> It does make sense to keep SQLite on the other hand a Docker image with
all the components might just be as convenient?

I formed an actual opinion. :) IMO, SQLite is very convenient and should be
left in.

> Dropping stuff would lessen the burden of maintenance. There is a lot of
cruft that is not used. This reduces the surface for bugs and makes it
easier to do the optimizations you refer too. However, I’m not married to a
2.0 release now, but it does look fresh ;-).

What do you guys think about doing a 1.10 in January/Feb that has timezone
and RBAC (new UI), and targeting 2.0 after that?

On Mon, Dec 18, 2017 at 11:50 AM, Bolke de Bruin <[email protected]> wrote:

> Hi George,
>
> It does make sense to keep SQLite on the other hand a Docker image with
> all the components might just be as convenient?
>
> Dropping stuff would lessen the burden of maintenance. There is a lot of
> cruft that is not used. This reduces the surface for bugs and makes it
> easier to do the optimizations you refer too. However, I’m not married to a
> 2.0 release now, but it does look fresh ;-).
>
> Bolke
>
> Verstuurd vanaf mijn iPad
>
> > Op 18 dec. 2017 om 19:48 heeft George Leslie-Waksman <
> [email protected]> het volgende geschreven:
> >
> > I really do not think we should drop sqlite support. We use sqlite for
> > local testing/development; I used it when doing my initial evaluation of
> > Airflow; and there are regular comments in this group about people using
> > sqlite locally for their workflows. It feels like a critical feature for
> a
> > lot of use cases.
> >
> > I also feel like we're kind of jumping the gun on pushing for 2.0 (is it
> > just because 9 is the last single digit?). There is a lot of the code
> base
> > that's pretty hairy and a number of things that are fairly
> non-performant /
> > buggy. I don't think that we NEED to break backward compatibility to
> clean
> > things up and fix things. I don't particularly like pickle or the old
> > import style but I don't feel like supporting them is terribly onerous (I
> > could be wrong).
> >
> > New webserver, releasing the API, and even timezones are all things that
> > can be released in a non-breaking fashion.
> >
> > --George
> >
> >> On Thu, Dec 14, 2017 at 2:20 PM Alek Storm <[email protected]>
> wrote:
> >>
> >> We use sqlite for developing Airflow DAGs locally, and doing basic
> checks
> >> for syntax/import errors in our CI/CD pipeline.
> >>
> >> Alek
> >>
> >> On Thu, Dec 14, 2017 at 4:19 PM, Chris Riccomini <[email protected]
> >
> >> wrote:
> >>
> >>> Agree with Bolke that it's a good idea, but major work. I will push
> back
> >> on
> >>> this for 2.0 mostly due to time concerns. I don't want 2.0 to take
> months
> >>> and months to get out.
> >>>
> >>>> Btw: what about dropping sqlite support?
> >>>
> >>> I'm fine with this, but isn't this really useful for demos? Seems to me
> >>> that the out of the box experience from airflow is one of the things
> that
> >>> make sit so compelling.
> >>>
> >>> On Thu, Dec 14, 2017 at 2:12 PM, Bolke de Bruin <[email protected]>
> >> wrote:
> >>>
> >>>> That will take quite some work. It is a a good idea but also a major
> >>>> change. Not sure if we should target that.
> >>>>
> >>>> Btw: what about dropping sqlite support?
> >>>>
> >>>> Verstuurd vanaf mijn iPad
> >>>>
> >>>>> Op 14 dec. 2017 om 21:19 heeft Gael Magnan <[email protected]>
> >> het
> >>>> volgende geschreven:
> >>>>>
> >>>>> Hi,
> >>>>>
> >>>>> haven't been following much lately but on the import side of things,
> >>>> isn't
> >>>>> Airflow 2 the best moment to change to a pip plugin system for
> >> imports
> >>> of
> >>>>> third party stuff?
> >>>>> I.E being able to add a new type of credentials, operator etc..
> >> without
> >>>>> touching to the airflow code itself or having them in a special
> >> folder.
> >>>>>
> >>>>> Regards
> >>>>> Gael
> >>>>>
> >>>>>
> >>>>>
> >>>>> Le jeu. 14 déc. 2017 à 14:17, Chris Riccomini <[email protected]
> >>>
> >>> a
> >>>>> écrit :
> >>>>>
> >>>>>> I'm fine with sensor refactor. Added to Wiki.
> >>>>>>
> >>>>>> On Thu, Dec 14, 2017 at 11:16 AM, Chris Riccomini <
> >>>> [email protected]>
> >>>>>> wrote:
> >>>>>>
> >>>>>>> @Bolke,
> >>>>>>>
> >>>>>>>> Should we, before 2.0, start the graduation from the incubator?
> >>>>>>>
> >>>>>>> No, I'd rather keep them separate. We can certainly start
> >> graduation,
> >>>> but
> >>>>>>> I don't want to block 2.0. Can pursue them in parallel.
> >>>>>>>
> >>>>>>> On Thu, Dec 14, 2017 at 11:15 AM, Andy Hadjigeorgiou <
> >>>>>> [email protected]
> >>>>>>>> wrote:
> >>>>>>>
> >>>>>>>> Does it make sense to include sensors.py refactor in 2.0, so we
> >> can
> >>>>>>>> retire the old import structure easily and support the new sensors
> >>>>>> package
> >>>>>>>> import structure?
> >>>>>>>>
> >>>>>>>> - Andy
> >>>>>>>>
> >>>>>>>> On Thu, Dec 14, 2017 at 2:12 PM, Driesprong, Fokko
> >>>> <[email protected]
> >>>>>>>
> >>>>>>>> wrote:
> >>>>>>>>
> >>>>>>>>> Hi all,
> >>>>>>>>>
> >>>>>>>>> Good initiative. I would be happy to refactor the sensors
> >> package.
> >>> I
> >>>>>>>>> started on it but it changes a lot, all the imports will break.
> >>>>>>>>>
> >>>>>>>>> https://github.com/apache/incubator-airflow/pull/2875
> >>>>>>>>>
> >>>>>>>>> What do you guys think?
> >>>>>>>>>
> >>>>>>>>> Cheers, Fokko
> >>>>>>>>>
> >>>>>>>>> 2017-12-14 20:09 GMT+01:00 Chris Riccomini <
> >> [email protected]
> >>>> :
> >>>>>>>>>
> >>>>>>>>>> I have created a wiki here:
> >>>>>>>>>>
> >>>>>>>>>> https://cwiki.apache.org/confluence/display/AIRFLOW/Airflow+2.0
> >>>>>>>>>>
> >>>>>>>>>> To track features and progress.
> >>>>>>>>>>
> >>>>>>>>>> On Thu, Dec 14, 2017 at 11:08 AM, Chris Riccomini <
> >>>>>>>>> [email protected]>
> >>>>>>>>>> wrote:
> >>>>>>>>>>
> >>>>>>>>>>>> Re: #2: Is there a current ticket out for removing the legacy
> >>>>>>>>> import
> >>>>>>>>>>> style?
> >>>>>>>>>>>
> >>>>>>>>>>> No, I don't think so, but you can create one! :)
> >>>>>>>>>>>
> >>>>>>>>>>> On Thu, Dec 14, 2017 at 11:06 AM, Andy Hadjigeorgiou <
> >>>>>>>>>> [email protected]
> >>>>>>>>>>>> wrote:
> >>>>>>>>>>>
> >>>>>>>>>>>> This sounds great, something I'd like to see updated for 2.0
> >>>>>>>>> release (or
> >>>>>>>>>>>> before) is the Airflow documentation
> >>>>>>>>>>>> <http://airflow.readthedocs.io/en/latest/installation.html> (
> >>>>>>>>>>>> http://airflow.readthedocs.io/en/latest/installation.html).
> >> It
> >>>>>>>>> seems
> >>>>>>>>>> that
> >>>>>>>>>>>> updating the repo does not update this site - and given that
> >> we
> >>>>>>>>> will be
> >>>>>>>>>>>> removing certain deprecated features I imagine the docs will
> >>>>>> change
> >>>>>>>>>>>> substantially.
> >>>>>>>>>>>>
> >>>>>>>>>>>> Re: #2: Is there a current ticket out for removing the legacy
> >>>>>> import
> >>>>>>>>>>>> style?
> >>>>>>>>>>>> I'm happy to help drive that forward.
> >>>>>>>>>>>>
> >>>>>>>>>>>> - Andy
> >>>>>>>>>>>>
> >>>>>>>>>>>> On Thu, Dec 14, 2017 at 1:55 PM, Chris Riccomini <
> >>>>>>>>> [email protected]
> >>>>>>>>>>>
> >>>>>>>>>>>> wrote:
> >>>>>>>>>>>>
> >>>>>>>>>>>>> Hey all,
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> With 1.9.0 wrapping up soon (hopefully), there's been some
> >>>>>>>>> discussion
> >>>>>>>>>> on
> >>>>>>>>>>>>> the having the next release be Airflow 2.0 (rather than
> >> 1.10).
> >>>>>>>>> This
> >>>>>>>>>>>> would
> >>>>>>>>>>>>> allow us to break compatibility, and clean up some stuff.
> >>>>>> Proposed
> >>>>>>>>>>>> things
> >>>>>>>>>>>>> to include in 2.0 are:
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> 1) New webserver that Joy Gao has been working on.
> >>>>>>>>>>>>> 2) Remove the legacy import style that's been deprecated
> >> since
> >>>>>> at
> >>>>>>>>>> least
> >>>>>>>>>>>> 1.8
> >>>>>>>>>>>>> 3) New timzone feature
> >>>>>>>>>>>>> 4) Move API out of experimental
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> I want to keep the list fairly tight, preferably to things
> >> that
> >>>>>>>>> have
> >>>>>>>>>>>>> already been done, so that we can ship it fairly quickly (in
> >>> the
> >>>>>>>>> next
> >>>>>>>>>>>>> couple of months).
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> Does this sound like a good list?
> >>>>>>>>>>>>>
> >>>>>>>>>>>>> Cheers,
> >>>>>>>>>>>>> Chris
> >>>>>>
> >>>>
> >>>
> >>
>

Reply via email to