hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vinayakumar B (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-6247) Avoid timeouts for replaceBlock() call by sending intermediate responses to Balancer
Date Fri, 25 Jul 2014 17:55:39 GMT

    [ https://issues.apache.org/jira/browse/HDFS-6247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14074674#comment-14074674
] 

Vinayakumar B commented on HDFS-6247:
-------------------------------------

I feel 30 seconds or may be 0.5 * socketTimeout would be fine.
Since this socketTimeout is based on the configuration at client side, only it can be assumed
that this configuration is same in both client and datanode.
bq.  How about .1 * socketTimeout for the hearbeat interval? Does that make sense?
I feel making at same as socketTimeout can create timeout problems at the boundary. Like datanode
might send the response, but just before receiving that response client might get timeout.

So I feel better will be half of it. i.e *0.5 * socketTimeout*  will that be fine?

> Avoid timeouts for replaceBlock() call by sending intermediate responses to Balancer
> ------------------------------------------------------------------------------------
>
>                 Key: HDFS-6247
>                 URL: https://issues.apache.org/jira/browse/HDFS-6247
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: balancer, datanode
>    Affects Versions: 2.4.0
>            Reporter: Vinayakumar B
>            Assignee: Vinayakumar B
>         Attachments: HDFS-6247.patch, HDFS-6247.patch, HDFS-6247.patch, HDFS-6247.patch
>
>
> Currently there is no response sent from target Datanode to Balancer for the replaceBlock()
calls.
> Since the Block movement for balancing is throttled, complete block movement will take
time and this could result in timeout at Balancer, which will be trying to read the status
message.
>  
> To Avoid this during replaceBlock() call in in progress Datanode  can send IN_PROGRESS
status messages to Balancer to avoid timeouts and treat BlockMovement as  failed.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message