Grokbase Groups HBase dev June 2010
FAQ
Hey everyone,

As mentioned before, I'm hoping to have a development release we can talk up
at the Hadoop Summit at the end of this month.

To that end, I'm planning on branching trunk on Monday to try to go through
some short stabilization testing for a dev release at the end of the week.
I'm happy to do this as an internal branch at Cloudera, and just do a
release for CDH, or I can do a branch on the Apache SVN and an official
Apache release. Does anyone prefer one way or the other? Hopefully I will
branch from the state of trunk on Monday, but if that proves to be too
unstable I may rewind a week or two before the recent master changes, and
then cherry-pick some other fixes on top.

Note: this release is intended as a "technology preview" - we can get some
users trying out trunk on real clusters early so that we'll have an easier
time releasing when we think we're actually done. I plan on including a list
of known bugs and deficiencies - there's no chance we'll close out all the
blockers or finish the master rewrite, but I think the release will still be
valuable.

Any thoughts on this?

Thanks
-Todd

--

Todd Lipcon
Software Engineer, Cloudera

Search Discussions

  • Jonathan Gray at Jun 19, 2010 at 1:11 am
    So then is the plan to branch and then cherry-pick fixes until the first release of it?

    Following the release, would we continue to cherry-pick bug fixes and possibly have a second dev release?

    I'm thinking cutting a branch soon and cherry-picking fixes between now and the summit might make sense. But then we should not separately support that release (as in, if bugs are found that are no longer relevant or are implemented differently at the tip of trunk, we don't fix them in the branch, only in trunk). I think there's a decent chance of this because of some remaining largish changes going in.

    A follow-up dev release would still be possible but should be a re-cut off a fresher trunk with more goodness in it rather than a patched version of the first dev release.

    JG
    -----Original Message-----
    From: Todd Lipcon
    Sent: Friday, June 18, 2010 4:58 PM
    To: dev
    Subject: Development release next week

    Hey everyone,

    As mentioned before, I'm hoping to have a development release we can
    talk up
    at the Hadoop Summit at the end of this month.

    To that end, I'm planning on branching trunk on Monday to try to go
    through
    some short stabilization testing for a dev release at the end of the
    week.
    I'm happy to do this as an internal branch at Cloudera, and just do a
    release for CDH, or I can do a branch on the Apache SVN and an official
    Apache release. Does anyone prefer one way or the other? Hopefully I
    will
    branch from the state of trunk on Monday, but if that proves to be too
    unstable I may rewind a week or two before the recent master changes,
    and
    then cherry-pick some other fixes on top.

    Note: this release is intended as a "technology preview" - we can get
    some
    users trying out trunk on real clusters early so that we'll have an
    easier
    time releasing when we think we're actually done. I plan on including a
    list
    of known bugs and deficiencies - there's no chance we'll close out all
    the
    blockers or finish the master rewrite, but I think the release will
    still be
    valuable.

    Any thoughts on this?

    Thanks
    -Todd

    --

    Todd Lipcon
    Software Engineer, Cloudera
  • Jonathan Gray at Jun 19, 2010 at 1:22 am
    And in general, I'd prefer the route of official Apache development releases. I think there's value to all users and not just those that will be running CDH. Should be a version up on JIRA so we can appropriately flag it on issues and such.

    We might consider trying to get a dev release of the Hadoop 0.20-append branch cut along with it? At the least I suppose we'd ship with jars and maven config to pull from a 0.20-append hadoop?
    -----Original Message-----
    From: Jonathan Gray
    Sent: Friday, June 18, 2010 6:11 PM
    To: dev@hbase.apache.org
    Subject: RE: Development release next week

    So then is the plan to branch and then cherry-pick fixes until the
    first release of it?

    Following the release, would we continue to cherry-pick bug fixes and
    possibly have a second dev release?

    I'm thinking cutting a branch soon and cherry-picking fixes between now
    and the summit might make sense. But then we should not separately
    support that release (as in, if bugs are found that are no longer
    relevant or are implemented differently at the tip of trunk, we don't
    fix them in the branch, only in trunk). I think there's a decent
    chance of this because of some remaining largish changes going in.

    A follow-up dev release would still be possible but should be a re-cut
    off a fresher trunk with more goodness in it rather than a patched
    version of the first dev release.

    JG
    -----Original Message-----
    From: Todd Lipcon
    Sent: Friday, June 18, 2010 4:58 PM
    To: dev
    Subject: Development release next week

    Hey everyone,

    As mentioned before, I'm hoping to have a development release we can
    talk up
    at the Hadoop Summit at the end of this month.

    To that end, I'm planning on branching trunk on Monday to try to go
    through
    some short stabilization testing for a dev release at the end of the
    week.
    I'm happy to do this as an internal branch at Cloudera, and just do a
    release for CDH, or I can do a branch on the Apache SVN and an official
    Apache release. Does anyone prefer one way or the other? Hopefully I
    will
    branch from the state of trunk on Monday, but if that proves to be too
    unstable I may rewind a week or two before the recent master changes,
    and
    then cherry-pick some other fixes on top.

    Note: this release is intended as a "technology preview" - we can get
    some
    users trying out trunk on real clusters early so that we'll have an
    easier
    time releasing when we think we're actually done. I plan on including a
    list
    of known bugs and deficiencies - there's no chance we'll close out all
    the
    blockers or finish the master rewrite, but I think the release will
    still be
    valuable.

    Any thoughts on this?

    Thanks
    -Todd

    --

    Todd Lipcon
    Software Engineer, Cloudera
  • Todd Lipcon at Jun 19, 2010 at 6:17 am

    On Fri, Jun 18, 2010 at 6:10 PM, Jonathan Gray wrote:

    So then is the plan to branch and then cherry-pick fixes until the first
    release of it?

    Following the release, would we continue to cherry-pick bug fixes and
    possibly have a second dev release?
    I don't think so - a second dev release would be great, but we'd do a new
    branch off trunk - ie these releases aren't intended to be supported... just
    snapshots of trunk with a little spit and polish prior to release to make
    sure it basically works. If there are bugs we'll tell people "thanks for
    reporting, hopefully they'll be gone in the next dev release" and leave it
    at that, in my opinion. Otherwise we'll be too much distracted from our
    primary mission of getting a real stable trunk release out this summer.

    I'm thinking cutting a branch soon and cherry-picking fixes between now and
    the summit might make sense. But then we should not separately support that
    release (as in, if bugs are found that are no longer relevant or are
    implemented differently at the tip of trunk, we don't fix them in the
    branch, only in trunk). I think there's a decent chance of this because of
    some remaining largish changes going in.

    A follow-up dev release would still be possible but should be a re-cut off
    a fresher trunk with more goodness in it rather than a patched version of
    the first dev release.
    Yep. I wrote the above paragraph and then read yours, and appears we're in
    100% agreement :)

    JG
    -----Original Message-----
    From: Todd Lipcon
    Sent: Friday, June 18, 2010 4:58 PM
    To: dev
    Subject: Development release next week

    Hey everyone,

    As mentioned before, I'm hoping to have a development release we can
    talk up
    at the Hadoop Summit at the end of this month.

    To that end, I'm planning on branching trunk on Monday to try to go
    through
    some short stabilization testing for a dev release at the end of the
    week.
    I'm happy to do this as an internal branch at Cloudera, and just do a
    release for CDH, or I can do a branch on the Apache SVN and an official
    Apache release. Does anyone prefer one way or the other? Hopefully I
    will
    branch from the state of trunk on Monday, but if that proves to be too
    unstable I may rewind a week or two before the recent master changes,
    and
    then cherry-pick some other fixes on top.

    Note: this release is intended as a "technology preview" - we can get
    some
    users trying out trunk on real clusters early so that we'll have an
    easier
    time releasing when we think we're actually done. I plan on including a
    list
    of known bugs and deficiencies - there's no chance we'll close out all
    the
    blockers or finish the master rewrite, but I think the release will
    still be
    valuable.

    Any thoughts on this?

    Thanks
    -Todd

    --

    Todd Lipcon
    Software Engineer, Cloudera


    --
    Todd Lipcon
    Software Engineer, Cloudera
  • Stack at Jun 19, 2010 at 5:49 am

    On Fri, Jun 18, 2010 at 4:57 PM, Todd Lipcon wrote:
    To that end, I'm planning on branching trunk on Monday to try to go through
    some short stabilization testing for a dev release at the end of the week.
    Sounds great. Will be better being able to point at a sampler for the
    new release at hadoop/hbase summit, something that will basically run,
    something we cannot guarantee for trunk considering the big changes
    coming in.

    I'm happy to do this as an internal branch at Cloudera, and just do a
    release for CDH, or I can do a branch on the Apache SVN and an official
    Apache release.
    Should be an Apache release. I'd like to help out on stabilization.

    Good on you Todd,
    St.Ack
  • Dhruba Borthakur at Jun 19, 2010 at 6:09 am
    ith jars and maven config to pull from a 0.20-append hadoop
    Yes, I agree that will be good.

    There are a few remaining patches that will go into the 0,20-append branch,
    but I am waiting for them to et committed into Hadoop trunk before pulling
    them into 0-20-append.

    thanks,
    dhruba
    On Fri, Jun 18, 2010 at 10:49 PM, Stack wrote:
    On Fri, Jun 18, 2010 at 4:57 PM, Todd Lipcon wrote:

    To that end, I'm planning on branching trunk on Monday to try to go through
    some short stabilization testing for a dev release at the end of the
    week.

    Sounds great. Will be better being able to point at a sampler for the
    new release at hadoop/hbase summit, something that will basically run,
    something we cannot guarantee for trunk considering the big changes
    coming in.

    I'm happy to do this as an internal branch at Cloudera, and just do a
    release for CDH, or I can do a branch on the Apache SVN and an official
    Apache release.
    Should be an Apache release. I'd like to help out on stabilization.

    Good on you Todd,
    St.Ack


    --
    Connect to me at http://www.facebook.com/dhruba
  • Stack at Jun 19, 2010 at 7:05 pm

    On Fri, Jun 18, 2010 at 11:08 PM, Dhruba Borthakur wrote:
    There are a few remaining patches that will go into the 0,20-append branch,
    but I am waiting for them to et committed into Hadoop trunk before pulling
    them into 0-20-append.
    Which issues Dhruba? Are they tagged 0.20-append? Can we help get
    with getting them committed on TRUNK?
    Thanks,
    St.Ack
  • Dhruba Borthakur at Jun 20, 2010 at 8:45 am
    Sure, tey are all marked as "Fix for 0.20-append"..

    http://issues.apache.org/jira/secure/IssueNavigator.jspa?reset=true&mode=hide&sorter/order=DESC&sorter/field=priority&resolution=-1&pid=12310942&fixfor=12315103

    thanks,
    dhruba
    On Sat, Jun 19, 2010 at 12:04 PM, Stack wrote:
    On Fri, Jun 18, 2010 at 11:08 PM, Dhruba Borthakur wrote:
    There are a few remaining patches that will go into the 0,20-append branch,
    but I am waiting for them to et committed into Hadoop trunk before pulling
    them into 0-20-append.
    Which issues Dhruba? Are they tagged 0.20-append? Can we help get
    with getting them committed on TRUNK?
    Thanks,
    St.Ack


    --
    Connect to me at http://www.facebook.com/dhruba
  • Todd Lipcon at Jun 21, 2010 at 6:17 am
    I tried to do some cluster testing of trunk this afternoon/tonight and ran
    into several issues that blocked my ability to get up and running.
    Particularly, HBASE-2758, plus the inability to disable a table (forget the
    JIRA number, but it's up there)

    So, unless I hear any big -1s, I think it's best to revert the recent master
    changes for this dev release branch this week. There's still plenty of new
    stuff for users to try out, and I think the new master stuff is a bit too
    unstable yet for users to try. Let's do another dev release in a couple
    weeks once the new master stuff is more fully baked, and use this one to
    suss out bugs in all the other new stuff.

    Again, a reminder that we don't plan to support this branch post-release -
    it's just a preview for users, so I don't think doing a release without the
    new master stuff is going to take steam out of continued fixup of these bugs
    in trunk proper.

    Sound good by everyone?

    -Todd
    On Fri, Jun 18, 2010 at 4:57 PM, Todd Lipcon wrote:

    Hey everyone,

    As mentioned before, I'm hoping to have a development release we can talk
    up at the Hadoop Summit at the end of this month.

    To that end, I'm planning on branching trunk on Monday to try to go through
    some short stabilization testing for a dev release at the end of the week.
    I'm happy to do this as an internal branch at Cloudera, and just do a
    release for CDH, or I can do a branch on the Apache SVN and an official
    Apache release. Does anyone prefer one way or the other? Hopefully I will
    branch from the state of trunk on Monday, but if that proves to be too
    unstable I may rewind a week or two before the recent master changes, and
    then cherry-pick some other fixes on top.

    Note: this release is intended as a "technology preview" - we can get some
    users trying out trunk on real clusters early so that we'll have an easier
    time releasing when we think we're actually done. I plan on including a list
    of known bugs and deficiencies - there's no chance we'll close out all the
    blockers or finish the master rewrite, but I think the release will still be
    valuable.

    Any thoughts on this?

    Thanks
    -Todd

    --

    Todd Lipcon
    Software Engineer, Cloudera


    --
    Todd Lipcon
    Software Engineer, Cloudera
  • Stack at Jun 21, 2010 at 7:33 pm
    +1 Sounds good to me. Excising the commit might be a little tough
    given other stuff went in on top of it to fix up issues found.
    St.Ack
    On Sun, Jun 20, 2010 at 11:16 PM, Todd Lipcon wrote:
    I tried to do some cluster testing of trunk this afternoon/tonight and ran
    into several issues that blocked my ability to get up and running.
    Particularly, HBASE-2758, plus the inability to disable a table (forget the
    JIRA number, but it's up there)

    So, unless I hear any big -1s, I think it's best to revert the recent master
    changes for this dev release branch this week. There's still plenty of new
    stuff for users to try out, and I think the new master stuff is a bit too
    unstable yet for users to try. Let's do another dev release in a couple
    weeks once the new master stuff is more fully baked, and use this one to
    suss out bugs in all the other new stuff.

    Again, a reminder that we don't plan to support this branch post-release -
    it's just a preview for  users, so I don't think doing a release without the
    new master stuff is going to take steam out of continued fixup of these bugs
    in trunk proper.

    Sound good by everyone?

    -Todd
    On Fri, Jun 18, 2010 at 4:57 PM, Todd Lipcon wrote:

    Hey everyone,

    As mentioned before, I'm hoping to have a development release we can talk
    up at the Hadoop Summit at the end of this month.

    To that end, I'm planning on branching trunk on Monday to try to go through
    some short stabilization testing for a dev release at the end of the week.
    I'm happy to do this as an internal branch at Cloudera, and just do a
    release for CDH, or I can do a branch on the Apache SVN and an official
    Apache release. Does anyone prefer one way or the other? Hopefully I will
    branch from the state of trunk on Monday, but if that proves to be too
    unstable I may rewind a week or two before the recent master changes, and
    then cherry-pick some other fixes on top.

    Note: this release is intended as a "technology preview" - we can get some
    users trying out trunk on real clusters early so that we'll have an easier
    time releasing when we think we're actually done. I plan on including a list
    of known bugs and deficiencies - there's no chance we'll close out all the
    blockers or finish the master rewrite, but I think the release will still be
    valuable.

    Any thoughts on this?

    Thanks
    -Todd

    --

    Todd Lipcon
    Software Engineer, Cloudera


    --
    Todd Lipcon
    Software Engineer, Cloudera

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupdev @
categorieshbase, hadoop
postedJun 18, '10 at 11:58p
activeJun 21, '10 at 7:33p
posts10
users4
websitehbase.apache.org

People

Translate

site design / logo © 2022 Grokbase