hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Wei-Chiu Chuang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-12886) Ignore minReplication for block recovery
Date Fri, 20 Jul 2018 20:47:02 GMT

    [ https://issues.apache.org/jira/browse/HDFS-12886?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16551264#comment-16551264

Wei-Chiu Chuang commented on HDFS-12886:

I stand corrected: the data corruption bug caused by concurrent recoverLease() and close()
is *NOT* fixed by HDFS-12886. In fact, it is accidentally fixed by HDFS-8909.

Basically, HDFS-8909 added an assertion that if a block recovery command is being asked, it
checks if the block is actually under recovery. So after HDFS-8909, the recoverLease() call
is ignored. However, note that under a more delicate condition, concurrent recoverLease()
+ close() could still corrupt block. Ignoring block recovery if the block is not under recovery
is not the ultimate solution. Instead, the close() should be ignored/rejected when block is
being recovered.

> Ignore minReplication for block recovery
> ----------------------------------------
>                 Key: HDFS-12886
>                 URL: https://issues.apache.org/jira/browse/HDFS-12886
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: hdfs, namenode
>            Reporter: Lukas Majercak
>            Assignee: Lukas Majercak
>            Priority: Major
>             Fix For: 3.1.0, 2.10.0, 2.9.1, 3.0.3
>         Attachments: HDFS-12886.001.patch, HDFS-12886.002.patch, HDFS-12886_branch-2.000.patch
> Ignore minReplication for blocks that went through recovery, and allow NN to complete
them and replicate.

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: hdfs-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-help@hadoop.apache.org

View raw message