airflow-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ivan Vergiliev (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (AIRFLOW-1035) Exponential backoff retry logic should use 2 as base
Date Mon, 10 Apr 2017 21:20:41 GMT

    [ https://issues.apache.org/jira/browse/AIRFLOW-1035?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15963492#comment-15963492
] 

Ivan Vergiliev commented on AIRFLOW-1035:
-----------------------------------------

[~criccomini] I'm not really in a rush (disabled exponential backoff as soon as I found this),
but the PR for this issue fixes a bug that existed in the old logic as well: https://github.com/apache/incubator-airflow/pull/2196#issuecomment-290045055
. It effectively causes the scheduler to stop scheduling any new tasks after an operator with
exponential backoff fails a few times. So I'm guessing it might be worth including in the
bug fix release.

> Exponential backoff retry logic should use 2 as base
> ----------------------------------------------------
>
>                 Key: AIRFLOW-1035
>                 URL: https://issues.apache.org/jira/browse/AIRFLOW-1035
>             Project: Apache Airflow
>          Issue Type: Bug
>            Reporter: Alex Guziel
>            Assignee: Alex Guziel
>             Fix For: 1.9.0
>
>
> Right now, the exponential backoff logic computes it as 
> (retry_period) ^ (retry_number) instead of retry_period * 2 ^ retry_number. 
> See https://en.wikipedia.org/wiki/Exponential_backoff



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message