[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: referrals in use



On 07/30, Pierangelo Masarati wrote:
> Dieter Kluenter wrote:
> >Csillag Tamas <cstamas@digitus.itk.ppke.hu> writes:
> >>On 07/29, Pierangelo Masarati wrote:
> >>>Csillag Tamas wrote:
> >
> >[...]
> >  
> >>Subordinate needs a common prefix for the two databases, do they?
> >>(If I understand correctly.)
> >>    
> >
> >Yes, in principle, but but rwm might do the trick, but I'm not sure.
> >  
> "" is common to any base.

Oh that is really simple.

> >  
> >>That's what I was trying to avoid, with referrals.  I'm ready to go the
> >>way you suggest just want to make sure what and how to do it correctly.
> >>dc=itk,dc=ppke is in production, dc=mkpk,dc=hu is a new suffix, I do not
> >>want to make big changes in dc=itk,dc=ppke, but can move dc=mkpk,dc=hu
> >>to dc=mkpk,dc=ppke.
> >>
> >>So I need to create a dc=ppke root element to create a common root
> >>prefix. Then create dc=mkpk,dc=ppke and set the subordinate flag for
> >>this database. Both must have a same rootdn.
> >>If I start a search against dc=ppke I can search both databases.
> >>Please correct me if I wrong.
> >>    
> >
> >This would be a better task for back-meta and rwm, although I do have
> >my struggle with rwm.
> >  
> back-meta means yet another layer (and possibly another server) in 
> between; slapo-glue (i.e. "subordinate") sounds better for the desired 
> use.  But slapo-glue doesn't work too well with slapo-rwm(5).
> 
> In this case, one could use:
> 
> # ...
> 
> database bdb
> suffix "dc=ppke"
> subordinate
> # ...
> 
> database bdb
> suffix "dc=hu"
> subordinate
> # ...
> 
> database bdb
> suffix ""
> # ...
> 
> or, as an alternative:
> 
> # ...
> 
> database bdb
> suffix "dc=itk,dc=ppke"
> # ...
> 
> database bdb
> suffix "dc=mkpk,dc=hu"
> # ...
> 
> database meta
> suffix ""
> uri "ldap:///dc=itk,dc=ppke";
> uri "ldap:///dc=mkpk,dc=hu";
> # ...
> 
> so that requests for any of the specific naming contexts get satisfied 
> by the appropriate database, while any other request gets routed to both 
> if appropriate (i.e. if the empty base is used).  The second solution is 
> a bit less efficient, but parallelizes lookups; the first approach 
> serializes operations but should be much lighter.

Thanks for the full explanation.
Now it is clear to me.

Thanks to all.
 
-- 
CSILLAG Tamas (cstamas) - http://digitus.itk.ppke.hu/~cstamas