[
https://issues.apache.org/jira/browse/HDFS-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13500036#comment-13500036
]
Sanjay Radia commented on HDFS-2802:
------------------------------------
Minutes of Meetup on Snapshots for HDFS
Date: 8 Nov 2012, Location: Hortonworks Offices
Attendance: approx 25 to 30 in person and about 7 calls into webex.
Format: Suresh presented slides (attached in this Jira HDFS-2802) and a
discussion followed along with the slides.
*Main areas of discussions*
# CLI - most felt that we shouldn't use the fs command, introduce a new command
instead.
# Scheduling snapshots - most folks felt it was best to leave it outside
# File attributes (permissions, replication factor etc)
** Most folks felt that we should preserve the original attributes in the
snapshots, however no access time to be recorded for snapshots.
** # replicas is max of current and the snapshots
# Quotas
** Everyone agreed that snapshot metadata should count into the regular
snapshot quotas.
** Most felt that max snapshot count per snapshottable directory is not needed.
A system level snapshot count however may be necessary as system limit.
# Length of file open-for-writing: most agreed with the proposal to update
length at hearbeats and for applications that care about a specific length to
do a SuperFlush/SuperSync.
# HBase and snapshots: HBase snapshots will be able to use the proprosed
solution.
# Exclude list - most (everyone?) disliked the idea of a "Snapshot exclude
directory list"
# Restores
An interesting messy example was raised: A subtree was renamed to outside of
snapshoted subtree and then that snapshot is restored. What happens to the
subtree that is outside - does it get deleted?
Again this would be discussed further in Jiras
# Should we introduce the notion of volumes? This would incorporate
** a) snapshot able-dir,
** b) where quotas are specified,
** c) not allow renames across volumes
It was noted that c) would break apps.
Allen W and Sanjay were in favor, Most were not in favor of the rename
restriction. This topic to be discussed further in a Jira.
> 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.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
