I personally think this change should be small enough to just be done via jira without introducing a CEP. But since paging generally has seen some problems, it'd be good to understand the design before flushing it out in code.
One of the ways where we probably should be considerate is the paging state. Paging state is passed from the coordinator to client, and then from the client to a (potentially different) coordinator. If the coordinator speaks the internode protocol of a different version, this can potentially lead to problems. That said, if we have transactional metadata implemented and can reliably track versions in the cluster, and switch between epochs in which an effective minimal version for all coordinators supports a specific version of client state, this should not be a problem at all. On Fri, Oct 22, 2021 at 9:07 AM Jacek Lewandowski < lewandowski.ja...@gmail.com> wrote: > Hi > > I wanted to start working on extending the paging mechanism to support > memory based limits. My question is whether such a change requires a CEP or > discussion in Jira is enough? > > Thanks, > - - -- --- ----- -------- ------------- > Jacek Lewandowski > -- alex p
