hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sushanth Sowmyan <khorg...@gmail.com>
Subject Re: Preparation for Hive-1.2 release
Date Thu, 30 Apr 2015 18:57:02 GMT
Hi Folks,

As of yesterday, we had a dummy RC0 ready, and the time draws near to
start getting all the issues set about in
https://cwiki.apache.org/confluence/display/Hive/Hive+1.2+Release+Status
to getting them resolved, and then rolling out an RC1. To do this, we
must first freeze the list over there for further inclusion requests,
and for that, I've been phasing in strictness for the bar to include
into that list.

As of right now, the bar says "bugfixes only".

As my original plan said, at 3:01PM PDT today, I had planned to raise
that bar further, to become "critical bugfixes or trivial changes
only". Since I did not send a reminder email yesterday night about
this deadline, I extend this deadline to midnight tonight (defined as
00:01 AM PDT Friday May 1st 2015) for requests for inclusion for
general purpose bugfixes.

After this deadline passes, only bugfixes that are deemed to be
product outages (i.e. no workarounds known, or a major feature of hive
is disabled), or regressions discovered during testing of the RCs will
be allowed to be included. Any jiras already present in that list will
be included. After the deadline passes, I will create a release JIRA,
and mark it blocked by all the jiras in the wiki.

Examples of (fake) bugs that will be allowed in after this deadline:

a) NPE whenever someone tries to use a union in a subquery clause =>
Product outage - only workaround might be to not use unions or
subqueries, which makes it impact a major feature of hive. Also,
probably a regression.
b) Changing an error message from "could not fnid appropriate config"
to "could not find appropriate config" => trivial, and does not carry
significant testing risk.
c) TestSchemaTool discovered to fail for postgres. => Regression.

Examples of (fake) bugs that would not be allowed for inclusion past
this deadline:
a) Implement OGRE (new fileformat, say) support in HCatalog =>
Feature, not bug, and not critical bug
b) Fix NPE when HCatalog attempts to read from OGRE (new fileformat,
say) => Same as above, but phrased to sound like a bug - easy to test
if we ask if it's a regression.
c) Formatting of output from beeline to be customizable => Feature,
not bug, and not critical bug, although smaller in scope than above.
d) Fix default for orc.compress.dummy.setting to be 42 =>
questionable. Might fly under the radar for being trivial enough if
the setting does not carry enough testing risk, but at the same time,
should have been requested for inclusion earlier.

Issues that have already been asked for inclusion in the list at
https://cwiki.apache.org/confluence/display/Hive/Hive+1.2+Release+Status
will still be honoured, and considered release blockers unless the
committer requesting inclusion and release manager agree that it can
be removed.

TL;DR : Deadline for general purpose bugfix inclusion into 1.2 is
tonight, after that, I'm going to get more strict on whether or not I
allow something in. If you already have a jira on that list, I will
wait on its fix before moving forward, so you have a ticket on the 1.2
train. That said, if you have a jira on that list, I will of course be
breathing down your necks to make sure it gets in in time so that I
can get the release process cracking. :p

Thanks for reading!
-Sushanth

On Tue, Apr 28, 2015 at 1:33 AM, Sushanth Sowmyan <khorgath@gmail.com> wrote:
> Hi Folks,
>
> As per the timeline set on
> https://cwiki.apache.org/confluence/display/Hive/Hive+1.2+Release+Status
> , the time for requests for inclusion of any new features to
> branch-1.2 is now closed. Any JIRAS already marked in the lists
> previously will be honoured and considered release blockers, but no
> new feature JIRAs will be included.
>
> The list is still open for any new bugfix jira to be added to the
> list. I will start getting RC0 ready now. The purpose of RC0 is not
> for it to be a true release-candidate, but simply to go through the
> process of creating a release, verifying instructions set to create a
> release, so that we can discover any hiccups along the way. Once that
> is done, the next deadline of note for the community is 15:01,
> Thursday, 30th Apr 2015 PDT, which is the deadline for requesting any
> bugfix jiras for inclusion.
>
> After this deadline, the list will go into further lockdown, and allow
> inclusion of only jiras already in the currently accepted lists, and
> either critical bugfixes which are product breakages without
> workarounds, or bugs discovered during testing of the RCs.
>
> Thanks,
> -Sushanth
>
>
>
> On Mon, Apr 27, 2015 at 4:51 PM, Sushanth Sowmyan <khorgath@gmail.com> wrote:
>> HIVE-10509 and HIVE-10510 have been created with the individual
>> patches that upgraded master and branch-1.2 respectively, and marked
>> as resolved.
>>
>> I'll also add it in to the HowToRelease doc to ensure that from the
>> next release onward, we create a jira for this.
>>
>> Also, friendly reminder for folks :  If you have any other jira you
>> want included for 1.2, please go and add it to
>> https://cwiki.apache.org/confluence/display/Hive/Hive+1.2+Release+Status
>>
>> Tonight midnight will be the deadline for feature include requests,
>> and this list will go on lockdown for any further requests for feature
>> inclusions. Bugfixes will still be okay to request for inclusion.
>>
>>
>>
>> On Mon, Apr 27, 2015 at 4:41 PM, Sushanth Sowmyan <khorgath@gmail.com> wrote:
>>> Also, Agreed with Sergio/Szehon, we should have a jira that's marked
>>> as committed with a patch attached for the version bump patch.
>>> Creating one.
>>>
>>> On Mon, Apr 27, 2015 at 4:39 PM, Sushanth Sowmyan <khorgath@gmail.com>
wrote:
>>>> Vinod,
>>>>
>>>> If 2.7.0 is "not-yet-ready", then we should not be bumping 1.2 to
>>>> 2.7.0. But, we should be okay to bump master(1.3) up to 2.7.0 so that
>>>> we start general-purpose testing against that.
>>>>
>>>> Please go ahead and create a jira for that, and we should be able to
>>>> bump up trunk to go against that.
>>>>
>>>> Thanks,
>>>> -Sushanth

Mime
View raw message