[
https://issues.apache.org/jira/browse/HBASE-11085?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Demai Ni updated HBASE-11085:
-----------------------------
Attachment: HLogPlayer.java
attached a customized HLogPlayer, which is used in incremental backup for the
'convert' feature. That is convert from HLog to HFile. The code is very similar
to WALPlayer except that it is offline(w/o live hbase cluster). Currently the
code is kind of messy and only for backup code.
So posted here only to show as prototype.
We are working on a general code in
[HBASE-11170|https://issues.apache.org/jira/browse/HBASE-11170]. so the real
patch will be provided it.
> Incremental Backup Restore support
> ----------------------------------
>
> Key: HBASE-11085
> URL: https://issues.apache.org/jira/browse/HBASE-11085
> Project: HBase
> Issue Type: New Feature
> Reporter: Demai Ni
> Assignee: Demai Ni
> Fix For: 0.99.0
>
> Attachments:
> HBASE-11085-trunk-v1-contains-HBASE-10900-trunk-v4.patch,
> HBASE-11085-trunk-v1.patch,
> HBASE-11085-trunk-v2-contain-HBASE-10900-trunk-v4.patch,
> HBASE-11085-trunk-v2.patch, HLogPlayer.java
>
>
> h2. Feature Description
> the jira is part of
> [HBASE-7912|https://issues.apache.org/jira/browse/HBASE-7912], and depend on
> full backup [HBASE-10900| https://issues.apache.org/jira/browse/HBASE-10900].
> for the detail layout and frame work, please reference to [HBASE-10900|
> https://issues.apache.org/jira/browse/HBASE-10900].
> When client issues an incremental backup request, BackupManager will check
> the request and then kicks of a global procedure via HBaseAdmin for all the
> active regionServer to roll log. Each Region server will record their log
> number into zookeeper. Then we determine which log need to be included in
> this incremental backup, and use DistCp to copy them to target location. At
> the same time, a dependency of backup image will be recorded, and later on
> saved in Backup Manifest file.
> Restore is to replay the backuped WAL logs on target HBase instance. The
> replay will occur after full backup.
> As incremental backup image depends on prior full backup image and
> incremental images if exists. Manifest file will be used to store the
> dependency lineage during backup, and used during restore time for PIT
> restore.
> h2. Use case(i.e example)
> {code:title=Incremental Backup Restore example|borderStyle=solid}
> /*******************************************************************************************/
> /* STEP1: FULL backup from sourcecluster to targetcluster
> /* if no table name specified, all tables from source cluster will be
> backuped
> /*******************************************************************************************/
> [sourcecluster]$ hbase backup create full
> hdfs://hostname.targetcluster.org:9000/userid/backupdir t1_dn,t2_dn,t3_dn
> ...
> 14/05/09 13:35:46 INFO backup.BackupManager: Backup request
> backup_1399667695966 has been executed.
> /*******************************************************************************************/
> /* STEP2: In HBase Shell, put a few rows
>
> /*******************************************************************************************/
> hbase(main):002:0> put 't1_dn','row100','cf1:q1','value100_0509_increm1'
> hbase(main):003:0> put 't2_dn','row100','cf1:q1','value100_0509_increm1'
> hbase(main):004:0> put 't3_dn','row100','cf1:q1','value100_0509_increm1'
> /*******************************************************************************************/
> /* STEP3: Take the 1st incremental backup
>
> /*******************************************************************************************/
> [sourcecluster]$ hbase backup create incremental
> hdfs://hostname.targetcluster.org:9000/userid/backupdir
> ...
> 14/05/09 13:37:45 INFO backup.BackupManager: Backup request
> backup_1399667851020 has been executed.
> /*******************************************************************************************/
> /* STEP4: In HBase Shell, put a few more rows.
>
> /* update 'row100', and create new 'row101'
>
> /*******************************************************************************************/
> hbase(main):005:0> put 't3_dn','row100','cf1:q1','value101_0509_increm2'
> hbase(main):006:0> put 't2_dn','row100','cf1:q1','value101_0509_increm2'
> hbase(main):007:0> put 't1_dn','row100','cf1:q1','value101_0509_increm2'
> hbase(main):009:0> put 't1_dn','row101','cf1:q1','value101_0509_increm2'
> hbase(main):010:0> put 't2_dn','row101','cf1:q1','value101_0509_increm2'
> hbase(main):011:0> put 't3_dn','row101','cf1:q1','value101_0509_increm2'
> /*******************************************************************************************/
> /* STEP5: Take the 2nd incremental backup
>
> /*******************************************************************************************/
> [sourcecluster]$ hbase backup create incremental
> hdfs://hostname.targetcluster.org:9000/userid/backupdir
> ...
> 14/05/09 13:39:33 INFO backup.BackupManager: Backup request
> backup_1399667959165 has been executed.
> /*******************************************************************************************/
> /* STEP7: Restore from PIT of the 1st incremental backup
>
> /* specified the backup ID of the 1st incremental
>
> /* option -automatic, will trigger the restore of full backup first, then 1st
>
> /* incremental backup image
>
> /* t1_dn,etc are the original table names. All tables will be restored if not
> specified
> /* t1_dn_restore, etc. are the restored table. if not specified, orginal
> table name will be used
> /*******************************************************************************************/
> [sourcecluster]$ hbase restore -automatic
> hdfs://hostname.targetcluster.org:9000/userid/backupdir backup_1399667851020
> t1_dn,t2_dn,t3_dn t1_dn_restore1,t2_dn_restore1,t3_dn_restore1
> /*******************************************************************************************/
> /* STEP8: Restore from PIT of the 2nd incremental backup
>
> /* specified the backup ID of the 1st incremental
>
> /* option -automatic, will trigger the restore of full backup first, then 1st
>
> /* incremental backup image, and finally 2nd incremental backup image
> /*******************************************************************************************/
> [sourcecluster]$ hbase restore -automatic
> hdfs://hostname.targetcluster.org:9000/userid/backupdir backup_1399667959165
> t1_dn,t2_dn,t3_dn t1_dn_restore2,t2_dn_restore2,t3_dn_restore2
> {code}
> h2. Patch history
> Since this jira depends on [HBASE-10900|
> https://issues.apache.org/jira/browse/HBASE-10900], for each version, two
> patches will be uploaded. One would be the real patch for this
> incremental-update jira; another patch will contain the depended patch, so
> that 1) easy to review; 2) can by applied by HadoopQA
> * Version 1 (https://reviews.apache.org/r/21492/)
> **
> [HBASE-11085-trunk-v1.patch|https://issues.apache.org/jira/secure/attachment/12644214/HBASE-11085-trunk-v1.patch]:
> incremental update/restore code
> **
> [HBASE-11085-trunk-v1-contains-HBASE-10900-trunk-v4.patch|https://issues.apache.org/jira/secure/attachment/12644215/HBASE-11085-trunk-v1-contains-HBASE-10900-trunk-v4.patch]:
> contain both [HBASE-11085-trunk-v1.patch|
> https://issues.apache.org/jira/secure/attachment/12644214/HBASE-11085-trunk-v1.patch]
> and
> [HBASE-10900-trunk-v4.patch|https://issues.apache.org/jira/secure/attachment/12644142/HBASE-10900-trunk-v4.patch]
> * Version 2
> ** [HBASE-11085-trunk-v2.patch |
> https://issues.apache.org/jira/secure/attachment/12646987/HBASE-11085-trunk-v2.patch]:
> incremental. Add UT, address review comments and javadoc warnings
> ** [HBASE-11085-trunk-v2-contain-HBASE-10900-trunk-v4.patch |
> https://issues.apache.org/jira/secure/attachment/12646989/HBASE-11085-trunk-v2-contain-HBASE-10900-trunk-v4.patch]
> : combined patch that can be applied directly to trunk
--
This message was sent by Atlassian JIRA
(v6.2#6252)
