Namespace recovery from the secondary image should be documented.

• Key: HADOOP-3541
• Type: Task
• Status: Closed
• Resolution: Fixed
• Priority: Blocker
• Assignee: Konstantin Shvachko
• Reporter: Konstantin Shvachko
• Component/s: documentation
• Affects Version/s: 0.18.0
• Fix Version/s: 0.18.0

Description

HADOOP-2585 introduced a new feature which lets starting name-node from an image stored on the secondary node.
This need to be documented in "HDFS Admin Guide".

Comments

[Konstantin Shvachko]

Here is the documentation.

[Lohit Vijayarenu]

+1 Looks good

• Can we describe create an empty storage directory, to be more explicit saying we should not have name/current directory
• Would it be good idea to document about fs.checkpoint.period and fs.checkpoint.size.

PS: Shouldnt the patch contain only .xml file changes?

[Konstantin Shvachko]

It is a good idea to document fs.checkpoint.period and .size properties. I did that.
By saying "create empty directory" I actually mean creating an empty directory that is the one that does not
contain anything inside. Don' know how to clarify it better.
Yes you are right that the implementation checks only that the directory does not contain a legal image,
but I think the documentation should say empty directory, that way there will be no confusion if we ever
change any thing in the internal directory structure.
I thought we commit xml files as well as the generated by forrest html and pdf files. So my patch contains all three.
Please let me know if this is wrong.

[Lohit Vijayarenu]

+1 Looks good.

[Konstantin Shvachko]

I just committed this.

[Hudson]

Integrated in Hadoop-trunk #529 (See http://hudson.zones.apache.org/hudson/job/Hadoop-trunk/529/)