You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@beam.apache.org by "Luke Cwik (Jira)" <ji...@apache.org> on 2020/04/23 22:45:00 UTC
[jira] [Commented] (BEAM-9488) Python SDK sending unexpected
MonitoringInfo
[ https://issues.apache.org/jira/browse/BEAM-9488?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17090981#comment-17090981 ]
Luke Cwik commented on BEAM-9488:
---------------------------------
I was able to reproduce this still at commit 48db942ebee8d2a12298d7f22b6f2967674f7041 with ./gradlew :sdks:python:test-suites:portable:py2:flinkValidatesRunner and some simple logging in Flinks progress handler: https://github.com/apache/beam/blob/a905f69056c1ae9bd27a577471800d3623b47e49/runners/flink/src/main/java/org/apache/beam/runners/flink/translation/functions/FlinkExecutableStageFunction.java#L154
> Python SDK sending unexpected MonitoringInfo
> --------------------------------------------
>
> Key: BEAM-9488
> URL: https://issues.apache.org/jira/browse/BEAM-9488
> Project: Beam
> Issue Type: Bug
> Components: sdk-py-core
> Reporter: Ruoyun Huang
> Assignee: Luke Cwik
> Priority: Minor
>
> element_count metrics is supposed to be tied with pcollection ids, but by inspecting what is sent over by python sdk, we see there are monitoringInfo sent wit ptransforms in it.
> [Double checked the job graph, these seem to be redundant. i.e. the corresponding pcollection does have its own MonitoringInfo reported.]
> Likely a bug.
> Proof:
> urn: "beam:metric:element_count:v1"
> type: "beam:metrics:sum_int_64"
> metric {
> counter_data {
> int64_value: 1
> }
> }
> labels {
> key: "PTRANSFORM"
> value: "start/MaybeReshuffle/Reshuffle/RemoveRandomKeys-ptransform-85"
> }
> labels {
> key: "TAG"
> value: "None"
> }
> timestamp {
> seconds: 1583949073
> nanos: 842402935
> }
--
This message was sent by Atlassian Jira
(v8.3.4#803005)