[Mulgara-dev] [Topaz-dev] unable to restore live database from backup

Life is hard, and then you die ronald at innovation.ch
Tue Apr 15 01:36:02 UTC 2008


On Tue, Apr 15, 2008 at 10:56:15AM +1000, Andrae Muys wrote:
> 
> On 15/04/2008, at 2:53 AM, Russell Uman wrote:
> > is there anything i can to do get more information on what's failing?
> >
> > is there anything i can to do attempt to recover the broken backups?
> 
> I'm not convinced the backups are broken, but is it possible for me  
> to look at one of them?  I'm happy to log into a dmz box to take a  
> look if you don't want me poking around your production server or  
> making your data public :).

But he's not using itql backup - he's copying the raw files. Russ,
there's no guarantee that doing an rsync on a running server will give
you something valid, the reason being that rsync does not first take a
snapshot but instead copies files one by one, leaving a window between
files in which things can go weird. If you were to either do an lvm
snapshot or grab the mulgara write-lock first then this posibility
wouldn't be an isuse (I'm not saying there still isn't a problem,
btw).

> > is it possible that this is related to the NPE we get trying to use  
> > the
> > backup ITQL command while the server is running, and there's some kind
> > of corruption in the live datastore?
> 
> Looking at the error message you are seeing I doubt it.  The -1 at  
> the bottom of the error message indicates the server isn't  
> recognising the model URI - I suspect both the backup and the restore  
> are fine, it's just the current model-URI handling code is tripping  
> up.  This has been a rich source of bugs in the past, and I suspect  
> it's biting us again.

Russ, what does

  select $s $p $o from <local:///topazproject#> where $s $p $o;

return? I.e. what models does it show?


  Cheers,

  Ronald




More information about the Mulgara-dev mailing list