[ANNOUNCE] Weekly Community Update 2020/26-27

[Konstantin Knauf <kn...@apache.org>]

Dear community,

happy to share this (and last) week's community update. Flink 1.11 is
finally ready to be released. Besides that a few design discussions in
different areas of Apache Flink, like enhanced fan out for Flink's Kinesis
Source or Temporal Table support in pure Flink SQL, and of course a bit
more.

Flink Development
==============

* [releases] Zhinjiang has published release candidate #4 for Flink 1.11.0
last Tuesday. The vote [1] passed this morning, so we will see the release
of Flink 1.11 very soon.

* [sql] A while ago I started a discussion on supporting Temporal Table
Joins via pure Flink SQL. As of now, a user either needs to register a
Temporal Table Function in the Table API or the environments configuration
of the SQL CLI. This became a more involved discussion than anticipated
that Leonard Xu is doing a great job in moving forward. It seems that we
are close to a FLIP document now. [2]

* [connectors] Danny Cranmer has started the discussion [3] and -
subsequently - the vote [4] on FLIP-128, which adds support for enhanced
fan out for Flink's Kinesis source. With enhanced fan out each consumer
receives dedicated data output per shard, as opposed to competing for the
per-shared data output with other consumers.

* [apis] Aljoscha has started a discussion about what kind of compatibility
guarantees the community would like to give for the APIs that are commonly
used by packaged, third-party or custom connectors. Not too much feedback
so far, but right now it seems that we would like it to be safe to use
connectors across patch releases (1.x.y -> 1.x.z), but not across minor
releases (1.u -> 1.v). Based on the recent discussions [5] on the
guarantees for @PublicEvolving this means that connectors could only use
APIs that are annotated @Public or @PublicEvolving. [6]

* [state] Etienne Chauchot has published a design document for FLINK-17073,
which introduces a backpressure mechanism for checkpoints: when checkpoints
can not be cleaned up as quickly as they are created triggering new
checkpoints will be delayed. This change was motivated by an OOME on the
Jobmanger resulting from too many queued checkpoint clean up tasks. [7]

[1]
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/VOTE-Release-1-11-0-release-candidate-4-tp42829.html
[2]
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-FLINK-16824-Creating-Temporal-Table-Function-via-DDL-tp40333.html
[3]
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-FLIP-128-Enhanced-Fan-Out-for-AWS-Kinesis-Consumers-tp42728.html
[4]
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/VOTE-FLIP-128-Enhanced-Fan-Out-for-AWS-Kinesis-Consumers-tp42846.html
[5]
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-Cross-version-compatibility-guarantees-of-Flink-Modules-Jars-tp42746.html
[6]
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/DISCUSS-Stability-guarantees-for-PublicEvolving-classes-tp41459.html
[7]
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/FLINK-17073-checkpoint-backpressure-design-doc-tp42788.html

Notable Bugs
==========

* [FLINK-18452] [1.11] [1.10.1] SQL queries that use "Top-N" can not be
restored from a savepoint due to a incorrectly implemented Object#equals in
one of the state objects. [8]

[8] https://issues.apache.org/jira/browse/FLINK-18452

Events, Blog Posts, Misc
===================

* On the Ververica blog Jaehyeuk Oh & Gihoon Yeom explain how HyperConnect
is using Apache Flink for match making in their real-time communication app
Azar. [9]

* On the Flink blog, Jeff Zhang has published the second part of his blog
post series on Flink on Zeppelin. [10]

[9]
https://www.ververica.com/blog/data-driven-matchmaking-at-azar-with-apache-flink
[10]
https://flink.apache.org/ecosystem/2020/06/23/flink-on-zeppelin-part2.html

Cheers,

Konstantin

-- 

Konstantin Knauf

https://twitter.com/snntrable

https://github.com/knaufk