[
https://issues.apache.org/jira/browse/PHOENIX-2948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15544486#comment-15544486
]
Sergey Soldatov commented on PHOENIX-2948:
------------------------------------------
I suggest to improve this approach to not only taking snapshot, but make a
backup table feature. There are two approaches:
1. Create snapshots for all required tables and run a separate MR job for
storing data from snapshots to file/files on hdfs. Metadata for table need to
be stored separately.
2. Use new HBase backup feature (hopefully it will be accepted in the nearest
releases of HBase). But in this case it will work only with the latest HBase.
First way is quite similar to HBase backup and implementing it may be painful
(for HBase team it's already took about a year and still in progress). We may
try to implement it in easier way, but not sure whether it's possible or not.
Also there is a questions how to use it. Whether it should be a new command for
sqlline or external tool or both.
All thoughts are welcome, I will try to collect all of them and make a summary
document.
> Snapshot support for Phoenix Tables
> -----------------------------------
>
> Key: PHOENIX-2948
> URL: https://issues.apache.org/jira/browse/PHOENIX-2948
> Project: Phoenix
> Issue Type: New Feature
> Affects Versions: 4.7.0
> Reporter: Anil Gupta
> Assignee: Sergey Soldatov
> Labels: Backup/Restore
>
> As discussed in PhoenixCon, here is the description of desired Snapshot
> feature in Phoenix:
> Currently, Phoenix does not supports taking snapshots of Table.
> HBase has support for taking snapshot of table so Phoenix should also support
> taking snapshot for taking backups of data.
> At present, taking backup of a Phoenix is highly manual and detail oriented
> process and in some cases its not even possible to take a snapshot because
> Metadata of all Phoenix tables is stored in one system table. We cant take
> snapshot of entire system table for making a backup/restore of one particular
> table.
> These are scenarios that we should handle in Phoenix Snapshot:
> 1. Handling Global and Local secondary index.
> 2. Snapshotting Table schema.
> 3. Handling views. I am not sure how. But just highlighting it.
> I think #2 above requires most of the work. IMO, #1 should be easy to handle.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
