hadoop-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Eric Baldeschwieler <eri...@hortonworks.com>
Subject Re: [VOTE] Release hadoop-0.22.0-rc0
Date Sat, 03 Dec 2011 02:48:47 GMT
I'm fine with the idea of alpha releases. I think they are important.  I'm a bit confused on
where this is going. 

what are your criteria for declaring the 0.22 branch stable?  Is compatibility with 1.0 and
security a goal for this release line?  

E14 - typing on glass

On Dec 2, 2011, at 1:15 AM, Konstantin Shvachko <shv.hadoop@gmail.com> wrote:

> This is a well tested state of the branch, which imo worth releasing.
> Adding changes at this point will invalidate the testing.
> 0.23 has "substantial regressions from the stable line" as well, and
> is not stable.
> 1.0 does not have important features including some in MR that 0.22 has.
> What is a regression!
> 0.22 is progress in at least one area, which is HDFS.
> People can actually start appending data to files, use links, and
> build features (e.g. HA) based on BackupNode.
> HBase clusters should not be segregated anymore as HBase-only ones,
> but will allow general load as well.
> Speaking of confusing. What isn't in Hadoop this days. Say, when you
> transition from Hadoop 1.0 to 0.22 or 0.23 you call it upgrade.
> Seriously. This project needs to move forward beyond the 0.20 line. I
> tested 0.22 in the last three months and found it working with HBase,
> Hive, Pig, Cascading, Zookeeper, Oozie. What else do you need to
> release? It is not perfect? Agreed. Have anybody seen a perfect piece
> of software? I haven't.
> Eric, thank you for raising the question. It is important to
> understand the pros and cons of different versions.
> You should try the candidate, who knows it might just work for you.
> --Konstantin
> On Thu, Dec 1, 2011 at 10:26 PM, Eric Baldeschwieler
> <eric14@hortonworks.com> wrote:
>> So this release contains substantial regressions from the stable line and contains
a subset of features in 0.23.
>> Wouldn't it be prudent to wait until the branch is compelling when compared to one
of the other release lines before releasing?  Otherwise this just seems confusing.
>> ---
>> E14 - typing on glass
>> On Dec 1, 2011, at 6:10 PM, Konstantin Boudnik <cos@apache.org> wrote:
>>> +1
>>> Checksum are correct, license is in place
>>> Single node cluster (Ubuntu 11.04) installation and a mapreduce job worked
>>> without any issues.
>>> Cos
>>> On Tue, Nov 29, 2011 at 02:47AM, Konstantin Shvachko wrote:
>>>> I created a release candidate for hadoop-0.22.0 available for review at:
>>>> http://people.apache.org/~shv/hadoop-0.22.0-rc0/
>>>> The candidate incorporates more than 700 fixed jiras counting from the
>>>> release of 0.21.0.
>>>> The branch was cut on November 17 last year. During this year the
>>>> branch was tested by different groups in the Hadoop community.
>>>> Recently it has been extensively tested on a mid-size cluster.
>>>> It contains important improvements, optimizations and bug fixes for
>>>> HDFS and MapReduce compared to 0.21.0. Some features introduced in the
>>>> latest offspring releases of hadoop-security are not yet ported to
>>>> this branch, which is mostly related to security and MapReduce
>>>> improvements. Compared to 0.23.0 this release will not contain
>>>> Federation or Yarn.
>>>> The hadoop-0.22.0 release build is fully automated thanks to the
>>>> BigTop team. rc0 artifacts are available and reproducible on Hadoop
>>>> Jenkins. The validity of rc0 has been also verified with BigTop.
>>>> I think this could be a usable release and an important step forward
>>>> in the Hadoop roadmap.
>>>> Many people contributed to hadoop-0.22. Thank you everybody who
>>>> developed code, built, tested, found and fixed bugs or supported this
>>>> effort by any other means.
>>>> Please try it out and vote.
>>>> Thanks,
>>>> --Konstantin

View raw message