Hi Yang, The main difference between 2.6 and 3.0 is the new real-time OLAP feature. Hadoop 2 users can select either of them, depends on whether they need the real-time feature.
After 3.0, the next major features would be the Flink cube engine (planned in v3.1) and the Parquet storage (early stage, maybe in v4.0). When the parquet storage is released, as the dependency on HBase can be dropped, then we assume the API issue will easier than today. We can re-evaluate the possibility to support Hadoop 2. So I think the impact on today's Hadoop 2 users is acceptable. Not mention that they still can manually compile that. Best regards, Shaofeng Shi 史少锋 Apache Kylin PMC Email: [email protected] Apache Kylin FAQ: https://kylin.apache.org/docs/gettingstarted/faq.html Join Kylin user mail group: [email protected] Join Kylin dev mail group: [email protected] Li Yang <[email protected]> 于2020年2月27日周四 上午7:37写道： > The proposal means Kylin 3.0 will be the last major version that supports > Hadoop 2. > > What will be recommended version for Hadoop 2 users after this? I feel the > latest stable version of 2.6 is better than 3.0. > > Anyway, I'm fine with moving focus to Hadoop 3. That is the direction. > However we shall also think about what it means for Hadoop 2 users. > Questions like below shall also be answered. > > - What is the recommended version/branch for Hadoop 2? (Btw, 3.0 does not > sound right here.) > - How that version/branch will be maintained? > > +1 in general > > Regards > -Yang > > > On Wed, Feb 26, 2020 at 5:36 PM Zhou Kang <[email protected]> wrote: > > > +1 > > > > > > > 2020年2月26日 下午3:48，ShaoFeng Shi <[email protected]> 写道： > > > > > > Hello, Kylin users and developers, > > > > > > As we know Hadoop 3 and HBase 2 have released for some time. Kylin > > starts to support Hadoop 3 since v2.5.0 in Sep 2018. As the APIs of > HBase > > 1 and 2 are incompatible, we need to keep different branches for them. > And > > in each release, we need to build separate packages and do a round of > > testing for them separately. Furthermore, Cloudera's API difference with > > the Apache release makes the situation worse; We need to build 4 binary > > packages for reach release. That has spent much of our manual effort and > > computing resources. > > > > > > Today, Hadoop 3 + HBase 2 becomes enough mature and stable for > > production use; And we see more and more users are starting to use the > new > > versions. We think it is time for Kylin to totally upgrade to the new > > version. So that we can focus more on Kylin itself, instead of > environments. > > > > > > Here is my proposal: > > > 1) From Kylin 3.1, Hadoop/HBase version upgrades to 3.1/2.1 (or a > close > > version); > > > 2) Hadoop 2 and HBase 1 users can use Kylin 3.0 and previous releases; > > > 3) We will re-evaluate the need for building binary packages for > > Cloudera release. (we may raise another discuss) > > > > > > Please let us know your comments. And please also understand with the > > limited resource we couldn't support multiple Hadoop versions... > > > > > > Thanks! > > > > > > Best regards, > > > > > > Shaofeng Shi 史少锋 > > > Apache Kylin PMC > > > Email: [email protected] > > > > > > Apache Kylin FAQ: > https://kylin.apache.org/docs/gettingstarted/faq.html > > > Join Kylin user mail group: [email protected] > > > Join Kylin dev mail group: [email protected] > > > > > > > > > > >
