[
https://issues.apache.org/jira/browse/HBASE-15862?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15330125#comment-15330125
]
Vladimir Rodionov edited comment on HBASE-15862 at 6/14/16 6:40 PM:
--------------------------------------------------------------------
Ok, let us suppose that last full backup was done on a table when it had 10
regions. The last incremental backup was done when table had 100 regions. We
restore everything into table with 10 regions. When restore finishes, every
flush will trigger split until we reach ~ 100 regions - super split storm.
The current restore code does not use direct wal replay to HBase, [~jerryhe].
We combine all WALs into a single M/R job, create HFiles and then load them
into a table. This is why is so important to have region boundaries after full
backup restore as much closer to a PIT boundaries as possible - to avoid
subsequent region splits.
was (Author: vrodionov):
Ok, let us suppose that last full backup was done on a table when it had 10
regions. The last incremental backup was done when table had 100 regions. We
restore everything into table with 10 regions. When restore finishes, every
flush will trigger split until we reach ~ 100 regions - super split storm.
> Backup - Delete- Restore does not restore deleted data
> ------------------------------------------------------
>
> Key: HBASE-15862
> URL: https://issues.apache.org/jira/browse/HBASE-15862
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.0.0
> Reporter: Vladimir Rodionov
> Assignee: Vladimir Rodionov
> Labels: backup
> Fix For: 2.0.0
>
> Attachments: HBASE-15862-v1.patch, HBASE-15862-v2.patch,
> HBASE-15862-v3.patch
>
>
> This was discovered during testing. If we delete row after full backup and
> perform immediately restore, the deleted row still remains deleted.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
