Thank you for getting the partial rollback in. I will close https://issues.apache.org/jira/browse/BEAM-5180 as fixed. Ankur: if you have a more nuanced fix in mind, please open a new JIRA ticket to track and update us on this thread.
On Fri, Aug 24, 2018 at 10:42 AM Ankur Goenka <[email protected]> wrote: > Replies on the Jira and PR. > For now we should go ahead with rollback to unblock 2.7 > > On Fri, Aug 24, 2018 at 10:21 AM Udi Meiri <[email protected]> wrote: > >> +Ankur Goenka <[email protected]> (Kenneth is out of office) >> >> On Fri, Aug 24, 2018 at 3:20 AM Tim Robertson <[email protected]> >> wrote: >> >>> Thanks Jozef for bringing this to dev@ and your work in reporting Jiras >>> and offering fixes. >>> >>> I propose we consider BEAM-5180, BEAM-2277 blockers on 2.7.0. They break >>> word count and file IO writing on HDFS unless the workaround is used (see >>> BEAM-2277 commentary). >>> >>> In addition the performance of writing to HDFS is bad (to the point of >>> being unusable) due to BEAM-5036 (blocked by BEAM-4861) which you have also >>> observed. I don’t think we’ll make the cut for 2.7.0 on those because I >>> anticipate significant testing is needed. I will focus on that as soon as I >>> have time but I propose we make those blockers on 2.8.0 though. >>> >>> >>> >>> On Fri, Aug 24, 2018 at 7:53 AM Jozef Vilcek <[email protected]> >>> wrote: >>> >>>> Hello, >>>> >>>> does this JIRA have a change to be part of 2.7.0 release? >>>> https://issues.apache.org/jira/browse/BEAM-5180 >>>> >>>> It is rather simple ask, but was not decided yet if attached PR is a >>>> correct way of fixing it. >>>> >>>> Thanks, >>>> Jozef >>>> >>>
