samldb and associated patches

Matthias Dieter Wallnöfer mdw at samba.org
Sat Sep 18 04:09:47 MDT 2010


Hi Kamen,

Kamen Mazdrashki wrote:
> Hi Matthias,
>
> Here are my 2c about you patch :)
>
> http://gitweb.samba.org/samba.git/?p=mdw/samba.git;a=commitdiff;h=7de6c1f019fdd2ab4367c026c56de334afdd09d8
> Right fix for this problem is to make FSMO role transfer handling in 
> "net fsmo"
> tool synchronous. Increasing/decreasing this timeout is just a cosmetic
> workaround, that can byte us again at any time.
I had to apply that since otherwise the test always failed.
>
> http://gitweb.samba.org/samba.git/?p=mdw/samba.git;a=commitdiff;h=fe854c00d6105850f82784ce953d3de36162e438
> If I understood correctly, this commit should allow for object with 
> 'isDeleted' set
> to be removed when they are not in "Deleted Objects" container right?
No. The flag simply prevents the move of deleted objects (tombstones 
ecc.) into "CN=Deleted Objects". Everything else remains the same (so I 
understand it from MS-ADTS 3.1.1.5.5).
> IMHO, such a situation is a bug (or older version of Samba).
> I think we should honor _RELAX_OID control here, so that upgrade provision
> can fix this when upgrading.
> Btw, can you recall how you reproduced this to happen?
Well with my patchset applied it happened.
> I am observing recently a problem with subtree_rename.c module - it 
> doesn't
> allow repl_meta_data.c module to move an object under "Deleted Objects"?
> Unfortunately I can't reproduce this always, so it should be a bug 
> somewhere else.
Exactly this is the issue. Probably not all constraints from 
"subtree_rename" should apply on replicated stuff. Please use the relax 
control if you would like to bypass these constraints.
But exactly here also the mentioned system flag comes into mind: It 
prevents move operations of deleted stuff. And for example all objects 
under cn=sites wear that one, so they souldn't go into cn=deleted objects.

Greets,
Matthias



More information about the samba-technical mailing list