You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by Ted Yu <yu...@gmail.com> on 2011/07/28 18:47:39 UTC
HBase 0.92 branch
Hi,
I want to conduct a small poll on when we should create 0.92 branch.
Since build 2047, TRUNK build has been on a steady decline. I believe even
if this trend is reversed, it is hard to guarantee that TRUNK build would be
healthy in the future.
Now that the issues on http://s.apache.org/x4 have come down to 18, I think
we should consider branching 0.92 soon (maybe after HFile v2 and HBASE-4027
go in and we fix the broken build).
After branching, we can focus on the following:
1. every checkin to 0.92 branch shouldn't break its build. This requires the
committer to perform at least one complete test suite run before checking
in.
2. patches for issues on http://s.apache.org/x4 are allowed to go into 0.92
branch. New issues outside the list have to collect 3 +1 from committers
before going in.
0.92 branch would contain coprocessor, HFile v2 and HBASE-4027.
We should let people try it out soon rather than later.
Thanks for your attention.
Re: HBase 0.92 branch
Posted by Stack <st...@duboce.net>.
A bunch of the majors' have patches against them that need
review/commit. HBASE-3842, "CP Compaction API" looks like it needs to
be done for 0.92. Should it be a blocker (Would be a pain having to
do this post-0.92)? Similar with HBASE-4132, some api additions wanted
by Dhruba.
St.Ack
On Fri, Jul 29, 2011 at 9:40 AM, Stack <st...@duboce.net> wrote:
> On Fri, Jul 29, 2011 at 9:36 AM, Stack <st...@duboce.net> wrote:
>>> Shall we branch when we have a successful TRUNK build after the three
>>> blockers on http://s.apache.org/x4, HFile V2 and HBASE-4027 get checked in ?
>>>
>>
>> I'm fine w/ branching after the blockers are fixed.. I think a couple
>> of our criticals should graduate to blockers. Let me do quick review.
>>
>
> I promoted three criticals to blockers (one is nearly done). The rest
> of the criticals don't look too bad, like we could leave them to 0.94.
> St.Ack
>
Re: HBase 0.92 branch
Posted by Stack <st...@duboce.net>.
On Fri, Jul 29, 2011 at 9:36 AM, Stack <st...@duboce.net> wrote:
>> Shall we branch when we have a successful TRUNK build after the three
>> blockers on http://s.apache.org/x4, HFile V2 and HBASE-4027 get checked in ?
>>
>
> I'm fine w/ branching after the blockers are fixed.. I think a couple
> of our criticals should graduate to blockers. Let me do quick review.
>
I promoted three criticals to blockers (one is nearly done). The rest
of the criticals don't look too bad, like we could leave them to 0.94.
St.Ack
Re: HBase 0.92 branch
Posted by Stack <st...@duboce.net>.
On Thu, Jul 28, 2011 at 5:13 PM, Ted Yu <yu...@gmail.com> wrote:
>>> Our practice is RTC with one +1 other than patch creator being sufficient
> for commit, and CTR for trivial changes. I'm -1 on changing this.
> Fair.
>
I think Ted was just trying to put a bit of commit friction else
fellas like me will keep piling on the features and the 0.92 will
never go out.
But yeah, 3 +1s per commit is a little onerous.
> Shall we branch when we have a successful TRUNK build after the three
> blockers on http://s.apache.org/x4, HFile V2 and HBASE-4027 get checked in ?
>
I'm fine w/ branching after the blockers are fixed.. I think a couple
of our criticals should graduate to blockers. Let me do quick review.
St.Ack
Re: HBase 0.92 branch
Posted by Andrew Purtell <ap...@apache.org>.
I was talking about feature branches. So, no, I do not support making a feature branch at this time. What feature would that be?
Release branches and trunk should maintain the same convention we have right now: RTC with one +1 from a committer not the originator of the patch, or CTR for changes marked as trivial.
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)
>________________________________
>From: Ted Yu <yu...@gmail.com>
>To: dev@hbase.apache.org; Andrew Purtell <ap...@apache.org>
>Sent: Thursday, July 28, 2011 5:13 PM
>Subject: Re: HBase 0.92 branch
>
>>> Our practice is RTC with one +1 other than patch creator being sufficient
>for commit, and CTR for trivial changes. I'm -1 on changing this.
>Fair.
>
>Shall we branch when we have a successful TRUNK build after the three
>blockers on http://s.apache.org/x4, HFile V2 and HBASE-4027 get checked in ?
>
>Thanks
>
>On Thu, Jul 28, 2011 at 11:10 AM, Andrew Purtell <ap...@apache.org>wrote:
>
>> > New issues outside the list have to collect 3 +1 from committers before
>> going in.
>>
>> -1
>>
>> Do we actually have bylaws?
>>
>> Our practice is RTC with one +1 other than patch creator being sufficient
>> for commit, and CTR for trivial changes. I'm -1 on changing this.
>>
>> Hadoop core recently ran a vote to increase the threshold for RTC to three
>> +1s for commits that represent a branch merge. I support this for HBase. So
>> if people think some big "new issue" needs additional review, then we make a
>> feature branch for it and require three +1s for merge commit.
>>
>> Best regards,
>>
>>
>> - Andy
>>
>> Problems worthy of attack prove their worth by hitting back. - Piet Hein
>> (via Tom White)
>>
>>
>> >________________________________
>> >From: Stack <st...@duboce.net>
>> >To: dev@hbase.apache.org
>> >Sent: Thursday, July 28, 2011 10:04 AM
>> >Subject: Re: HBase 0.92 branch
>> >
>> >On Thu, Jul 28, 2011 at 9:47 AM, Ted Yu <yu...@gmail.com> wrote:
>> >> Since build 2047, TRUNK build has been on a steady decline. I believe
>> even
>> >> if this trend is reversed, it is hard to guarantee that TRUNK build
>> would be
>> >> healthy in the future.
>> >
>> >I think we can fix it.
>> >
>> >> Now that the issues on http://s.apache.org/x4 have come down to 18, I
>> think
>> >> we should consider branching 0.92 soon (maybe after HFile v2 and
>> HBASE-4027
>> >> go in and we fix the broken build).
>> >
>> >I suppose I'd like the issues to go to zero before we branched but I
>> >can go along w/ the above; if we held to my way of doing things we
>> >might never branch.
>> >
>> >
>> >> After branching, we can focus on the following:
>> >> 1. every checkin to 0.92 branch shouldn't break its build. This requires
>> the
>> >> committer to perform at least one complete test suite run before
>> checking
>> >> in.
>> >
>> >I'll set up a build of the branch soon as we branch. I've been
>> >responsible for build breakage of late. Will reform myself.
>> >
>> >> 2. patches for issues on http://s.apache.org/x4 are allowed to go into
>> 0.92
>> >> branch. New issues outside the list have to collect 3 +1 from committers
>> >> before going in.
>> >>
>> >
>> >I'd say 3+1s if its a 'big' change. The small stuff should just let go
>> through.
>> >
>> >Good stuff,
>> >St.Ack
>> >
>> >
>> >
>>
>
>
>
Re: HBase 0.92 branch
Posted by Ted Yu <yu...@gmail.com>.
>> Our practice is RTC with one +1 other than patch creator being sufficient
for commit, and CTR for trivial changes. I'm -1 on changing this.
Fair.
Shall we branch when we have a successful TRUNK build after the three
blockers on http://s.apache.org/x4, HFile V2 and HBASE-4027 get checked in ?
Thanks
On Thu, Jul 28, 2011 at 11:10 AM, Andrew Purtell <ap...@apache.org>wrote:
> > New issues outside the list have to collect 3 +1 from committers before
> going in.
>
> -1
>
> Do we actually have bylaws?
>
> Our practice is RTC with one +1 other than patch creator being sufficient
> for commit, and CTR for trivial changes. I'm -1 on changing this.
>
> Hadoop core recently ran a vote to increase the threshold for RTC to three
> +1s for commits that represent a branch merge. I support this for HBase. So
> if people think some big "new issue" needs additional review, then we make a
> feature branch for it and require three +1s for merge commit.
>
> Best regards,
>
>
> - Andy
>
> Problems worthy of attack prove their worth by hitting back. - Piet Hein
> (via Tom White)
>
>
> >________________________________
> >From: Stack <st...@duboce.net>
> >To: dev@hbase.apache.org
> >Sent: Thursday, July 28, 2011 10:04 AM
> >Subject: Re: HBase 0.92 branch
> >
> >On Thu, Jul 28, 2011 at 9:47 AM, Ted Yu <yu...@gmail.com> wrote:
> >> Since build 2047, TRUNK build has been on a steady decline. I believe
> even
> >> if this trend is reversed, it is hard to guarantee that TRUNK build
> would be
> >> healthy in the future.
> >
> >I think we can fix it.
> >
> >> Now that the issues on http://s.apache.org/x4 have come down to 18, I
> think
> >> we should consider branching 0.92 soon (maybe after HFile v2 and
> HBASE-4027
> >> go in and we fix the broken build).
> >
> >I suppose I'd like the issues to go to zero before we branched but I
> >can go along w/ the above; if we held to my way of doing things we
> >might never branch.
> >
> >
> >> After branching, we can focus on the following:
> >> 1. every checkin to 0.92 branch shouldn't break its build. This requires
> the
> >> committer to perform at least one complete test suite run before
> checking
> >> in.
> >
> >I'll set up a build of the branch soon as we branch. I've been
> >responsible for build breakage of late. Will reform myself.
> >
> >> 2. patches for issues on http://s.apache.org/x4 are allowed to go into
> 0.92
> >> branch. New issues outside the list have to collect 3 +1 from committers
> >> before going in.
> >>
> >
> >I'd say 3+1s if its a 'big' change. The small stuff should just let go
> through.
> >
> >Good stuff,
> >St.Ack
> >
> >
> >
>
Re: HBase 0.92 branch
Posted by Andrew Purtell <ap...@apache.org>.
> New issues outside the list have to collect 3 +1 from committers before going in.
-1
Do we actually have bylaws?
Our practice is RTC with one +1 other than patch creator being sufficient for commit, and CTR for trivial changes. I'm -1 on changing this.
Hadoop core recently ran a vote to increase the threshold for RTC to three +1s for commits that represent a branch merge. I support this for HBase. So if people think some big "new issue" needs additional review, then we make a feature branch for it and require three +1s for merge commit.
Best regards,
- Andy
Problems worthy of attack prove their worth by hitting back. - Piet Hein (via Tom White)
>________________________________
>From: Stack <st...@duboce.net>
>To: dev@hbase.apache.org
>Sent: Thursday, July 28, 2011 10:04 AM
>Subject: Re: HBase 0.92 branch
>
>On Thu, Jul 28, 2011 at 9:47 AM, Ted Yu <yu...@gmail.com> wrote:
>> Since build 2047, TRUNK build has been on a steady decline. I believe even
>> if this trend is reversed, it is hard to guarantee that TRUNK build would be
>> healthy in the future.
>
>I think we can fix it.
>
>> Now that the issues on http://s.apache.org/x4 have come down to 18, I think
>> we should consider branching 0.92 soon (maybe after HFile v2 and HBASE-4027
>> go in and we fix the broken build).
>
>I suppose I'd like the issues to go to zero before we branched but I
>can go along w/ the above; if we held to my way of doing things we
>might never branch.
>
>
>> After branching, we can focus on the following:
>> 1. every checkin to 0.92 branch shouldn't break its build. This requires the
>> committer to perform at least one complete test suite run before checking
>> in.
>
>I'll set up a build of the branch soon as we branch. I've been
>responsible for build breakage of late. Will reform myself.
>
>> 2. patches for issues on http://s.apache.org/x4 are allowed to go into 0.92
>> branch. New issues outside the list have to collect 3 +1 from committers
>> before going in.
>>
>
>I'd say 3+1s if its a 'big' change. The small stuff should just let go through.
>
>Good stuff,
>St.Ack
>
>
>
Re: HBase 0.92 branch
Posted by Ted Yu <yu...@gmail.com>.
>> if we held to my way of doing things we might never branch.
That was the motivation for my poll. Without fixing all issues on
http://s.apache.org/x4, there would still be a 0.92 build for people to try
out.
>> I'd say 3+1s if its a 'big' change. The small stuff should just let go
through.
I agree.
On Thu, Jul 28, 2011 at 10:04 AM, Stack <st...@duboce.net> wrote:
> On Thu, Jul 28, 2011 at 9:47 AM, Ted Yu <yu...@gmail.com> wrote:
> > Since build 2047, TRUNK build has been on a steady decline. I believe
> even
> > if this trend is reversed, it is hard to guarantee that TRUNK build would
> be
> > healthy in the future.
>
> I think we can fix it.
>
> > Now that the issues on http://s.apache.org/x4 have come down to 18, I
> think
> > we should consider branching 0.92 soon (maybe after HFile v2 and
> HBASE-4027
> > go in and we fix the broken build).
>
> I suppose I'd like the issues to go to zero before we branched but I
> can go along w/ the above; if we held to my way of doing things we
> might never branch.
>
>
> > After branching, we can focus on the following:
> > 1. every checkin to 0.92 branch shouldn't break its build. This requires
> the
> > committer to perform at least one complete test suite run before checking
> > in.
>
> I'll set up a build of the branch soon as we branch. I've been
> responsible for build breakage of late. Will reform myself.
>
> > 2. patches for issues on http://s.apache.org/x4 are allowed to go into
> 0.92
> > branch. New issues outside the list have to collect 3 +1 from committers
> > before going in.
> >
>
> I'd say 3+1s if its a 'big' change. The small stuff should just let go
> through.
>
> Good stuff,
> St.Ack
>
Re: HBase 0.92 branch
Posted by Stack <st...@duboce.net>.
On Thu, Jul 28, 2011 at 9:47 AM, Ted Yu <yu...@gmail.com> wrote:
> Since build 2047, TRUNK build has been on a steady decline. I believe even
> if this trend is reversed, it is hard to guarantee that TRUNK build would be
> healthy in the future.
I think we can fix it.
> Now that the issues on http://s.apache.org/x4 have come down to 18, I think
> we should consider branching 0.92 soon (maybe after HFile v2 and HBASE-4027
> go in and we fix the broken build).
I suppose I'd like the issues to go to zero before we branched but I
can go along w/ the above; if we held to my way of doing things we
might never branch.
> After branching, we can focus on the following:
> 1. every checkin to 0.92 branch shouldn't break its build. This requires the
> committer to perform at least one complete test suite run before checking
> in.
I'll set up a build of the branch soon as we branch. I've been
responsible for build breakage of late. Will reform myself.
> 2. patches for issues on http://s.apache.org/x4 are allowed to go into 0.92
> branch. New issues outside the list have to collect 3 +1 from committers
> before going in.
>
I'd say 3+1s if its a 'big' change. The small stuff should just let go through.
Good stuff,
St.Ack