Re: Possible bug?

["Kurt D. Zeilenga" <Kurt@OpenLDAP.org>]

I may have misdiagnosed the problem.  It might be that daemon.c
is mishaving.  However, it shouldn't be in a tight loop only
which only calls time().  Other calls, namely select(), should
be intermixed.

At 01:28 PM 7/12/00 -0400, William Yodlowsky wrote:
>On Wed, 12 Jul 2000, Kurt D. Zeilenga wrote:
>
>> This is likely just BerkeleyDB spinlocking.  It's generally viewed
>> as feature, not a bug.  The fact that it causes load is inherit.
>> 
>> If you don't like BerkeleyDB spinlocking, you can may be able
>> to disable it (see their configuration documentation) or use
>> an alternative DBM (such as GDBM).
>
>Well, I can attest to it not being just 1.2.11 that does it (1.2.10 with
>db2 does it too).  Didn't think about the backend... I'll try that.
>
>BTW disabling threads dropped the yield() in the 
>truss(1) traces.  However, I still get thousands of time() calls in a
>row and the load is still pretty high (90%+ on an idle system?)...
>
>> >time()                                          = 963368437
>> >time()                                          = 963368437
>> >time()                                          = 963368437
>> >time()                                          = 963368437
>> >time()                                          = 963368437
>> >time()                                          = 963368437
>
>`-
>William Yodlowsky
><wyodlows@nj.devry.edu>