Re: Spark 1.6 Release Schedule

[Sean Owen <so...@cloudera.com>]

I like the idea, but I think there's already a lot of triage backlog. Can
we more concretely address this now and during the next two weeks?

1.6.0 stats from JIRA:

344 issues targeted at 1.6.0, of which
  253 are from committers, of which
    215 are improvements/other, of which
       5 are blockers
    38 are bugs, of which
       4 are blockers
      11 are critical

Tip: It's really easy to manage saved queries for this and other things
with the free JIRA Client (http://almworks.com/jiraclient/overview.html)
that now works with Java 8.

It still looks like a lot for a point where 1.6.0 is supposed to be being
tested in theory. Lots of (most?) things that were said to be done for
1.6.0 for several months aren't going to be, and that still surprises me as
a software development practice.

Well, life is busy and chaotic out here in OSS land. I'd still like to push
even more on lightweight triage and release planning, centering around
Target Version, if only to make visible what's happening with intention and
reality:

1. Any JIRAs that seem to have been targeted at 1.6.0 by a non-committer
are untargeted, as they shouldn't be to begin with

2. This week, maintainers and interested parties review all JIRAs targeted
at 1.6.0 and untarget/retarget accordingly

3. Start of next week (the final days before an RC), non-Blocker non-bugs
untargeted, or in a few cases pushed to 1.6.1 or beyond

4. After next week, non-Blocker and non-Critical bugs are pushed, as the RC
is then late.

5. No release candidate until no Blockers are open.

6. (Repeat 1 and 2 more regularly through the development period for 1.7
instead of at the end.)

On Sat, Oct 31, 2015 at 11:25 AM, Michael Armbrust <mi...@databricks.com>
wrote:

> Hey All,
>
> Just a friendly reminder that today (October 31st) is the scheduled code
> freeze for Spark 1.6.  Since a lot of developers were busy with the Spark
> Summit last week I'm going to delay cutting the branch until Monday,
> November 2nd.  After that point, we'll package a release for testing and
> then go into the normal triage process where bugs are prioritized and some
> smaller features are allowed in on a case by case basis (if they are very
> low risk/additive/feature flagged/etc).
>
> As a reminder, release window dates are always maintained on the wiki and
> are updated after each release according to our 3 month release cadence:
>
> https://cwiki.apache.org/confluence/display/SPARK/Wiki+Homepage
>
> Thanks!
>
> Michael
>
>

Re: Spark 1.6 Release Schedule

[Michael Armbrust <mi...@databricks.com>]

Sorry for the delay due to traveling...

The branch has been cut.  At this point anything that we want to go into
Spark 1.6 will need to be cherry-picked.  Please be cautious when doing so,
and contact me if you are uncertain.

Michael

On Sun, Nov 1, 2015 at 4:16 AM, Sean Owen <so...@cloudera.com> wrote:

> I like the idea, but I think there's already a lot of triage backlog. Can
> we more concretely address this now and during the next two weeks?
>
> 1.6.0 stats from JIRA:
>
> 344 issues targeted at 1.6.0, of which
>   253 are from committers, of which
>     215 are improvements/other, of which
>        5 are blockers
>     38 are bugs, of which
>        4 are blockers
>       11 are critical
>
> Tip: It's really easy to manage saved queries for this and other things
> with the free JIRA Client (http://almworks.com/jiraclient/overview.html)
> that now works with Java 8.
>
> It still looks like a lot for a point where 1.6.0 is supposed to be being
> tested in theory. Lots of (most?) things that were said to be done for
> 1.6.0 for several months aren't going to be, and that still surprises me as
> a software development practice.
>
> Well, life is busy and chaotic out here in OSS land. I'd still like to
> push even more on lightweight triage and release planning, centering around
> Target Version, if only to make visible what's happening with intention and
> reality:
>
> 1. Any JIRAs that seem to have been targeted at 1.6.0 by a non-committer
> are untargeted, as they shouldn't be to begin with
>
> 2. This week, maintainers and interested parties review all JIRAs targeted
> at 1.6.0 and untarget/retarget accordingly
>
> 3. Start of next week (the final days before an RC), non-Blocker non-bugs
> untargeted, or in a few cases pushed to 1.6.1 or beyond
>
> 4. After next week, non-Blocker and non-Critical bugs are pushed, as the
> RC is then late.
>
> 5. No release candidate until no Blockers are open.
>
> 6. (Repeat 1 and 2 more regularly through the development period for 1.7
> instead of at the end.)
>
> On Sat, Oct 31, 2015 at 11:25 AM, Michael Armbrust <michael@databricks.com
> > wrote:
>
>> Hey All,
>>
>> Just a friendly reminder that today (October 31st) is the scheduled code
>> freeze for Spark 1.6.  Since a lot of developers were busy with the Spark
>> Summit last week I'm going to delay cutting the branch until Monday,
>> November 2nd.  After that point, we'll package a release for testing and
>> then go into the normal triage process where bugs are prioritized and some
>> smaller features are allowed in on a case by case basis (if they are very
>> low risk/additive/feature flagged/etc).
>>
>> As a reminder, release window dates are always maintained on the wiki and
>> are updated after each release according to our 3 month release cadence:
>>
>> https://cwiki.apache.org/confluence/display/SPARK/Wiki+Homepage
>>
>> Thanks!
>>
>> Michael
>>
>>
>