[ 
https://issues.apache.org/jira/browse/HDFS-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13489659#comment-13489659
 ] 

Aaron T. Myers commented on HDFS-2802:
--------------------------------------

bq. ** Current design solves single-client consistency but not multiple-client 
consistency.

This is with respect to multiple distinct DFSClients which are simultaneously 
writing to open files at the time of snapshot creation. All of the designs 
discussed so far discuss single client ordering of namespace operations/data 
writes, but not multiple clients. The implication is that multi client 
consistency would likely require such far-reaching changes as to be unrealistic.

bq. *** Would admins like to restrict based on users as well as directories for 
snapshots?

We discussed the possibility of adding a configurable white list of users who 
were allowed to create snapshots, and/or an ACL per snapshottable directory of 
users allowed to create snapshots of that directory.

bq. ** Current design adds extra overhead for every inode,

The design for storing snapshots in the NN in the proposal I posted would add 
integer start_snap and end_snap fields to every INode in the FS in order to 
track which snapshots a given INode should be included in. This is probably the 
biggest difference between the design proposed by Suresh/Nicholas and the 
design I've proposed. That proposal adds a "diff" per INodeDirectory which has 
been snapshotted to keep track of changes made to that directory across 
different snapshots. This has the advantage of saving NN memory space for 
files/directories which have never had snapshots created of them.
                
> Support for RW/RO snapshots in HDFS
> -----------------------------------
>
>                 Key: HDFS-2802
>                 URL: https://issues.apache.org/jira/browse/HDFS-2802
>             Project: Hadoop HDFS
>          Issue Type: New Feature
>          Components: data-node, name-node
>            Reporter: Hari Mankude
>            Assignee: Hari Mankude
>         Attachments: HDFS-2802.20121101.patch, 
> HDFS-2802-meeting-minutes-121101.txt, HDFSSnapshotsDesign.pdf, snap.patch, 
> snapshot-design.pdf, snapshot-design.tex, snapshot-one-pager.pdf, 
> Snapshots20121018.pdf, Snapshots20121030.pdf
>
>
> Snapshots are point in time images of parts of the filesystem or the entire 
> filesystem. Snapshots can be a read-only or a read-write point in time copy 
> of the filesystem. There are several use cases for snapshots in HDFS. I will 
> post a detailed write-up soon with with more information.

--
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

Reply via email to