[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
Re: Possible bug?
On Wed, 12 Jul 2000, William Yodlowsky wrote:
> I'm rebuilding with gdbm so I should be able to see if that helps soon.
Well, I rebuilt like so:
env \
ac_cv_func_pthread_create=no ol_cv_kthread_flag=no \
ol_cv_pthread_flag=no ol_cv_pthreads_flag=no \
ol_cv_thread_flag=no \
./configure --with-ldbm-api=gdbm --enable-wrappers
To make threads, gdbm, and tcpwrappers work (from the FAQ-O-Matic).
I still see
yield() = 0
time() = 963431301
yield() = 0
time() = 963431301
by the thousands in truss outputs, however the load has gone WAY down to
50% CPU (sustained). There are always 6 threads running, and top(1)
always shows slapd in a "run" state.
(When using BerkeleyDB it was in a "sleep" state, with 6 threads, and 95%
CPU in the same conditions.)
Actually I spoke too soon. As I'm typing this, I have top(1) watching the
machine in question and truss(1) in another window. slapd has just hit
90% and is still climbing and truss shows:
Occasionally I will see:
[thousands of time() calls in a row snipped]
time() = 963431449
time() = 963431449
time() = 963431449
time() = 963431449
time() = 963431449
lwp_sema_post(0xFDF05E78) = 0
lwp_sema_wait(0xFDF05E78) = 0
time() = 963431449
time() = 963431449
lwp_sema_post(0xFE40FE78) = 0
lwp_sema_wait(0xFE40FE78) = 0
time() = 963431449
time() = 963431449
time() = 963431449
time() = 963431449
time() = 963431449
[thousands more time() calls in a row snipped]
So I guess it wasn't db2 specific.
What can I do to help debug this?
Thanks for all your help.
`-
William Yodlowsky
<wyodlows@nj.devry.edu>