[
https://issues.apache.org/jira/browse/HDFS-4167?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14010822#comment-14010822
]
Guo Ruijing commented on HDFS-4167:
-----------------------------------
I agree that it's better to keep it as a standalone HDFS metadata operation. It
is easy to restore any snapshot if block is copied out for append as HDFS-6087
proposal.
what's change in append?
1. file f1 includes (Block1, Block2, Block3)
2. append to f1
a) client request block3 information from namenode
b) client request to datanode to copy block3 as block4
c) append to block4
d) commit block4 to namenode
what happend in snapshot?
snap1: f1 include (block1, block2, block3)
snap2: f2 include (block1, block2, block4)
how to restore snapshot? just restore snap1 as current file since no partial
blocks are shared by different snap.
> Add support for restoring/rolling back to a snapshot
> ----------------------------------------------------
>
> Key: HDFS-4167
> URL: https://issues.apache.org/jira/browse/HDFS-4167
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: namenode
> Reporter: Suresh Srinivas
> Assignee: Jing Zhao
> Attachments: HDFS-4167.000.patch, HDFS-4167.001.patch,
> HDFS-4167.002.patch, HDFS-4167.003.patch, HDFS-4167.004.patch
>
>
> This jira tracks work related to restoring a directory/file to a snapshot.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
