You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@beam.apache.org by Daniel Oliveira <da...@google.com> on 2022/03/24 00:20:35 UTC
Potentially release-blocking performance regressions in Streaming Python load tests.
Hey everyone,
I'm the current release manager for version 2.38.0, which I'm in the
process of cutting. As per the Beam Release Guide
<https://beam.apache.org/contribute/release-guide/#4-investigate-performance-regressions>
I checked Beam's performance metrics and found regressions in Streaming
Python's ParDo Load Tests
<http://metrics.beam.apache.org/d/MOi-kf3Zk/pardo-load-tests?orgId=1&var-processingType=streaming&var-sdk=python>,
and GBK Load Tests
<http://metrics.beam.apache.org/d/UYZ-oJ3Zk/gbk-load-tests?orgId=1&var-processingType=streaming&var-sdk=python&from=now-30d&to=now>,
both starting around March 17.
As per the guide, I've filed a JIRA
<https://issues.apache.org/jira/browse/BEAM-14163> and I'm informing
everyone with this email. I won't be delaying the cut for this fix, but the
procedure is to block the release until a fix can be implemented and
cherry-picked in. So this is essentially a call for help figuring out this
performance regression ASAP.
Thanks,
Daniel Oliveira
Re: Potentially release-blocking performance regressions in Streaming Python load tests.
Posted by Kerry Donny-Clark <ke...@google.com>.
Yes, release blocking. @Valentyn Tymofieiev <va...@google.com> can you
root cause this?
On Thu, Mar 24, 2022 at 2:45 PM Ahmet Altay <al...@google.com> wrote:
> @Valentyn Tymofieiev <va...@google.com> @Kerry Donny-Clark
> <ke...@google.com> - in case you would consider these release blockers.
> (please feel free to add others who might comment.)
>
> On Wed, Mar 23, 2022 at 5:20 PM Daniel Oliveira <da...@google.com>
> wrote:
>
>> Hey everyone,
>>
>> I'm the current release manager for version 2.38.0, which I'm in the
>> process of cutting. As per the Beam Release Guide
>> <https://beam.apache.org/contribute/release-guide/#4-investigate-performance-regressions>
>> I checked Beam's performance metrics and found regressions in Streaming
>> Python's ParDo Load Tests
>> <http://metrics.beam.apache.org/d/MOi-kf3Zk/pardo-load-tests?orgId=1&var-processingType=streaming&var-sdk=python>,
>> and GBK Load Tests
>> <http://metrics.beam.apache.org/d/UYZ-oJ3Zk/gbk-load-tests?orgId=1&var-processingType=streaming&var-sdk=python&from=now-30d&to=now>,
>> both starting around March 17.
>>
>> As per the guide, I've filed a JIRA
>> <https://issues.apache.org/jira/browse/BEAM-14163> and I'm informing
>> everyone with this email. I won't be delaying the cut for this fix, but the
>> procedure is to block the release until a fix can be implemented and
>> cherry-picked in. So this is essentially a call for help figuring out this
>> performance regression ASAP.
>>
>> Thanks,
>> Daniel Oliveira
>>
>
Re: Potentially release-blocking performance regressions in Streaming Python load tests.
Posted by Ahmet Altay <al...@google.com>.
@Valentyn Tymofieiev <va...@google.com> @Kerry Donny-Clark
<ke...@google.com> - in case you would consider these release blockers.
(please feel free to add others who might comment.)
On Wed, Mar 23, 2022 at 5:20 PM Daniel Oliveira <da...@google.com>
wrote:
> Hey everyone,
>
> I'm the current release manager for version 2.38.0, which I'm in the
> process of cutting. As per the Beam Release Guide
> <https://beam.apache.org/contribute/release-guide/#4-investigate-performance-regressions>
> I checked Beam's performance metrics and found regressions in Streaming
> Python's ParDo Load Tests
> <http://metrics.beam.apache.org/d/MOi-kf3Zk/pardo-load-tests?orgId=1&var-processingType=streaming&var-sdk=python>,
> and GBK Load Tests
> <http://metrics.beam.apache.org/d/UYZ-oJ3Zk/gbk-load-tests?orgId=1&var-processingType=streaming&var-sdk=python&from=now-30d&to=now>,
> both starting around March 17.
>
> As per the guide, I've filed a JIRA
> <https://issues.apache.org/jira/browse/BEAM-14163> and I'm informing
> everyone with this email. I won't be delaying the cut for this fix, but the
> procedure is to block the release until a fix can be implemented and
> cherry-picked in. So this is essentially a call for help figuring out this
> performance regression ASAP.
>
> Thanks,
> Daniel Oliveira
>