Hi Arina,
The three Jira’s are ready (+1, squashed, rebased, etc.):
• DRILL-5616: Add memory checks, plus minor metrics changes
(Commit ID: c7ad56b7e075e84761896c9ce660b18b09d9e49a )
• DRILL-5665: Add the option planner.force_2phase_aggr to override small inputs
(Commit ID: 1917ec084bdef74962370fbe5b253172c634c86e )
• DRILL-5669: Add a configurable option for minimum memory allocation to buffered ops
(Commit ID: 4d7a3fe3e7ae52db57b0fd1f0619c6d1d64308c8 )
Thanks,
Boaz
On 7/18/17, 3:04 AM, "Arina Yelchiyeva" <[EMAIL PROTECTED]> wrote:
Hi Parth,
no problem, I did not start preparing the release since there is one
blocker has left:
*DRILL-5669 - Boaz Ben-Zvi (fix will be ready in a couple of days)*
Kind regards
Arina
On Mon, Jul 17, 2017 at 11:26 PM, Parth Chandra <[EMAIL PROTECTED]> wrote:
> Hi Arina,
>
> I just realized that I did not push the C++ client fix ( DRILL-5659) to
> master. I have just done so.
>
> Sorry if I messed things up for the release, but we should get this fix
> in.
>
> Parth
>
> On Fri, Jul 14, 2017 at 8:15 AM, Paul Rogers <[EMAIL PROTECTED]> wrote:
>
> > I looked into one NPE on CSV. Turns out it is already fixed by previous
> > 1.11 PRs, so we’re good.
> >
> > There is another reported by a user on the user list, but I’ve not heard
> > back about getting sample data. Let’s not hold the release for that.
> >
> > - Paul
> >
> > > On Jul 14, 2017, at 2:38 AM, Arina Yelchiyeva <
> > [EMAIL PROTECTED]> wrote:
> > >
> > > *Blockers*
> > > DRILL-5659 - Parth (done)
> > > DRILL-5669 - Boaz Ben-Zvi (fix will be ready in a couple of days)
> > >
> > > *Possible blockers*
> > > NPE on CSV - Paul can you please provide an update?
> > >
> > > Kind regards
> > > Arina
> > >
> > > On Thu, Jul 13, 2017 at 2:46 PM, Arina Yelchiyeva <
> > > [EMAIL PROTECTED]> wrote:
> > >
> > >> *Blockers*
> > >> DRILL-5659 - problem is identified, Parth is working on the fix.
> > >>
> > >> *Not blockers*
> > >> DRILL-5660 - Vitalii
> > >> DRILL-5468 - Jinfeng
> > >>
> > >> *Possible blockers*
> > >> DRILL-5669 - Boaz Ben-Zvi / Paul
> > >> NPE on CSV - Paul
> > >>
> > >> Kind regards
> > >> Arina
> > >>
> > >>
> > >> On Thu, Jul 13, 2017 at 1:28 AM, Paul Rogers <[EMAIL PROTECTED]>
> wrote:
> > >>
> > >>> I withdraw my suggestion to fix DRILL-5660. Was a big hassle for us
> to
> > >>> track down the incompatibility the issue internally, but production
> > users
> > >>> probably don’t attempt to use one build of Drill with metadata files
> > >>> created from another. We can worry about the issue later only if a
> user
> > >>> actually runs into it.
> > >>>
> > >>> As suggested earlier, let’s just document this obscure case in a
> > release
> > >>> note.
> > >>>
> > >>> - Paul
> > >>>
> > >>>> On Jul 12, 2017, at 3:04 PM, Jinfeng Ni <[EMAIL PROTECTED]> wrote:
> > >>>>
> > >>>> Upgrade from 1.10 to 1.11 is irreversible only when parquet metadata
> > >>> cache
> > >>>> file is involved.
> > >>>>
> > >>>> I might be wrong, but I thought that if someone is running on
> 1.10.0,
> > >>> and
> > >>>> he wants to go back to 1.0.0, where the first version of parquet
> > >>> meatadata
> > >>>> cache was introduced, can we guarantee that Drillbit would work
> > >>> correctly?
> > >>>> I'm trying to understand what DRILL-5660 asks for is a new
> > requirement,
> > >>> or
> > >>>> an requirement being there starting from Drill 1.0.0?
> > >>>>
> > >>>> Regression failures are normally regarded as release blocker.
> > >>>>
> > >>>>
> > >>>>
> > >>>> On Wed, Jul 12, 2017 at 2:54 PM, Paul Rogers <[EMAIL PROTECTED]>
> > wrote:
> > >>>>
> > >>>>> If we don’t do DRILL-5660 for 1.11 then we don’t need to do it at
> > all,
> > >>> so
> > >>>>> that is a time savings. We could handle this via a release note
> item
> > >>> that
> > >>>>> simply states that an upgrade from 1.10 to 1.11 is irreversible: we
> > do
> > >>> not
> > >>>>> support the ability to fall back to the old version. Everyone OK
> with
> > >>> that?
> > >>>>>
> > >>>>> We are seeing some stress test regression failures that, I’m told,
> > will
> > >>>>> show up as JIRA entries shortly.
> > >>>>>
> > >>>>> - Paul
> > >>>>>
> > >>>>>> On Jul 12, 2017, at 11:05 AM, Jinfeng Ni <[EMAIL PROTECTED]> wrote:
> > >>>>>>
> > >>>>>> I put some analysis in DRILL-5468.
> > >>>>>>
> > >>>>>> Also, IMHO, DRILL-5660 is not a release blocker.
> > >>>>>>
> > >>>>>>
> > >>>>>>
> > >>>>>> On Tue, Jul 11, 2017 at 11:27 AM, Parth Chandra <
> [EMAIL PROTECTED]>
> > >>>>> wrote:
> > >>>>>>
> > >>>>>>> Re DRILL-5634:
> > >>>>>>> From the Apache commons web page [1] -
> > >>>>>>>
> > >>>>>>> Please note that Apache Commons Crypto doesn't implement the
> > >>>>> cryptographic
> > >>>>>>> algorithm such as AES directly.* It wraps to Openssl or JCE which
> > >>>>> implement
> > >>>>>>> the algorithms.*
> > >>>>>>>
> > >>>>>>>
> > >>>>>>> Since we are cleared for using Openssl and JCE, we are OK to
> > include
> > >>> the
> > >>>>>>> AES code in the UDFs.
> > >>>>>>>
> > >>>>>>>
> > >>>>>>> Parth
> > >>>>>>>
> > >>>>>>> [1]
https://commons.apache.org/proper/commons-crypto/ > > >>>>>>>
> > >>>>>>> On Tue, Jul 11, 2017 at 10:56 AM, Arina Yelchiyeva <
> > >>>>>>> [EMAIL PROTECTED]> wrote:
> > >>>>>>>
> > >>>>>>>> During today hangouts we agreed on approximate cut-off date - 14
> > >>>