[Date Prev][Date Next]
Re: commit: ldap/servers/slapd/back-bdb back-bdb.h config.c tools.c
- To: Howard Chu <email@example.com>
- Subject: Re: commit: ldap/servers/slapd/back-bdb back-bdb.h config.c tools.c
- From: Hallvard B Furuseth <firstname.lastname@example.org>
- Date: Thu, 16 Sep 2004 20:47:48 +0200
- Cc: OpenLDAP Commit <openldap-commit2devel@OpenLDAP.org>
- In-reply-to: <4104B199.email@example.com>
- References: <200407222032.i6MKWbHg029409@cantor.openldap.org> <firstname.lastname@example.org> <HBF.email@example.com> <4104B199.firstname.lastname@example.org>
Howard Chu writes:
>Hallvard B Furuseth wrote:
>>Howard Chu writes:
>>>>Added linearindex option for slapindex. Improves performance when database
>>>>size exceeds dbcache size by minimizing cache thrashing.
>>>Another note that I forgot to mention - with the linearindex option,
>>>slapadd does no indexing at all. (Otherwise it would defeat the purpose
>>>of this feature.) You must run slapindex after slapadd in this case. I
>>>suppose this should be in the manpage too.
>> Wouldn't it be better to have slapadd run slapindex before it exits?
>> It might not even have to exec() slapindex, since they are the same
>> executable now.
> Maybe. That might require some other restructuring in the
> slapadd/slapindex frontend. Currently the patch is isolated to the
> back-bdb backend, and it could be messy to expose this change in the
> frontend. I haven't really investigated what it would take, and don't
> have the time to look into it at the moment.
Me neither, as you can tell:-)
Still, a slapd.conf-option which changes which commands one must run
like that does make me a bit nervous. Another way might be to give
slapadd a "don't run slapindex" option, and fail if that option is not
specified but the database being updated has linearindex. Then we can
leave "make slapadd run slapindex at the end if linearindex is given"
on the TODO-list.