[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bacula-devel] Trying out the VirtualFull feature.


On Thursday 28 August 2008 13:41:57 Graham Keeling wrote:
> On Wed, Aug 27, 2008 at 07:28:08PM +0200, Kern Sibbald wrote:
> > On Tuesday 26 August 2008 12:40:29 Graham Keeling wrote:
> > > I wonder - if you try the virtual-backup-test (with my patch from my
> > > last email) on your setup, does the test pass or succeed?
> >
> > No, I didn't try it because the patch guarantees that the test will fail.
> >  If you backup a file in the build directory then delete it before the
> > restore, you are *guaranteed* that the test will fail.  The build
> > directory is sort of the base, and the results of the restore are
> > compared to it, so when you delete something from the build directory,
> > you can be sure the test won't work.
>
> OK, sorry, I misunderstood what 'check_restore_diff' was doing. I thought
> that it was comparing the two restores, whereas it is actually comparing
> the restore after the VirtualFull with the original directory.
>
> So, I moved my 'rm' command to immediately before 'end_test' at the end of
> the script and tried again.
> I also made 'check_restore_diff' in scripts/functions print the output of
> the diff.

I turned on some debug code, which I have been wanting to do for some time, 
and there was indeed a problem that crept in after the first implementation.  
It is now picking up all the jobs it should and I have verified that the 
consolidation bsr is identical to the first restore bsr (the consolidation 
uses totally new code implemented for accurate).

However, there is still another problem, probably a confusion in the SD with 
the Virtual Full level that causes the consolidation to write out all the 
data but there is something messed up with the JobMedia records, which means 
a restore after the consolidation doesn't yet work.   Hopefully it won't be 
long before I get that resolved.  In the mean time, I have committed the 
changes so far, which give sufficient debug output to see that it is actually 
processing the three jobids rather than just the Full.

I like your idea of having a file that actually changes (your date stamp).  I 
haven't yet applied it to the Virtual backup test, but will do so, and I 
think I'll add it to a number of other "incremental" tests.

Best regards,

Kern

>
> Here is what I get...
>
> > invader regress # ./tests/virtual-backup-test
> >
> >
> >  === Starting virtual-backup-test at 11:50:12 ===
> >
> > diff -r build/date
> > tmp/bacula-restores/home/graham/bacula/regress/build/date 1c1
> > < Thu Aug 28 11:50:41 BST 2008
> > ---
> >
> > > Thu Aug 28 11:50:15 BST 2008
> >
> >   !!!!! virtual-backup-test failed!!! 11:51:31 !!!!!
> >   !!!!! Restored files differ          !!!!!
> >    Status: backup=0 restore=0 diff=1
>
> I then commented out the second restore in the test script and ran it
> again, so that the diff was being run on the original directory and the
> restore
>
> before the Virtual Backup:
> > invader regress # ./tests/virtual-backup-test
> >
> >
> >  === Starting virtual-backup-test at 12:18:50 ===
> >
> >   ===== virtual-backup-test OK 12:19:57 =====
>
> On Wed, Aug 27, 2008 at 07:28:08PM +0200, Kern Sibbald wrote:
> > It could potentially work if Accurate is turned on, but the
> > virtual-backup-test script does not have Accurate turned on,
>
> I tried the above test with and without 'Accurate = yes' added to all of
> the 'Job' sections in scripts/bacula-dir-migration.conf. The result was the
> same in both cases.
>
> > but the current test is only meant to test Virtual Backup ...
>
> Unless I am failing to understand what Virtual Backup is meant to do,
> my adjustments to the current test shows that it doesn't work.
> I would expect my 'date' file in the original directory, and after each of
> the two restores, to all be identical. I would expect this with either
> Accurate turned on or off, as the 'date' file is present at the beginning
> of the test, before the first Full backup.
> They are not identical.
>
> The current test (without my adjustments) is 'touch'ing files in the build
> directory between backups, which only adjusts the timestamps.
> The command that you are using to then check that the restores are the same
> is 'diff -r', which doesn't compare the timestamps. It compares whether
> the contents of the files actually are different or not.
> So, I would suggest that the current test is broken (or at the very least,
> not testing what I want it to test), as it cannot tell whether the Virtual
> Backup has included the Differentials and Incrementals.
>
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's
> challenge Build the coolest Linux based applications with Moblin SDK & win
> great prizes Grand prize is a trip for two to an Open Source event anywhere
> in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> Bacula-devel mailing list
> Bacula-devel@xxxxxxxxxxxxxxxxxxxxx
> https://lists.sourceforge.net/lists/listinfo/bacula-devel



-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Bacula-devel mailing list
Bacula-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/bacula-devel


This mailing list archive is a service of Copilotco.