[ZODB-Dev] Who's experiencing FileStorage corruption?

[Christian Robottom Reis]

On Thu, Jul 01, 2004 at 03:24:35PM -0400, Tim Peters wrote:
> > In any case, during all the time we at the MEMS Exchange used ZODB we
> > never experienced data loss or corruption due to ZODB or ZOE.  One time we
> > noticed that bad memory in one of our servers caused the FileStorage to
> > become corrupted.  Fortunately, in large part due to the simplicity of
> > the FileStorage, I was able to reconstruct the database using the current
> > file and backups without losing data.
> 
> That's a match to Zope Corp's experience, and thanks for confirming that
> FileStorage users aren't generally plagued with corruption problems!  Since
> (as you know) a FileStorage always grows by appending, it's hard to imagine
> how clumps of NUL bytes can end up in the interior except via non-ZODB
> system problems (although I remain open to the possibility that ZODB is at
> fault).

We've had a fairly large (~1GB) database being used in production for
about a year now. It has about 15 simultaneous users doing OLTP-style
work every weekday. We don't use Zope, but rather IndexedCatalog and a
PyGTK UI.

We've never had a single instance of any sort of corruption. We don't
pack very often (once every month or so), and nothing much happens on
the server apart from running the ZEO and a few instances of the
application for people with slow machines. 

We do keep daily backups, of course.
--
Christian Robottom Reis | http://async.com.br/~kiko/ | [+55 16] 3361 2331