Re: [CANCEL][VOTE] Release 2.3.0, release candidate #1

+1 for option 1. It is almost certainly repeated proto parsing. That is fixed for ParDo via generalized caching but not for other transforms. Flink also has a proto round trip, but the design might avoid the problem anyhow. We should investigate before releasing, or just preemptively roll it back

Re: [DISCUSS] What to do about widespread KinesisIO breakage

Kenn, JB, Thank you for raising this question. I was working mostly on writing support for KinesisIO (PR is waiting to be merged) but I’ll take a look on flaky read tests as well. Regards, Alexey > On 6 Feb 2018, at 06:27, Jean-Baptiste Onofré wrote: > > Yes, I don't

Build failed in Jenkins: beam_PostRelease_NightlySnapshot #18

See Changes: [iemejia] [BEAM-3592] Fix spark-runner profile for Nexmark after move to Spark 2.x [Pablo] Moving User metrics to be in the PTransform proto for Fn API. [ccy] [BEAM-3566] Replace

Re: [CANCEL][VOTE] Release 2.3.0, release candidate #1

Possibly stupid question: are new RCs created from master, or by cherrypicks of desired changes on top of previous RCs? (it affects whether or not my recently merged change on "master" will be in 2.3.0; I'd like it to be :) ) On Tue, Feb 6, 2018 at 1:49 PM Lukasz Cwik wrote: >

Re: [CANCEL][VOTE] Release 2.3.0, release candidate #1

Changes are cherrypicked to the release-2.3.0 branch, or applied directly there. On Tue, Feb 6, 2018 at 2:08 PM, Eugene Kirpichov wrote: > Possibly stupid question: are new RCs created from master, or by > cherrypicks of desired changes on top of previous RCs? > (it

Re: [CANCEL][VOTE] Release 2.3.0, release candidate #1

+1 for rollback as well. On Tue, Feb 6, 2018 at 12:37 PM, Reuven Lax wrote: > +1 for the rollback. > > On Tue, Feb 6, 2018 at 9:24 AM, Kenneth Knowles wrote: > >> +1 for option 1. It is almost certainly repeated proto parsing. That is >> fixed for ParDo via

Re: [CANCEL][VOTE] Release 2.3.0, release candidate #1

+1 for the rollback. On Tue, Feb 6, 2018 at 9:24 AM, Kenneth Knowles wrote: > +1 for option 1. It is almost certainly repeated proto parsing. That is > fixed for ParDo via generalized caching but not for other transforms. Flink > also has a proto round trip, but the design

Re: Jenkins build is still unstable: beam_PostCommit_Java_ValidatesRunner_Dataflow #4896

This postcommit has had a failing ViewTest for a couple days. I'm looking into it. https://issues.apache.org/jira/browse/BEAM-3627 On Tue, Feb 6, 2018 at 11:25 AM, Apache Jenkins Server < jenk...@builds.apache.org> wrote: > See

Re: Jenkins build is still unstable: beam_PostCommit_Java_ValidatesRunner_Dataflow #4896

Looks definitely related to the valueInEmptyWindows change. Had ValidatesRunner green on the PR but likely Dataflow changed underneath. On Tue, Feb 6, 2018 at 11:30 AM, Kenneth Knowles wrote: > This postcommit has had a failing ViewTest for a couple days. I'm looking > into it.

Python tests broken by https://builds.apache.org/view/A-D/view/Beam/job/beam_PostCommit_Python_Verify/4152/

Hi all, It looks like https://github.com/apache/beam/pull/4529 may have broken https://builds.apache.org/view/A-D/view/Beam/job/beam_PostCommit_Python_Verify/4152/. I filed https://issues.apache.org/jira/browse/BEAM-3628 since I didn't manage to find a tracking ticket. Kenn

Re: [CANCEL][VOTE] Release 2.3.0, release candidate #1

Hi all, As you might have seen, I found the cause of performance degradation on the direct runner (BEAM-3617). We have basically three options for RC2: 1. We revert the change for 2.3.0. I think it's the fastest and more secure way. I created a PR for that

Build failed in Jenkins: beam_Release_NightlySnapshot #677

See Changes: [Pablo] Moving User metrics to be in the PTransform proto for Fn API. [ccy] [BEAM-3566] Replace apply_* hooks in DirectRunner with [ccy] Address reviewer comments [jiangkai] Covariance

Re: [CANCEL][VOTE] Release 2.3.0, release candidate #1

Hi Reuven, it's what I'm suspecting. git bisect should give us more information (still in progress, 3 more steps to complete ;)). I keep you posted. Regards JB On 02/06/2018 08:36 AM, Reuven Lax wrote: > Could this be related to any of the portability changes? > > On Mon, Feb 5, 2018 at 7:51