[
https://issues.apache.org/jira/browse/HDFS-4749?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13641225#comment-13641225
]
Jing Zhao commented on HDFS-4749:
---------------------------------
Another issue of using path is that it cannot handle (or it's very difficult to
handle) the fsimage after the following operations:
1. Create directory /dir1, /dir2, /dir1/foo, and /dir1/foo/bar
2. Allow and make snapshot s1 on /dir1
3. Rename /dir1/foo/bar to /dir2/bar
4. Rename /dir1/foo to /dir2/bar/foo
The current implementation will write a path /dir1/.snapshot/s1/foo/bar/foo to
the fsimage, which cannot be correctly resolved.
> Use INodeId to identify the corresponding directory node for FSImage
> saving/loading
> -----------------------------------------------------------------------------------
>
> Key: HDFS-4749
> URL: https://issues.apache.org/jira/browse/HDFS-4749
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: datanode, namenode
> Reporter: Jing Zhao
> Assignee: Jing Zhao
>
> Currently in fsimage, we use the path to locate a directory node for later
> loading, i.e., when loading a subtree from fsimage, we first read the path of
> the directory node, and resolve the path to identify the directory node. This
> brings extra complexity since we need to generate path for directory nodes in
> both the current tree and snapshot copies.
> As a simplification, we can use INodeId to identify the directory node.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
