You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@airflow.apache.org by "ASF subversion and git services (Jira)" <ji...@apache.org> on 2020/02/02 11:22:00 UTC

[jira] [Commented] (AIRFLOW-6522) Sensors in reschedule mode with S3TaskHandler can cause log duplication

    [ https://issues.apache.org/jira/browse/AIRFLOW-6522?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17028393#comment-17028393 ] 

ASF subversion and git services commented on AIRFLOW-6522:
----------------------------------------------------------

Commit c7ad2c3f1a5c39999fd46edbbc390b1229e01e64 in airflow's branch refs/heads/v1-10-test from rconroy293
[ https://gitbox.apache.org/repos/asf?p=airflow.git;h=c7ad2c3 ]

[AIRFLOW-6522] Clear task log file before starting to fix duplication in S3TaskHandler (#7120)

The same task instance (including try number) can be run on a worker
when using a sensor in "reschedule" mode. Accordingly, this clears the
local log file when re-initializing the logger so that the old log
lines aren't uploaded again when the logger is closed.

(cherry picked from commit 88608caa56bf3621807af860a6a378242220de47)


> Sensors in reschedule mode with S3TaskHandler can cause log duplication
> -----------------------------------------------------------------------
>
>                 Key: AIRFLOW-6522
>                 URL: https://issues.apache.org/jira/browse/AIRFLOW-6522
>             Project: Apache Airflow
>          Issue Type: Bug
>          Components: logging
>    Affects Versions: 1.10.6
>            Reporter: Robert Conroy
>            Assignee: Robert Conroy
>            Priority: Minor
>             Fix For: 1.10.8
>
>
> With sensors using {{reschedule}} mode and {{S3TaskHandler}} for logging, the task instance log gets a bunch of duplicate messages. I believe this is happening because contents of the local log file are appended to what's already in S3. The local log file may contain log messages that have already been uploaded to S3 if the task is sent back to a worker that had already processed a poke for that task instance.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)