Hi liuwb9, thanks for reporting this issue. Please keep a few things in mind, though: - The community only supports the two most-recently released versions of Flink. Your version 1.14 is too old (issues might have been fixed already in newer releases). Trying to reproduce the behavior in a supported Flink version is appreciated. - The dev mailing list is usually used for development/architectural discussions (user questions could be asked in the user mailing list) [1] - The description of the issue sounds a bit vague. Is your conclusion of a bug based on the fact that the submission took longer than expected? That information is not enough to start an investigation. Usually, providing Flink logs helps to support your case and makes it easier for others to analyze what's going on. - Jira [2] might be the best channel if you are certain that the behavior you observed indicates a bug. Additional artifacts (like logs) can be attached to your bug report there as well.
Thanks, Matthias [1] https://flink.apache.org/community/#mailing-lists [2] https://flink.apache.org/community/#issue-tracker On Thu, Jul 20, 2023 at 11:27 AM liu...@midea.com <liu...@midea.com> wrote: > > Hi all, > > I encountered a suspected bug in Flink. > A flink job associates with 30+ HBase2.2 asynchronous dimension > tables（Temporal > Join AND lookup.async = 'true'), > When I submitted this job, it took a very long time to submit to the YARN > cluster. > During the generation of the job graph and optimization of the operator > chain, recursion occurs in the process of locating and discovering. > > Flink source code location: version 1.14 > > > org.apache.flink.streaming.api.graph.StreamingJobGraphGenerator#areOperatorsChainable > > > liu...@midea.com >
