directory-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Emmanuel Lécharny <>
Subject Re: Suggestions for troubleshooting filters.
Date Thu, 06 Mar 2014 06:49:06 GMT
Le 3/6/14 6:54 AM, Mark D a écrit :
> Well, you may have helped uncover something.  I thought I had indexes
> When I look at config in ApacheStudio it shows the indexes on the partitions.
> When I looked at config file, I noted the following index attributes were missing.
> ads-enabled: TRUE
> ads-indexcachesize: 100
> Note, a few partitions I was testing were created dynamically and not through
> the Studio.  Hence I was responsible for those missing.

Definitively, the ads-enabled attribute *must* be TRUE (FALSE is teh
default AFAIR). The other one has a default valu of 100.

> So, I added ads-enabled: TRUE to the UID index for the partition.
> Restarted AD.  Ran test search which failed with same results.
> I then added a new entry to the partition and ran search again.
> It now finds the entry with UID.
> So, it this possibly the issue ? 
Yes, absolutely.

>   When I search with filter (objectClass=namedObject)
> it always worked.  Does 'x=*' have to be indexed, whereas '=' searches anything ?
Those are two different indexes. the (xxx=*) will match every entry
having the xxx attributeType, and we use the prsenceIndex for that, when
the (xxx=blah) use the xxx index to fetch entries having an
attributeType xxx having the value blah.

Note that the (xxx=*) filter will only work *if* xxx is indexed...

One vicious thing is that if uid is not indexed, filters like (uid=blah)
will work, but (uid=*) won't. The reason is that when we have and
indexed attribute like uid, we will use the index to fetch entries , so
here, as we always have a presence index, we will use it, and as uid is
not indexed, we won't find any entry associated with (uid=*).

Now, thinking out loud I think that we are doing a very big disservice
here : just because we do'nt have an index on uid should *never* be an
issue for the user, except that the response time will be way lower. In
other words, indexes should speedup the serahces, not break them.

That deserves a JIRA, IMO.
> I have been suffering for a while with so many filters not working with (x=*).
I can understand...
> My confusion has been based on
> 1) I was testing with one partition which evidently was a partition created from the
> and I was comparing behavior to the new dynamic created partition.  
> 2) In ADStudio, the configs visually appear the same regarding indexes.
> I am back to recreating all test partitions etc and see if this changes things.
I'm quite sure it will work fine if you do that.

In the mean time, I'll create an issue to get the server respect our
users and *always* do the search and return valid entries except when
the index is created *after* the entry has already been created (we
do'nt reindex on the fly, sadly. Note that we should and we will, but
not in this current version).

> Thanks.
Thanks to you for having raised this issue !

Emmanuel Lécharny 

View raw message