Hi Marton, Yanquan, Thank you for your responses! Regarding the points brought up to discuss:
1. Supporting 1.20 definitely makes sense, but since there is quite a big gap to work down here now, I am not sure it should be done in 1 step. As my understanding, the externalized connector dev model [1] do not explicitly forbid that, but AFAIK there were external connector release that supported 3 different Flink minor versions. In this case, I think technically would be possible, but IMO supporting 3 Flink verisons adds more complexity to maintain. So what I would suggest to release 4.0 with Flink 1.18 and 1.19 support, and after that there can be a 4.1 that supports 1.19 and 1.20. 4.0 will only have patch support, probably minimizing Flink version specific problems. 2. Flink 1.17 had no JDK17 support, so those Hadoop related problems should not play a role if something needs to be released that supports 1.17. But if connector 4.0 is released, 3.x versions will not get any new releases (even not patch), cause 1.17 is out of support already. Best, Ferenc [1] https://cwiki.apache.org/confluence/display/FLINK/Externalized+Connector+development On Wednesday, 23 October 2024 at 05:19, Yanquan Lv <decq12y...@gmail.com> wrote: > > > Hi Feri, > Thank you for bringing up this discussion. > I agree to release a version to bump the newer version of Flink with partial > JDK versions support. I have two points to be discussed. > 1. I have heard many inquiries about supporting higher versions of Flink in > Slack, Chinese communities, etc., and a large part of them hope to use it on > Flink1.20. Should we consider explicitly supporting Flink1.20 on version 4.0, > otherwise users will have to wait for a relatively long release cycle. > 2. Currently supporting Flink1.17 is difficult, but are there any plans to > support it in the future? Do we need to wait for Hadoop related repositories > to release specific versions. > > > > 2024年10月22日 19:44，Ferenc Csaky ferenc.cs...@pm.me.INVALID 写道： > > > > Hello devs, > > > > I would like to start a discussion regarding a new HBase connector release. > > Currently, the > > externalized HBase connector has only 1 release: 3.0.0 that supports Flink > > 1.16 and 1.17. > > > > By stating this, it is obvious that the connector is already outdated for > > quite a while. There > > is a long-lasting ticket [1] to release a newer HBase version, which also > > contains a major version > > bump as HBase 1.x support is removed, but covering JDK17 with the current > > Hadoop related > > dependency mix is impossible, because there are parts that do not play well > > with it when you > > try to compile with JDK17+, and there are no runtime tests as well. > > > > Solving that properly will require to bump the HBase, Hadoop, and Zookeeper > > versions as well, > > but that will require more digging and some refactoring, at least on the > > test side. > > > > To cut some corners and move forward I think at this point it would make > > sense to release > > version 4.0 that supports Flink 1.18 and 1.19 but only on top of JDK8 and > > JDK11 just to close the > > current gap a bit. I am thinking about including the limitations in the > > java compat docs [2] to > > highlight users. > > > > WDYT? > > > > Best, > > Ferenc > > > > [1] https://issues.apache.org/jira/browse/FLINK-35136 > > [2] > > https://nightlies.apache.org/flink/flink-docs-release-1.19/docs/deployment/java_compatibility/
