phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Geoffrey Jacoby <gjac...@apache.org>
Subject Re: [DISCUSS] Omid and Tephra podlings
Date Mon, 28 Oct 2019 20:08:36 GMT
+1. I agree with Vincent and James. So long as we support transactions in
Phoenix, we're bound to keep at least one transaction manager working, and
we'd need to contribute at least compatibility and bug fixes. That's
probably more easily done under our umbrella if the Omid and Tephra
communities have become inactive.

If we were sure that most of our users would use one over the other, we
could potentially just adopt one, but since they have different strengths
and drawbacks, that might never be the case. (If I remember right, Lars
Hofhansl did some perf testing awhile back that showed Tephra better for
fewer, longer transactions and Omid better for more frequent short ones.)

At least for now, probably better to take on both.

Geoffrey

On Fri, Oct 25, 2019 at 6:14 PM Vincent Poon <vincentpoon@apache.org> wrote:

> +1
> I don't think pulling these under Phoenix would change their volume of
> activity in the near term.
> However, when/if usage of transactions with Phoenix increases, we'd want a
> path forward to be able to make modifications where necessary to these
> projects.
> This seems the best way to ensure there's an expedient way to do that.
>
> On Fri, Oct 25, 2019 at 10:36 AM James Taylor <jamestaylor@apache.org>
> wrote:
>
> > +1 to pulling these projects into Phoenix. We’ve already done some work
> in
> > those projects and so have some familiarity with them. We also have some
> > overlap in the committers with Omid. At a minimum we’d need to create new
> > compat modules when necessary for future HBase releases. The alternative
> > would be to rip out the transaction support which would be a mistake
> IMHO.
> >
> > On Fri, Oct 25, 2019 at 8:08 AM Josh Elser <elserj@apache.org> wrote:
> >
> > > I share your same hesitation. I'm worried about us adopting code that
> no
> > > one intends to actually maintain.
> > >
> > > My immediate concern is whether adoption of these codebases would
> impact
> > > the PMC's ability to develop on Phoenix -- I think there's a path
> > > forward to avoid that. Is that sufficient to say we "should" adopt
> them?
> > > I dunno.
> > >
> > > On 10/24/19 5:37 PM, Andrew Purtell wrote:
> > > > Tephra and Omid are interesting in that they serve a similar function
> > - a
> > > > transaction oracle - but scale differently per different choices in
> > > design,
> > > > so one is more appropriate for some kinds of transactional workloads
> vs
> > > the
> > > > other. It's akin to secondary indexing, there is not an index type
> that
> > > > fits all use cases. If you are going to consider one, you should
> > consider
> > > > both. That said my guess is you will find eventually one is the
> > 'winner'
> > > > per second order measures like contributions or user issues. This
> > should
> > > be
> > > > fine. As separate repositories they can move at their own speed and
> > only
> > > > consume bandwidth as usage and uptake actually demands.
> > > >
> > > > For what it's worth I'd vote as PMC +0 on accepting these code bases.
> > '+'
> > > > because Phoenix transactional indexes are a promising feature, and
> > could
> > > be
> > > > compelling, and they need one of these transaction oracles. '0'
> because
> > > it
> > > > would be unfair to commit someone else's time. I'm not around here
> > > much...
> > > > but may be around more going forward.
> > > >
> > > > On Thu, Oct 24, 2019 at 10:14 AM Josh Elser <elserj@apache.org>
> wrote:
> > > >
> > > >> Hiya folks,
> > > >>
> > > >> There's a discussion[1] on general@incubator about the Omid and
> > Tephra
> > > >> podlings, their decrease in volume (commits, discussion, activity),
> > and
> > > >> what to do about them. If you'd like to contribute to that
> discussion,
> > > >> please watch on general@incubator and the dev lists for those
> > podlings.
> > > >>
> > > >> One idea that seems to have resonated was that the Phoenix PMC could
> > > >> "adopt" the codebases for Omid and Tephra.
> > > >>
> > > >> While this is by no means a "done decision", but I thought it would
> be
> > > >> good for us to think about this, decide if it's something we think
> we
> > > >> want to entertain, and how would would technically do this.
> > > >>
> > > >> As far as a PMC goes, we are allowed to have multiple projects under
> > one
> > > >> PMC. We could move the tephra and omid repositories under the
> control
> > of
> > > >> our PMC, and manage them just like we do phoenix,
> phoenix-connectors,
> > > >> phoenix-queryserver, etc.
> > > >>
> > > >> Thankfully, with the work of the transaction abstraction layer, we
> > > >> shouldn't be in any position where Phoenix development would get
> > "stuck"
> > > >> by work that needed to be done in Omid or Tephra.
> > > >>
> > > >> What do folks think? Is this a good idea? Do we have enough interest
> > to
> > > >> keep the codebases healthy?
> > > >>
> > > >> - Josh
> > > >>
> > > >> [1]
> > > >>
> > > >>
> > >
> >
> https://lists.apache.org/thread.html/692a030a27067c20b9228602af502199cd4d80eb0aa8ed6461ebe1ee@%3Cgeneral.incubator.apache.org%3E
> > > >>
> > > >
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message