hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jonathan Hsieh <...@cloudera.com>
Subject Re: 0.94 Backports.
Date Tue, 12 Feb 2013 03:45:50 GMT
Sorry if I misinterpreted.  If it was commit speed is the concern I
generally agree -- but this patch had a +1 from one of the owners
(jimmy) so committing it wasn't unreasonable.  I think the bigger
point is that we need to be more vigilant about compatibility,
especially with late point releases.

Jon.

On Mon, Feb 11, 2013 at 7:36 PM, Andrew Purtell <apurtell@apache.org> wrote:
> I didn't say the revert is not reasonable.
>
>
>
> On Mon, Feb 11, 2013 at 7:32 PM, Jonathan Hsieh <jon@cloudera.com> wrote:
>
>> Andrew,
>>
>> I agree if a new patch under discussion and a commit was made -- bad
>> form to commit.
>>
>> However, a revert within 24 hours seems reasonable, especially if done
>> by the original committer.   A revert is done to undo harm (failed
>> build, massive test failures, or serious bug found with nontrivial
>> effort to repair).
>>
>> Personally, I'd rather have a bad commit, a revert and then a single
>> clean commit (even if this last one came a few days later) instead of
>> a bad commit, and then a series of addendums that come a few days
>> later.
>>
>> Jon.
>>
>> On Mon, Feb 11, 2013 at 7:20 PM, Andrew Purtell <apurtell@apache.org>
>> wrote:
>> > I'm also concerned that the revert happened here while discussion was
>> > ongoing. Given the latest comments on the issue, this could have been
>> > handled by a new issue that replaces the offending code with reflection.
>> I
>> > don't care about the revert per se but would ask we avoid making changes
>> > out from under a discussion until the matter is resolved with consensus.
>> We
>> > will have cleaner revision history and less churn overall as a result. I
>> > know many of us have to-do lists of HBase JIRAs to retire, but there is
>> no
>> > need to be hasty. Because we are all busy, unnecessary commit speed makes
>> > it more likely mistakes like this will slip by review in the first place
>> > too.
>> >
>> > For your consideration.
>> >
>> >
>> > On Mon, Feb 11, 2013 at 5:40 PM, Ted <yuzhihong@gmail.com> wrote:
>> >
>> >> No.
>> >> The release was cut before the revert.
>> >>
>> >> On Feb 11, 2013, at 5:35 PM, Enis Söztutar <enis.soz@gmail.com> wrote:
>> >>
>> >> > I was going to +1 the release, with the following checks I did:
>> >> > - Checked md5 sums
>> >> > - Checked gpg signature (gpg --verify )
>> >> > - Checked included documentation book.html, etc.
>> >> > - Running unit tests (passed on unsecure, secure)
>> >> > - Started in local mode, run LoadTestTool
>> >> > - integration tests (not working fully properly, but expected since
>> >> > HBASE-7521 is not in yet)
>> >> >
>> >> > I guess this means that the release candidate has sunk, right?
>> >> > Enis
>> >> >
>> >> >
>> >> >
>> >> > On Mon, Feb 11, 2013 at 4:59 PM, Stack <stack@duboce.net> wrote:
>> >> >
>> >> >> Good catch Jon.
>> >> >>
>> >> >> We need to be vigilant here all.
>> >> >>
>> >> >> Incompatibilities cost users and those following behind us as they
>> burn
>> >> >> cycles doing gymnastics trying to get over the incompatibility
-- if
>> it
>> >> is
>> >> >> possible to get over the incompatibility at all.  They make us
look
>> bad.
>> >> >> Worse, usually the incompatibility is found months later after
we
>> have
>> >> all
>> >> >> moved on and have long forgot what it was we committed (and even
>> why) so
>> >> >> all the more reason to be on the look out at commit time.
>> >> >>
>> >> >> St.Ack
>> >> >>
>> >> >>
>> >> >> On Mon, Feb 11, 2013 at 4:48 PM, Jonathan Hsieh <jon@cloudera.com>
>> >> wrote:
>> >> >>
>> >> >>> Apache Hat: What a particular vendor chooses to puts in its
releases
>> >> >>> shouldn't affect an Apache release and especially if we are
breaking
>> >> >>> the
>> >> >>> project's versioning / compatibility rules.
>> >> >>>
>> >> >>> Jon.
>> >> >>>
>> >> >>> On Mon, Feb 11, 2013 at 4:32 PM, Ted Yu <yuzhihong@gmail.com>
>> wrote:
>> >> >>>> I downloaded hadoop-0.20.2+737 from Cloudera website.
>> >> >>>>
>> >> >>>> I found getShortUserName() in UserGroupInformation
>> >> >>>>
>> >> >>>> Haven't checked other 0.20.x source code yet.
>> >> >>>>
>> >> >>>> FYI
>> >> >>>>
>> >> >>>> On Mon, Feb 11, 2013 at 4:24 PM, Jonathan Hsieh <jon@cloudera.com>
>> >> >>> wrote:
>> >> >>>>
>> >> >>>>> Hey guys, I saw HBASE-7814 [1] -- a backport committed
to 0.94
>> that
>> >> >>>>> makes HBase 0.94 now require Hadoop 1.0 (instead of
the older
>> >> >>>>> hadoops).  This was supposed to be a new requirement
for hbase
>> >> 0.96.0.
>> >> >>>>> [2]
>> >> >>>>>
>> >> >>>>> Are we ok with making the next 0.94 upgrade incompatible?
  (And
>> if
>> >> we
>> >> >>>>> are we need to release note this kind of stuff).
>> >> >>>>>
>> >> >>>>> Jon.
>> >> >>>>>
>> >> >>>>> [1] https://issues.apache.org/jira/browse/HBASE-7814
>> >> >>>>>
>> >> >>>>> [2]
>> >> >>
>> >>
>> http://mail-archives.apache.org/mod_mbox/hbase-dev/201210.mbox/%3CCADcMMgHtqx73JztE4schY04iqs9NPZP3u84HM2SM7iCL6r80mQ@mail.gmail.com%3E
>> >> >>>>>
>> >> >>>>>
>> >> >>>>> On Fri, Feb 8, 2013 at 11:56 AM, Enis Söztutar <
>> enis.soz@gmail.com>
>> >> >>> wrote:
>> >> >>>>>> The backporting situation for 0.94 is an exception
it seems,
>> because
>> >> >>> of
>> >> >>>>> the
>> >> >>>>>> fact that 96 is so late. But until 96 comes out,
we can keep up
>> the
>> >> >>>>> current
>> >> >>>>>> approach. It has worked mostly for the time being.
>> >> >>>>>>
>> >> >>>>>> Enis
>> >> >>>>>>
>> >> >>>>>>
>> >> >>>>>> On Thu, Feb 7, 2013 at 5:20 PM, Andrew Purtell
<
>> apurtell@apache.org
>> >> >>>
>> >> >>>>> wrote:
>> >> >>>>>>
>> >> >>>>>>> That said, let's make sure every backport has
meaningful
>> >> >>> justification
>> >> >>>>>>> (determined by consensus).
>> >> >>>>>>>
>> >> >>>>>>>
>> >> >>>>>>> On Thu, Feb 7, 2013 at 5:19 PM, Andrew Purtell
<
>> >> >> apurtell@apache.org>
>> >> >>>>>>> wrote:
>> >> >>>>>>>
>> >> >>>>>>>> -1 until we have an actual stable 0.96
release.
>> >> >>>>>>>>
>> >> >>>>>>>>
>> >> >>>>>>>> On Thu, Feb 7, 2013 at 3:15 PM, Elliott
Clark <
>> eclark@apache.org
>> >> >>>
>> >> >>>>> wrote:
>> >> >>>>>>>>
>> >> >>>>>>>>> Lately there have been a lot of issues
being committed to
>> trunk
>> >> >>> and
>> >> >>>>>>>>> also back-ported to 0.94 (I've done
it myself too).  Since
>> we're
>> >> >>> so
>> >> >>>>> far
>> >> >>>>>>>>> into 0.94's release cycle should we
think about not allowing
>> >> >> minor
>> >> >>>>>>>>> features
>> >> >>>>>>>>> and code clean ups to be back-ported
?
>> >> >>>>>>>
>> >> >>>>>>> --
>> >> >>>>>>> Best regards,
>> >> >>>>>>>
>> >> >>>>>>>   - Andy
>> >> >>>>>>>
>> >> >>>>>>> Problems worthy of attack prove their worth
by hitting back. -
>> Piet
>> >> >>> Hein
>> >> >>>>>>> (via Tom White)
>> >> >>>>>
>> >> >>>>>
>> >> >>>>>
>> >> >>>>> --
>> >> >>>>> // Jonathan Hsieh (shay)
>> >> >>>>> // Software Engineer, Cloudera
>> >> >>>>> // jon@cloudera.com
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>> --
>> >> >>> // Jonathan Hsieh (shay)
>> >> >>> // Software Engineer, Cloudera
>> >> >>> // jon@cloudera.com
>> >> >>
>> >>
>> >
>> >
>> >
>> > --
>> > Best regards,
>> >
>> >    - Andy
>> >
>> > Problems worthy of attack prove their worth by hitting back. - Piet Hein
>> > (via Tom White)
>>
>>
>>
>> --
>> // Jonathan Hsieh (shay)
>> // Software Engineer, Cloudera
>> // jon@cloudera.com
>>
>
>
>
> --
> Best regards,
>
>    - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein
> (via Tom White)



-- 
// Jonathan Hsieh (shay)
// Software Engineer, Cloudera
// jon@cloudera.com

Mime
View raw message