hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Brandon Li (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HDFS-5259) Support client which combines appended data with old data before sends it to NFS server
Date Thu, 03 Oct 2013 01:08:42 GMT

     [ https://issues.apache.org/jira/browse/HDFS-5259?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Brandon Li updated HDFS-5259:
-----------------------------

    Attachment: HDFS-5259.000.patch

Upload a patch to address the jumbo write for append. Basically NFS gateway tries to identify
the special access pattern and modifies the write to an real append write. 
The observed access pattern is, after reopen a file to append, if the previous written data
is still in client Kernel buffer cache, the client might combine the previous written data
with newly appended data for one WRITE NFS call for the first write.  Inside NFS gateway,
it checks if this is the only write, if yes, it drops the overlapped section and appends the
new data section.

> Support client which combines appended data with old data before sends it to NFS server
> ---------------------------------------------------------------------------------------
>
>                 Key: HDFS-5259
>                 URL: https://issues.apache.org/jira/browse/HDFS-5259
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: nfs
>            Reporter: Yesha Vora
>            Assignee: Brandon Li
>         Attachments: HDFS-5259.000.patch
>
>
> The append does not work with some Linux client. The Client gets "Input/output Error"
when it tries to append. And NFS server considers it as random write and fails the request.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message