Re: (ITS#7786) Impossible to modify cn=config if olcDbDirectory doesn't exist

[pierangelo.masarati@polimi.it]

On 02/06/2014 11:25 PM, Howard Chu wrote:
> Howard Chu wrote:
>> pierangelo.masarati@polimi.it wrote:
>>> On 01/15/2014 09:59 AM, raphael.ouazana@linagora.com wrote:
>>>> Full_Name: Raphael Ouazana
>>>> Version: 2.4.38
>>>> OS: Linux
>>>> URL: ftp://ftp.openldap.org/incoming/
>>>> Submission from: (NULL) (88.173.78.196)
>>>>
>>>>
>>>> Hi,
>>>>
>>>> I have an old configuration that I would like to export/reimport. The
>>>> olcDbDirectory item of this configuration contains a directory that 
>>>> does not
>>>> longer exist.
>>>> It it then impossible to modify the parameter:
>>>> - I was told not to edit directly LDIF config files
>>>> - if i try a slapcat -n0 I get:
>>>> 52d64c91 olcDbDirectory: value #0: invalid path: No such file or 
>>>> directory
>>>> 52d64c91 config error processing olcDatabase={2}hdb,cn=config: 
>>>> olcDbDirectory:
>>>> value #0: invalid path: No such file or directory
>>>> slapcat: bad configuration directory!
>>>>
>>>> I think slapcat should always allow to export a configuration.
>>>
>>> I see the point; slapcat is failing because to export the configuration
>>> (c->op == SLAP_CONFIG_EMIT) it needs to read the configuration first,
>>> and it reads the whole config tree.
>>>
>>> Perhaps when slapcat of only the config database is requested, config
>>> parsing should skip other databases, or at least ignore errors, if 
>>> possible.
>>
>> Agreed, we definitely need this.
>
> But it's not clear that it's entirely feasible without major 
> structural changes to back-config. In particular, later config items 
> may depend on earlier ones succeeding (e.g., loading a module or 
> reading a schema definition). So we can't simply no-op everything 
> while running slapcat, nor can we safely ignore all errors, because we 
> must still be able to parse all of the underlying config LDIF and some 
> of it will be unparseable without appropriate schema being loaded.
In the specific case, the fix is trivial: the user must create the 
folder.  In principle, when a "ignore non-blocking errors" flag is set, 
parsing of olcDbDirectory should not cause a failure.  Such flag would 
only be set for, say, database entries that are not strictly required 
for that slapcat (for any tool operation that does not need such 
database, for example).  However, in this specific case, the user could 
simply find the directory name by manually inspecting the offending LDIF 
file...

p.

-- 
Pierangelo Masarati
Associate Professor
Dipartimento di Scienze e Tecnologie Aerospaziali
Politecnico di Milano