asterixdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steven Jacobs <sjaco...@ucr.edu>
Subject Re: Modules that could be removed?
Date Thu, 01 Jun 2017 15:51:55 GMT
The two main BAD issues are:

1) BAD has a single "official" manager (me). There have been some weeks
were several code reviews are breaking BAD at once, which can be hard to
manage while trying to get research done. This has been alleviated somewhat
by the couchbase team, as they've started adding BAD changes themselves for
breaking changes (thanks!). But my concern is that experiments will have 0
"official" managers. I think if we did this then we need to strictly
require that the person who breaks the experiments build has to make the
experiments fix.

2) Maybe this is not completed related to BAD being a separate repository,
but BAD changes linger in review mode for months. The current Asterix-BAD
change set has been around for about a month. The Asterix side has gone
without comment for more than a week, and the BAD side has yet to receive
any comments. Since BAD is the focus of my work, by the time changes get
into master my local branch is several changes ahead. I'm still highly in
favor of the 72-hour rule, which I think got tabled.

Steven

On Thu, Jun 1, 2017 at 8:24 AM Michael Blow <mblow.apache@gmail.com> wrote:

> +1 on understanding the BAD issues.  CB folks are regularly making
> coordinated changes across two, three+ repos, and things work pretty well.
> We do have a way less painful way of linking changes instead of ref spec,
> but other than that things are probably not too dissimilar mechanics-wise.
>
> Our biggest issue is that unlike BAD breakages, there's not easy visibility
> by way of the -1 vote on Gerrit for the CB extension.  If experiments does
> move out (and gets an indicator on Gerrit), perhaps CB extension can
> provide feedback there as well.
>
> On Thu, Jun 1, 2017 at 10:53 AM Mike Carey <dtabass@gmail.com> wrote:
>
> > Likewise - I thought so too - so we should figure out what isn't working
> > with the extensions model!
> >
> > (We do need a good model for things that don't belong in the main open
> > source system but that we want to run tests on regularly so that we're
> > hyper-aware of breakages.)
> >
> > Note that I've always anecdotally gotten the impression that the
> > experiments package is kind of a mess/hack and makes use of things that
> > aren't necessarily intended to be used.  I would much much, much rather
> > that performance-related packages live outside the system and be banned
> > from doing things that normal clients cannot do...  I think this would
> > be a lot better for the integrity of the system (as otherwise it creates
> > weird dependencies, non-guaranteed assumptions/contracts, and also keeps
> > us from having a good/safe answer to outside cries for ways to replicate
> > our reported performance results).
> >
> > Just my $0.005.
> >
> > Cheers,
> >
> > Mike
> >
> >
> > On 5/31/17 10:39 PM, Till Westmann wrote:
> > > Hmm, I thought that the BAD model wasn’t working too badly … little do
> I
> > > know.
> > >
> > > On the other hand it seems that the module isn’t actually used - at
> > > least it
> > > seems that there were no changes in the module related to experiments
> > > since
> > > we merged repositories. Is that right? Or has somebody worked on that
> > > module
> > > recently?
> > >
> > > Cheers,
> > > Till
> > >
> > > On 31 May 2017, at 22:20, Steven Jacobs wrote:
> > >
> > >> My two cents on this is that the BAD model doesn't work well, so I'd
> > >> personally be against following it. I would vote to keep it inside as
> > >> well.
> > >> Steven
> > >>
> > >> On Wed, May 31, 2017 at 10:02 PM abdullah alamoudi <
> bamousaa@gmail.com>
> > >> wrote:
> > >>
> > >>> If that is the case (which makes sense), then keeping it inside and
> > >>> cleaning it up might be the better approach.
> > >>>
> > >>>
> > >>>> On May 31, 2017, at 8:45 PM, Ildar Absalyamov <
> > >>> ildar.absalyamov@gmail.com> wrote:
> > >>>>
> > >>>> The real challenge is that separate repo should not the treated
as
> > >>> /dev/null for old code.
> > >>>> I believe that committing changes, that causes this repo to be
> broken
> > >>> *must* be treated as -1 Jenkins bit.
> > >>>>
> > >>>>> On May 31, 2017, at 16:41, Yingyi Bu <buyingyi@gmail.com>
wrote:
> > >>>>>
> > >>>>> +1.
> > >>>>>
> > >>>>> Best,
> > >>>>> Yingyi
> > >>>>>
> > >>>>> On Wed, May 31, 2017 at 4:22 PM, Till Westmann <tillw@apache.org>
> > >>> wrote:
> > >>>>>
> > >>>>>> I read this discussion that Ildar’s suggestion to move
> > >>> asterix-experiments
> > >>>>>> to a different repo would be a solution that works for
everybody.
> > >>>>>>
> > >>>>>> Did I read that correctly or did someone disagree with
such an
> > >>> approach?
> > >>>>>>
> > >>>>>> Cheers,
> > >>>>>> Till
> > >>>>>>
> > >>>>>>
> > >>>>>> On 31 May 2017, at 16:14, Mike Carey wrote:
> > >>>>>>
> > >>>>>> This should be decided by more knowlegeable folks than
me - it
> just
> > >>> seems
> > >>>>>>> "weird" to me that that's in the main code base and
not outside
> > >>>>>>> (as a
> > >>> super
> > >>>>>>> useful client package).
> > >>>>>>>
> > >>>>>>>
> > >>>>>>> On 5/31/17 4:12 PM, Chris Hillery wrote:
> > >>>>>>>
> > >>>>>>>> If we move asterix-experiments out, we would need
to change the
> > >>> current
> > >>>>>>>> Perf job at Couchbase. However I think the change
would
> > >>>>>>>> actually make
> > >>>>>>>> that
> > >>>>>>>> job less complicated, at least if that allowed
one to use a
> > >>> pre-compiled
> > >>>>>>>> Asterix distribution for experiments rather than
requiring the
> > >>>>>>>> full
> > >>>>>>>> source
> > >>>>>>>> code to be available.
> > >>>>>>>>
> > >>>>>>>> Ceej
> > >>>>>>>> aka Chris Hillery
> > >>>>>>>>
> > >>>>>>>> On Wed, May 31, 2017 at 2:53 PM, Yingyi Bu <buyingyi@gmail.com>
> > >>> wrote:
> > >>>>>>>>
> > >>>>>>>> The code in asterix-experiments is not well organized,
e.g.:
> > >>>>>>>>>
> > >>>>>>>>> https://github.com/apache/asterixdb/tree/master/asterixdb/
> > >>>>>>>>>
> > >>>
> asterix-experiments/src/main/java/org/apache/asterix/experiment/builder
> > >>>>>>>>> https://github.com/apache/asterixdb/tree/master/asterixdb/
> > >>>>>>>>>
> > >>>
> asterix-experiments/src/main/java/org/apache/asterix/experiment/report
> > >>>>>>>>>
> > >>>>>>>>> Can anyone who actively uses asterix-experiments
clean up the
> > >>>>>>>>> code?
> > >>>>>>>>>
> > >>>>>>>>> An alternative approach would be to put it
in a separate github
> > >>> repo,
> > >>>>>>>>> if
> > >>>>>>>>> it is relatively independent of AsterixDB.
> > >>>>>>>>>
> > >>>>>>>>> Best,
> > >>>>>>>>> Yingyi
> > >>>>>>>>>
> > >>>>>>>>>
> > >>>>>>>>> On Wed, May 31, 2017 at 12:31 PM, Ian Maxon
<imaxon@uci.edu>
> > >>>>>>>>> wrote:
> > >>>>>>>>>
> > >>>>>>>>> I also would vouch for keeping asterix-experiments
in.
> > >>> 'asterix-tools'
> > >>>>>>>>>> and the others could probably go.
> > >>>>>>>>>>
> > >>>>>>>>>> On Wed, May 31, 2017 at 11:58 AM, Taewoo
Kim
> > >>>>>>>>>> <wangsaeu@gmail.com>
> > >>>>>>>>>> wrote:
> > >>>>>>>>>>
> > >>>>>>>>>>> @Ildar: +1
> > >>>>>>>>>>>
> > >>>>>>>>>>> Best,
> > >>>>>>>>>>> Taewoo
> > >>>>>>>>>>>
> > >>>>>>>>>>> On Wed, May 31, 2017 at 11:57 AM, Ildar
Absalyamov <
> > >>>>>>>>>>> ildar.absalyamov@gmail.com> wrote:
> > >>>>>>>>>>>
> > >>>>>>>>>>> I found Young-Seok’s asterix-experiment
package useful for
> > >>> everyone,
> > >>>>>>>>>>>>
> > >>>>>>>>>>> who
> > >>>>>>>>>
> > >>>>>>>>>> is doing any kind of experiments.
> > >>>>>>>>>>>> Can we instead make an ‘asterix-contrib’
repo and move it
> > >>>>>>>>>>>> there,
> > >>> the
> > >>>>>>>>>>>>
> > >>>>>>>>>>> same
> > >>>>>>>>>>
> > >>>>>>>>>>> way we did with asterix-bad?
> > >>>>>>>>>>>> We can also launch an automated
build in Jenkins to verify
> it
> > >>> builds
> > >>>>>>>>>>>> against master, again the same
way BAD is working. This
> > >>>>>>>>>>>> package
> > >>> does
> > >>>>>>>>>>>>
> > >>>>>>>>>>> not
> > >>>>>>>>>
> > >>>>>>>>>> have a lot of dependencies, so it will
be fairly painless to
> > >>> maintain
> > >>>>>>>>>>>>
> > >>>>>>>>>>> it.
> > >>>>>>>>>>
> > >>>>>>>>>>> On May 31, 2017, at 09:06, Yingyi Bu
<buyingyi@gmail.com>
> > >>>>>>>>>>> wrote:
> > >>>>>>>>>>>>>
> > >>>>>>>>>>>>> Hi dev,
> > >>>>>>>>>>>>>
> > >>>>>>>>>>>>>   I wonder if the following
potentially obsolete modules
> > >>>>>>>>>>>>> could
> > >>> be
> > >>>>>>>>>>>>>
> > >>>>>>>>>>>> moved
> > >>>>>>>>>>
> > >>>>>>>>>>> out of the AsterixDB code base:
> > >>>>>>>>>>>>>   -- asterix-experiment
> > >>>>>>>>>>>>>   -- asterix-tools
> > >>>>>>>>>>>>>   -- hyracks-dist
> > >>>>>>>>>>>>>   -- hyracks-sever
> > >>>>>>>>>>>>>
> > >>>>>>>>>>>>>   Any thoughts?
> > >>>>>>>>>>>>>
> > >>>>>>>>>>>>> Best,
> > >>>>>>>>>>>>> Yingyi
> > >>>>>>>>>>>>>
> > >>>>>>>>>>>> Best regards,
> > >>>>>>>>>>>> Ildar
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>>>>>>>>>
> > >>>>
> > >>>> Best regards,
> > >>>> Ildar
> > >>>>
> > >>>
> > >>>
> >
> >
>

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