Memory/Resources Errors On System That Has Abundent Resources Running Oracle Linux 6

Memory/Resources Errors On System That Has Abundent Resources Running Oracle Linux 6 (Doc ID 1578925.1)

In this Document
Symptoms
Changes
Cause
Solution
References

APPLIES TO:

Oracle Application Install - Version 12.1.3 to 12.1.3 [Release 12.1]
Linux x86-64
SYMPTOMS

Adpatch recommends the use of a high number of workers, ie 128. The machine is a blade server running 4 processors each with 8 cores and 2 threads per core. The machine has 128 Gb of RAM.

Now trying to apply a patch and when running XDOLoader.class the system is erroring out indicating that it does not have enough resources.

Reducing the number of workers to 8 fixed the issue, but clearly something is wrong since the machine should run 128 workers without problems

Checking a pid error log file from an adworker.log ie /oraapp/r12/oadev1/apps/apps_st/appl/admin/OADEV1/out/hs_err_pid14786.log shows the following

rlimit: STACK 8192k, CORE 0k, NPROC 1024, NOFILE 65535, AS infinity

But the limit.conf contains the values specified by the Installation note Oracle E-Business Suite Installation and Upgrade Notes Release 12 (12.1.1) for Linux x86-64 (Doc ID 761566.1)

 * hard nofile 65535
 * soft nofile 4096
 * hard nproc 16384
 * soft nproc 2047


CHANGES

Migration to RHEL6/OEL6

CAUSE

RHEL6 and OEL6 by inheritance have a flaw in which a user can override limits.conf by placing a file in /etc/security/limits.d.  The PAM Package put an obsolete value in nproc under limits.d which overrode the oracle settings causing the system to not have enough processes available.

In this case the adworkers were taking the value 1024 from limits.d instead of 2047 from /limits.conf, that's the lack of processes reason.



SOLUTION

1. Change the following:

cd /etc/security/limits.d/
ln -s ../limits.conf 99-zzzzzzzzzzzzzz.conf

2. Re-test.