Additional tdb flag

[Shlomi Yaakobovich]

I believe we've covered the cycles/loops issue in tdb_traverse, around a
year ago. Since that fix we haven't seen any loops or tdbbackup
processes stuck in 100% CPU. Have you this behavior since the last fix
(around May last year) ?

I think this is not a full bulletproof test of the tdb file, but it will
at least give a clear indication if the tdbbackup operation that will be
followed. We find it useful to know if the tdbbackup -v will restore the
tdb file or not, before actually doing it. And for this purpose, this
fix serves its purpose.

Shlomi
 

> -----Original Message-----
> From: 
> samba-technical-bounces+shlomi=exanet.com at lists.samba.org 
> [mailto:samba-technical-bounces+shlomi=exanet.com at lists.samba.
> org]On Behalf Of Volker Lendecke
> Sent: Friday, May 12, 2006 19:18
> To: Meital Hershcovitch
> Cc: samba-technical at lists.samba.org
> Subject: Re: Additional tdb flag
> 
> 
> On Tue, May 09, 2006 at 03:35:09PM +0300, Meital Hershcovitch wrote:
> > To handle this we added another flag to the tdbbackup "-c". 
> This flag
> > will only check if the tdb file was corrupted and return 
> non zero value
> > if it was corrupted.
> 
> Hmmm. I don't think running tdb_traverse is the best way to
> figure if a tdb is corrupted or not. I've for example seen
> tdb's with cycles in the freelist, ending up in 100% CPU for
> everyone trying to add something to it. This is not covered.
> 
> I'd hesitate to add this to the main tree, as I think it's
> really unreliable, and I don't want to give the illusion we
> really check something.
> 
> However, having a decent consistency checker for tdb's would
> be great, but this will probably involve a lot more work.
> 
> Volker
>