hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jay Booth (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HDFS-1123) Need HDFS Protocol Specification
Date Sun, 02 May 2010 18:35:00 GMT

     [ https://issues.apache.org/jira/browse/HDFS-1123?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Jay Booth updated HDFS-1123:

    Attachment: datanode-reads.txt

Here's a first pass at the protocol for reading from a DataNode.  Tried to be pretty thorough,
but it could probably use clarification in a few places and might have an error.  Could definitely
use checking for style.  Should we put this on a wiki somewhere?

The NameNode part of the process is entirely ipc.Server derived with Writables, I believe,
so for that item API and conceptual documentation is probably more useful than an itemized
description of each writable type.

> Need HDFS Protocol Specification
> --------------------------------
>                 Key: HDFS-1123
>                 URL: https://issues.apache.org/jira/browse/HDFS-1123
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>          Components: documentation
>            Reporter: bc Wong
>         Attachments: datanode-reads.txt
> It'd be great to document (in a spec, not in the code) the HDFS wire protocol:
> * The layout of the different request and reply messages.
> * The semantics of the various calls.
> * The semantics of the various fields.
> For example, I stumbled upon the goldmine of comments around DataNode.java:1150. It looks
correct, but the version number of 9 doesn't inspire confidence that it's up-to-date. (It's
also a random place to put such an important comment.)
> Having a formal spec is a big step forward for compatibility. It also highlights design
decisions and helps with protocol evolution.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message