[
https://issues.apache.org/jira/browse/BEAM-3937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17429037#comment-17429037
]
Brian Hulette edited comment on BEAM-3937 at 10/14/21, 11:01 PM:
-----------------------------------------------------------------
This ticket is tracking whether or not we should make the output_counter
enabled by default. There are still TODOs in the code for it:
https://github.com/apache/beam/blob/05428866cdbf1ea8e4c1789dd40327673fd39451/sdks/python/apache_beam/runners/common.py#L1198
[~robertwb] what do you think we should do here? If the metrics in the [comment
above|https://issues.apache.org/jira/browse/BEAM-3937?focusedCommentId=16472241&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16472241]
are still accurate it seems like we should keep it hidden behind a flag,
that's ~18% per-element overhead.
Is there a reason to enable this by default?
was (Author: bhulette):
This ticket is tracking whether or not we should make the output_counter
enabled by default. There are still TODOs in the code for it:
https://github.com/apache/beam/blob/05428866cdbf1ea8e4c1789dd40327673fd39451/sdks/python/apache_beam/runners/common.py#L1198
[~robertwb] what do you think we should do here? If the metrics in the comment
above
(https://issues.apache.org/jira/browse/BEAM-3937?focusedCommentId=16472241&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16472241)
are still accurate it seems like we should keep it hidden behind a flag,
that's ~18% per-element overhead.
Is there a reason to enable this by default?
> Track performance of output_counter before feature enabled
> ----------------------------------------------------------
>
> Key: BEAM-3937
> URL: https://issues.apache.org/jira/browse/BEAM-3937
> Project: Beam
> Issue Type: Task
> Components: sdk-py-harness
> Reporter: Boyuan Zhang
> Priority: P3
>
> To track performance of output_counter before feature enabled by default.
> Need to make sure there is no adversely reflect before enbale
> Related PR: [https://github.com/apache/beam/pull/4741]
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
