[Date Prev][Date Next] [Chronological] [Thread] [Top]

CPU usage and #ITS2383 (sched_yield())



Hi all,

We are currently migrating to openldap 2.1.x from openldap 2.0.27.
Our architecture is composed of one master and two slaves.
Everything seems to go fine until the master starts replication, at this
moment both slaves start to use 100% cpu.

We have include all the configuration advices (threads, checkpoint,
DB_CONFIG, logging, replogs ...) but we haven't achieve success.

Doing a "strace -p pid" to one of the slapd process we can see thousands
of lines like :

sched_yield()                           = 0

Any idea or advice ?

Thanks in advance,
Iñaki

* Software  :

RedHat 7.3 + kernel 2.4.21
Openldap-2.1.22
Berkeley DB 4.1.25
Openssl 0.9.7b

* Hardware

Master : 2x2.4 Xeon + 2 Gb RAM
Slaves : 2x1.2666 Mhz PIII + 1 Gb RAM

* Openldap configure :

env CPPFLAGS="-I/usr/local/etc2/openssl/include -I/usr/local/etc2/db-4.1.25/include" \
LDFLAGS="-L/usr/local/etc2/openssl/lib -L/usr/local/etc2/db-4.1.25/lib" \
./configure --with-tls --with-threads --enable-crypt --enable-monitor --prefix=/usr/local/etc2/openldap_2_1




Attachment: smime.p7s
Description: S/MIME Cryptographic Signature