[
https://issues.apache.org/jira/browse/HDFS-17924?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18082803#comment-18082803
]
ASF GitHub Bot commented on HDFS-17924:
---------------------------------------
kokonguyen191 opened a new pull request, #8509:
URL: https://github.com/apache/hadoop/pull/8509
### Description of PR
Currently, all 3 kinds of NNs trigger rollEditLog at some time interval and
based on the number of edits using a threshold check.
If a subcluster has a lot of observers, active can roll very frequently, up
to dozens of times per hour. This ticket aims improve that via 2 configurable
features:
- Only active should perform rollEditLog periodically, and it should
guarantee a roll as long as there's something to roll, at a reasonable interval.
- Standby and observer NNs can opt out of sending rollEditLog.
### How was this patch tested?
UT, and a small test cluster of a few nodes.
> Improve NameNode edit log rolling controls
> ------------------------------------------
>
> Key: HDFS-17924
> URL: https://issues.apache.org/jira/browse/HDFS-17924
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs
> Reporter: Felix N
> Assignee: Felix N
> Priority: Minor
>
> Currently, all 3 kinds of NNs trigger rollEditLog at some time interval and
> based on the number of edits using a threshold check.
> If a subcluster has a lot of observers, active can roll very frequently, up
> to dozens of times per hour. This ticket aims improve that via 2 configurable
> features:
> * Only active should perform rollEditLog periodically, and it should
> guarantee a roll as long as there's something to roll, at a reasonable
> interval.
> * Standby and observer NNs can opt out of sending rollEditLog.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
