Always give 100%. Unless you're giving blood."
>
> On Mon, Jan 13, 2020 at 6:10 PM Sagy Volkov <svolkov@redhat.com<mailto:svolkov@redhat.com>> wrote:
> >
> > I'll be happy to combine, how long will each of these sessions be?
> >
> > Thanks,
> >
> > -sv
> >
> > On Mon, Jan 13, 2020 at 10:06 AM Sage Weil <sweil@redhat.com<mailto:sweil@redhat.com>> wrote:
> >>
> >> Hi Blaine, Seb, Sagy,
> >>
> >> You all submitted Rook talks for Seoul. Blaine's and Seb's looked like
> >> general intro talks, and Sagy's abstract is looking more at how to achieve
> >> certain resiliency levels, how to think about failures, etc.
> >>
> >> We'd like to combine this into 2 talk slots. Presumably that's one that's
> >> more intro-ey, and one that's more advanced. Can the three of you work
> >> together on that?
> >>
> >> Note that there is also a 1.5 hour rook tutorial session, proposed by Ian
> >> Choi (MSFT) and John Haan (SK Telecom). Do any of you know them? We
> >> should also reach out to them to offer any review/assistance with the
> >> tutorial.
> >>
> >> Thanks!
> >> sage
> >>
> >
> >
> > --
> >
> > -sv
>
>
>
> --
>
> -sv
>