DarkMatter in Cyberspace

Subversion Backup and Restore

Backup Specified Revision

  1. Backup specified revision (here is 20):

    $ cd /opt/svnRepo
$ svnadmin dump deployTest/ -r 20 > deploy-r20.dump

  2. Restore backup:

    $ mv deploy-r20.dump /opt/Gcp/tmp
$ cd /opt/Gcp/tmp
$ svnadmin create deploy20
$ svnadmin load deploy20 < deploy-r20.dump

Verify

  1. Get revision 20 of "deploy.exp" in original repository

    $ svn cat -r 20 file:///opt/svnRepo/deployTest/CalcServer/deploy.exp > r20

  2. Get "deploy.exp" in restored repository

    $ svn cat file:///opt/Gcp/tmp/deploy20/CalcServer/deploy.exp > restore20

  3. Compare them:

    $ diff r20 restore20

    They are identical.

  4. Compare logs:

    [root@cloud141 /opt/svnRepo]$ svn log -r 20 file:///opt/svnRepo/deployTest/ 
------------------------------------------------------------------------ 
r20 | bvt | 2013-08-30 17:30:03 +0800 (Fri, 30 Aug 2013) | 1 line 
remove unittest in source file 
-------------------------------------------------------- 
[root@cloud141 /opt/svnRepo]$ svn log file:///opt/Gcp/tmp/deploy20 
------------------------------------------------------------------------ 
r1 | bvt | 2013-08-30 17:30:03 +0800 (Fri, 30 Aug 2013) | 1 line 
remove unittest in source file 
------------------------------------------------------------------------

So you can see the restored repository is identical to that part of original repository except the revision number.

Backup Specified Revisions

    $ svnadmin dump deployTest/ -r 10:20 > deploy-r20.dump

Incremental Backup

  1. Backup:

    $svnadmin dump myrepos -r 0:1000 > 0-1000.dumpfile 
$svnadmin dump myrepos -r 1001:2000 --incremental > 1001-2000.dumpfile 
$svnadmin dump myrepos -r 2001:3000 --incremental > 2001:3000.dumpfile

  2. Restore:

    $svnadmin load myrepos < 0-1000.dumpfile 
$svnadmin load myrepos < 1001-2000.dumpfile 
$svnadmin load myrepos < 2001:3000.dumpfile

Experiment on incremental dump

  1. Create a incremental backup: 

    [root@cloud141 /opt/svnRepo]$ svnadmin dump deployTest/ --incremental -r 9:11 > r9-11.dump

  2. Create a full backup: 

    [root@cloud141 /opt/svnRepo]$ svnadmin dump deployTest/ -r 9:11 > backup9-11.dump

  3. Compare the size of these two dump files, full backup is obviously larger than the incremental counterpart: 

    [root@cloud141 /opt/svnRepo]$ ll 
total 984 
-rw-r--r-- 1 root root 997105 Oct 31 19:08 backup9-11.dump 
drwxr-xr-x 6 root root 4096 Aug 22 18:09 deployTest 
-rw-r--r-- 1 root root 2509 Oct 31 19:07 r9-11.dump

  4. Restore a incremental backup in a fresh repository: 

    [root@cloud141 /opt/svnRepo]$ mv r9-11.dump ../Gcp/tmp 
[root@cloud141 /opt/Gcp/tmp]$ svnadmin create new9 
[root@cloud141 /opt/Gcp/tmp]$ svnadmin load new9 < r9-11.dump 
<<< Started new transaction, based on original revision 9 
svnadmin: File not found: transaction '0-0', path 'CalcServer/src/com/boco/deploy/ConfigLoader.java' 
  * editing path : CalcServer/src/com/boco/deploy/ConfigLoader.java ...[root@

    You can see the restore failed.

  5. Restore a incremental backup (revision 9~11) in a "root" repository (restored from revision 8) : 

    [root@cloud141 /opt/Gcp/tmp]$ svnadmin load backup8 < r9-11.dump 
... 
[root@cloud141 /opt/Gcp/tmp]$ svn log file:///opt/Gcp/tmp/backup8 
------------------------------------------------------------------------ 
r4 | bvt | 2013-08-30 16:35:10 +0800 (Fri, 30 Aug 2013) | 1 line 
initial import project 
------------------------------------------------------------------------ 
r3 | bvt | 2013-08-30 15:31:47 +0800 (Fri, 30 Aug 2013) | 1 line 
make new project basedir 
------------------------------------------------------------------------ 
r2 | bvt | 2013-08-29 14:32:15 +0800 (Thu, 29 Aug 2013) | 1 line 
remove unittest in source file to pass compile without junit 
------------------------------------------------------------------------ 
r1 | bvt | 2013-08-29 14:31:10 +0800 (Thu, 29 Aug 2013) | 1 line 
remove unittest in source file to pass compile without junit 
------------------------------------------------------------------------

    Now the restore succeed.

Summary

Backup a huge repository in the following steps:

  1. Execute a full backup at a revision. If the revision n is not 0, you will lose all revisions from 0 to n-1;

  2. Execute incremental backups weekly, use the revision arrange as file name: "inc-rev-from-to.dump", like "inc-rev-35-46.dump";

  3. When original repository corrupts, first load the full backup, then load the incremental ones according to their revision numbers;

Ref:

Published

Last Updated

Category

Tech

Tags

Contact