[Date Prev][Date Next]
Re: referral and performance
Andrea Albert <firstname.lastname@example.org> writes:
> well, the point is, that the clients asking for authentication belong
> to a different subtree as the referral. We have 2 ldap main servers
> running for testing in different subnets and both of them react like
> i've described below.
> Can it be a Berkeley DB problem? I have no DB_CONFIG to tune the
> database! Is it mandatory to ensure a normal functionality ?
> If the referral is up again, than the same clients, located in a
> different subtree as the referral, obtain a login permition whithin
> round about 5 secs.
> Norbert Klasen wrote:
>> --On Montag, 23. Juni 2003 12:41 +0200 Andrea Albert
>> <email@example.com> wrote:
>>> It works very fine!
>>> But if the server to which this referral points go down, then the
>>> response time of the main LDAP-Server (containing this referral) is >20
>>> sec. Many clients close connection before getting any answer for
>>> authentication. How can I avoid this problem?
>>> Pls help!!
>> This seems strange as the main server will not follow the referral
>> itself. It's the clients job to chase the referral. I suggest you
>> trace the connection between your client and server with something
>> like Ethereal to get an idea where exactly the delay occurs.
That could be a BDB problem. Just test your cachesize
./db_stat -h /path/to/openldap-data/ -m
If percentage for 'Requested pages found in cache' is below 80% you
should increase your cachesize.
Dieter Kluenter | Systemberatung
Tel:040.64861967 | Fax: 040.64891521