hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Yi Liu (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-8925) Move BlockReader to hdfs-client
Date Thu, 20 Aug 2015 14:07:51 GMT

    [ https://issues.apache.org/jira/browse/HDFS-8925?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14704957#comment-14704957

Yi Liu commented on HDFS-8925:

Thanks [~zhz] for raising the discussion and involving me here, I want to supplement some
background: for continuous block replication, DN get replica from local machine and transfer
to target DN(s); while for striped block reconstruction, DN needs to fetch some blocks of
the block group to reconstruct the missed block, currently DN uses the {{BlockReader}} to
fetch those blocks. For implementing some new logic to fetch the blocks in DN, I didn't see
the necessary for now, one exception is if the replica is DN local, we should read it directly.

Back to the issue, I think it's OK to move BlockReader to hdfs-client from the rule we follow
currently: if some component is required by both hdfs client and side side, then we move it
to the hdfs-client.  Actually as I mentioned in this umbrella, for long term, we may need
a "hdfs-common" project to put the common part. But at this point, it should be OK to put
it in hdfs-client side, since hdfs depends on hdfs-client and DN still can use it.

> Move BlockReader to hdfs-client
> -------------------------------
>                 Key: HDFS-8925
>                 URL: https://issues.apache.org/jira/browse/HDFS-8925
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: build
>            Reporter: Mingliang Liu
>            Assignee: Mingliang Liu
>             Fix For: 2.8.0
> This jira tracks the effort of moving the {{BlockReader}} class into the hdfs-client

This message was sent by Atlassian JIRA

View raw message