[
https://issues.apache.org/jira/browse/HDFS-16968?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17772512#comment-17772512
]
ASF GitHub Bot commented on HDFS-16968:
---------------------------------------
zhangshuyan0 commented on PR #6125:
URL: https://github.com/apache/hadoop/pull/6125#issuecomment-1750361000
If you want to improve the reliability of 2-replica writing, it is
recommended that you directly configure
`dfs.client.block.write.replace-datanode-on-failure.policy` to `ALWAYS`. The
current changes conflict with the design ideas of the `DEFAULT` policy. See:
https://github.com/apache/hadoop/blob/daa78adc888704e5688b84b404573ed1e28012db/hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml#L765-L783
> Corrupted blocks appear in files written with two replicas
> ----------------------------------------------------------
>
> Key: HDFS-16968
> URL: https://issues.apache.org/jira/browse/HDFS-16968
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: dfsclient
> Reporter: leo sun
> Priority: Major
> Labels: pull-request-available
>
> If the 2-replica files fail in the process of writing, only 1-replica will be
> recovered.
> If the recovered replica is broken before the block is completed, the file
> will miss.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
