[
https://issues.apache.org/jira/browse/FLINK-8357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17195991#comment-17195991
]
Robert Metzger commented on FLINK-8357:
---------------------------------------
I'm a bit skeptical if a one-hour rotation is the best thing for Flink. When a
job is just running fine for a long time, there might be hardly any log
statements (say you are checkpointing very infrequently and there are no errors
.. the JobManager won't log anything).
The problem is then that the web ui won't show any entries.
I would be more open to the idea of using size-based rotation. I believe this
is a more graceful approach, as users with low-volume logs will keep the
current experience, while users with a lot of logging will have to cope with
this change.
> enable rolling in default log settings
> --------------------------------------
>
> Key: FLINK-8357
> URL: https://issues.apache.org/jira/browse/FLINK-8357
> Project: Flink
> Issue Type: Improvement
> Components: Build System, Deployment / Scripts
> Reporter: Mingmin Xu
> Assignee: Jörn Kottmann
> Priority: Major
> Labels: usability
> Time Spent: 10m
> Remaining Estimate: 0h
>
> The release packages uses {{org.apache.log4j.FileAppender}} for log4j and
> {{ch.qos.logback.core.FileAppender}} for logback, which could results in very
> large log files.
> For most cases, if not all, we need to enable rotation in a production
> cluster, and I suppose it's a good idea to make rotation as default.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
