OHS 11g Poor Performance and Coredumps Frequently When LD_ASSUME_KERNEL is set (Doc ID 1324155.1)

Last updated on FEBRUARY 16, 2017

Applies to:

Oracle HTTP Server - Version 11.1.1.0.0 to 11.1.1.5.0 [Release Oracle11g]
Linux x86
Linux x86-64

Symptoms

OHS 11gR1 Performance is poor, it produces core-dump files frequently due to persistent crashes.
The followings are also might be noticed:

#0 0x0000002a95b9cf25 in __pthread_alt_unlock () from /lib64/libpthread.so.0 (gdb) bt
#0 0x0000002a95b9cf25 in __pthread_alt_unlock () from /lib64/libpthread.so.0
#1 0x0000002a95b99ead in pthread_mutex_unlock () from /lib64/libpthread.so.0
#2 0x0000002a95d54a44 in yodysMtxUnlock ()
from $ORACLE_HOME/lib/libdms2.so
#3 0x0000000000458daf in dms_thread_mmap ()
#4 0x0000002a95b98b6b in pthread_start_thread () from /lib64/libpthread.so.0
#5 0x0000002a985f9e13 in clone () from /lib64/libc.so.6
#6 0x726f746365726944 in ?? ()

 

DMS pid 15501 detected hung shared mutex (9c3e4260) -- shutting down.
MODPLSQL-00806: mod_plsql: Sensor create failed
mod_plsql: memory allocation failure - aborting
child process 25170 still did not exit, sending a SIGTERM
child process 15659 still did not exit, sending a SIGTERM
dms_thread_stop: thread wait timed out
dms_thread_stop: thread wait timed out

 

pthread_mutexattr_setpshared: Transport endpoint is not connected
pthread_mutexattr_setpshared: Invalid argument
pthread_mutexattr_setpshared: No such file or directory


At the time of OHS start - when no requests are made to OHS - the number of httpd.worker process is much more than expected, ($ ps -ef| grep httpd.worker) shows:

oracle 8248 8226 0 15:44 ? 00:00:00 $ORACLE_HOME/ohs/bin/httpd.worker -DSSL
oracle 8843 8248 0 15:44 ? 00:00:00 $ORACLE_HOME/ohs/bin/httpd.worker -DSSL
oracle 8844 8843 0 15:44 ? 00:00:00 $ORACLE_HOME/ohs/bin/httpd.worker -DSSL
oracle 9012 8984 0 15:44 ? 00:00:00 $ORACLE_HOME/ohs/bin/httpd.worker -DSSL
oracle 9013 8984 0 15:44 ? 00:00:00 $ORACLE_HOME/ohs/bin/httpd.worker -DSSL
oracle 9014 8984 0 15:44 ? 00:00:00 $ORACLE_HOME/ohs/bin/httpd.worker -DSSL
[..]


Total number of httpd.worker processes was 38 in this case while the expected number was just 2 processes, as defined by StartServers Directive in the MPM worker configuration in httpd.conf:

<IfModule mpm_worker_module>
StartServers 2
MaxClients 150
MinSpareThreads 15
MaxSpareThreads 25
ThreadsPerChild 25
MaxRequestsPerChild 0
AcceptMutex fcntl
LockFile "$INSTANCE_HOME/diagnostics/logs/OHS/ohs1/http_lock"
</IfModule>

Cause

Sign In with your My Oracle Support account

Don't have a My Oracle Support account? Click to get started

My Oracle Support provides customers with access to over a
Million Knowledge Articles and hundreds of Community platforms