[jira] [Updated] (HDFS-6840) Clients are always sent to the same datanode when read is off rack

["Andrew Wang (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HDFS-6840?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Andrew Wang updated HDFS-6840:
------------------------------
    Attachment: hdfs-6840.003.patch

Sorry for the delay in revving this. New patch removes the stale comment as per Jason's feedback.

Daryn, do you mind if we fix any seed issues in a separate JIRA? I think we depend on this behavior in other places too, so if/when said JDK change does hit, we could address them all at once.

> Clients are always sent to the same datanode when read is off rack
> ------------------------------------------------------------------
>
>                 Key: HDFS-6840
>                 URL: https://issues.apache.org/jira/browse/HDFS-6840
>             Project: Hadoop HDFS
>          Issue Type: Bug
>    Affects Versions: 2.5.0
>            Reporter: Jason Lowe
>            Assignee: Andrew Wang
>            Priority: Critical
>         Attachments: hdfs-6840.001.patch, hdfs-6840.002.patch, hdfs-6840.003.patch
>
>
> After HDFS-6268 the sorting order of block locations is deterministic for a given block and locality level (e.g.: local, rack. off-rack), so off-rack clients all see the same datanode for the same block.  This leads to very poor behavior in distributed cache localization and other scenarios where many clients all want the same block data at approximately the same time.  The one datanode is crushed by the load while the other replicas only handle local and rack-local requests.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)