You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@arrow.apache.org by "David Li (Jira)" <ji...@apache.org> on 2021/12/10 17:12:00 UTC

[jira] [Commented] (ARROW-15061) [C++] Add logging for kernel functions and exec plan nodes

    [ https://issues.apache.org/jira/browse/ARROW-15061?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17457279#comment-17457279 ] 

David Li commented on ARROW-15061:
----------------------------------

[PR#11906|https://github.com/apache/arrow/pull/11906] does this for kernel functions, but not for nodes, as a starter/something to build off of. It also doesn't add a span for the entire exec plan (that may be more annoying to #ifdef guard properly).

> [C++] Add logging for kernel functions and exec plan nodes
> ----------------------------------------------------------
>
>                 Key: ARROW-15061
>                 URL: https://issues.apache.org/jira/browse/ARROW-15061
>             Project: Apache Arrow
>          Issue Type: Sub-task
>          Components: C++
>            Reporter: Weston Pace
>            Priority: Major
>
> Using OT we should add spans that report the runtime of kernel functions.  We should  ensure that, if OT is disabled, there is no significant overhead.
> In addition we should add spans that report the runtime of each exec node.  This may be a little trickier than it would first seem because pipeline breakers like join, etc. don't do all of their work inside of InputReceived.  Those could potentially be added later in follow-up PRs if it is getting to be too tricky.  The simple nodes like FilterNode and ProjectNode should be more straightforward.  These spans should be the parent of any kernel functions launched by that node.
> There should also be a span for an entire exec plan run that will be the parent of all of this.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)