Validated Configuration Details

The goal of this program is to validate hardware with Oracle Linux and Oracle VM, and to identify issues and configuration details that would impact customer deployments. Oracle developed the program to go far beyond the standard certification testing normally performed with operating systems. Hardware vendors participating in this program perform thorough testing of the hardware in real-world configurations with Oracle Linux and Oracle VM.

This is not a database certification. Oracle Database is a component of the environment because it is demanding on hardware resources (processor, memory, cache, storage, network) and the operating system, and well suited for this advanced level of testing. The validation program uses Oracle Database as a tool and hardware is not certified with Oracle Database as a result of the validation program.

For more information please see the program's Frequently Asked Questions (FAQ).

Configuration Summary

Publication Date

Version

Server Platform

Storage Model

Oracle Software

Linux Distribution

Configuration Summary

OEL 5 x86_64 on IBM System X - x3550 (x336) and IBM System Storage DS8000 Family with Oracle Database 10g Release 2

Publication Date

15-APR-08

Version

Server Platform

IBM System X - x3550 (x336)

Storage Model

IBM System Storage DS8000 Family

Oracle Software

Oracle Database 10g Release 2 (10.2.0.3) for Linux x86_64

Linux Distribution

Oracle Enterprise Linux 5 x86_64

Server Model

4 X IBM System X - x3550 (x336)

Server Details

Additional Server models covered by this configuration:IBM System x - x3400 (x226), x3500 (x236), x3800 (x260),x3650 (x346), x3850 (x366), IBM BladeCenter HS20 and HS21

Processors

2 X Dual Core Intel(R) Xeon(TM) CPU 5160 3.00GHz

Memory

4GB RAM

Network/Interconnect

2 X Broadcom NetXtreme II BCM5708s

HBA

Dual port QLogic QLE2462 Fibre Channel HBA , Emulex PCI-E LPe11002 dual port

Storage Model

IBM System Storage DS8000 Family , DS8300 and DS8300 Turbo , DS8100 and DS8100 Turbo

Switch Model

Brocade 2005-B32

Switch Type

FC Switch

Additional Info.

None

Switch Model

Cisco Catalyst 6500-E

Switch Type

Network Switch

Additional Info.

None

Switch Model

Cisco Catalyst 3500 XL(private.)

Switch Type

Network Switch

Additional Info.

None

OEL 5 x86_64

Kernel

kernel-2.6.18-53.1.14.0.1.el5.x86_64.rpm or higher

File Name

Conf File Settings

Comments

/etc/modprobe.confoptions hangcheck-timer hangcheck_tick=30 hangcheck_margin=180 hangcheck_reboot=1 

/etc/multipath.confdefaults { polling_interval 30 failback immediate no_path_retry 5 rr_min_io 100 path_checker tur user_friendly_names yes }Latest settings are available at http://www-1.ibm.com/support/docview.wss? rs=540&context=ST52G7&dc=D430&uid=ssg1S4000107&loc=en_US&cs=utf-8<=en

device { vendor "IBM" product "2107900" path_grouping_policy group_by_serial }

/etc/security/limits.conforacle  soft    memlock 50000000# set memlock greater than or equal to the sga size to allow oracle to use hugepages if configured

oracle soft core unlimited

oracle hard nproc 131072

oracle hard memlock 50000000

 oracle  soft    nofile  131072# depending on size of db, these may need to be larger

oracle hard nofile 131072

oracle hard core unlimited

oracle soft nproc 131072

/etc/sysconfig/oracleasm

ORACLEASM_SCANORDER="mapper/mpath"

/etc/sysctl.conf

net.ipv4.ip_local_port_range=1024 65000

kernel.shmmax=2074114048

#Set to a value half the size of physical memory

fs.aio-max-nr=3145728

net.core.wmem_max=262144

net.core.wmem_default=262144

 net.core.rmem_max=2097152# rmem_max can be tuned based on workload to balance performance vs lowmem usage

 net.core.rmem_default=262144# rmem_default can be tuned based on workload to balance performance vs lowmem usage

kernel.sysrq=1

kernel.shmall=3279547

kernel.shmmni=4096

kernel.sem=250 32000 100 142

kernel.msgmnb = 65536

kernel.msgmax=8192

kernel.msgmni=2878

fs.file-max=327679

Filesystem

Mount Options

Details

ASMNoneUsing Automatic Storage Management Library Driver (asmlib) for datafiles

RAWNoneocr and voting disk located on raw partitions, for configuration with datafiles on ASM storage.

OCFS2

_netdev,datavolume,nointr

Datafiles, ocr and voting disk files on OCFS2

Driver Module

Hangcheck-timer module

Version

None

Settings

None

Additional Info.

The driver used is the hangcheck-timer module supplied with the distribution

Driver Module

Network driver module (bnx2)

Version

None

Settings

None

Additional Info.

The driver used is the bnx2 driver supplied with the distribution

Driver Module

Qlogic HBA (QLE2462)

Version

8.01.07-k1

Settings

None

Additional Info.

The driver used is the HBA driver supplied with the distribution

Driver Module

Emulex LightPulse Fibre Channel SCSI driver HBA

Version

8.1.10.3

Settings

None

Additional Info.

Firmware 2.50A6 (Z2F2.50A6). The driver used is the HBA driver module supplied with the distribution

Product

Oracle Database 10g Release 2, Single Instance and Oracle Real Application Clusters (RAC) for Linux x86_64

Version

10.2.0.3

Patches

5452672 , 5526245 , 6614205 , 5940792

Bug No

6045759

Summary

Applications using random numbers may hang or report errors if there is insufficient random data in the system

Affects

Oracle applications tools (e.g: java apps) requesting random data

Symptom

Oracle applications tools (e.g: java apps) requesting random data from /dev/random either hang or print out the following information message: Not enough random bytes available. Please do some other work to give the OS a chance to collect more entropy!

Solution/Workaround

If using e1000 or bnx2 network drivers on OEL5 the entropy=1 module option can be set for either of these drivers in /etc/modprobe.conf. Fixed in 2.6.18-8.0.0.4.1.el5 kernel or higher

Bug No

6075960

Summary

If an ORACLE_HOME resides on an nfs partition, a reboot while Oracle is running may cause a panic

Affects

Oracle deployments with Oracle_home on an nfs partition

Symptom

If the ORACLE_HOME resides on an nfs partition and the system is rebooted while an Oracle instance is still running, the system may panic. This does not cause a data corruption, but the system will need to be restarted

Solution/Workaround

Fixed in 2.6.18-53.0.0.0.1.el5 kernel or higher. For other kernel versions lower than this, please shutdown the database instance before reboot of the server.

Bug No

6079461

Summary

OCI program fails to execute with error: LIBNNZ10.SO: cannot restore segment prot after reloc: Permission denied

Affects

OCI programs when selinux is enabled.

Symptom

If selinux is enabled, running an OCI application may fail with the following error: libnnz10.so: cannot restore segment prot after reloc: Permission denied

Solution/Workaround

Disable selinux(either via the selinux=0 boot option or by setting SELINUX=disabled in /etc/sysconfig/selinux)

Bug No

6081490

Summary

With selinux enabled, ocfs2 fails to startup via /etc/init.d/o2cb start

Affects

Automated startup of ocfs2 via /etc/init.d/o2cb start

Symptom

With selinux enabled, starting ocfs2 via "/etc/init.d/o2cb start" fails with the following error: "mount: block device ocfs2_dlmfs is write-protected mounting read-only. mount: cannot mount block device ocfs2_dlmfs read-only. Unable to mount ocfs2_dlmfs filesystem Failed"

Solution/Workaround

For OEL5U3, Manually mount OCFS2_DLMFS using the command mount -t ocfs2_dlmfs ocfs2_dlmfs /dlm. Fixed in OEL5U4.

Bug No

6219364

Summary

Systems using kexec/kdump with Broadcom Tigon3 NICs will not be able to dump vmcore.

Affects

Troubleshooting of systems via kexec/kdump with Broadcom Tigon3 NICs.

Symptom

When troubleshooting with kexec/kdump, the system throws errors when trying to bring up the ethernet interface when trying to connect to the system on to which the vmcore should be dumped. This does not affect normal operations.

Solution/Workaround

Fixed in 2.6.18-53.0.0.0.1.el5 kernel

Bug No

6241862

Summary

Using sysrq+t with netconsole/tg3 configured causes the system to panic.

Affects

Troubleshooting of systems via "alt sysrq t" with netconsole/tg3 configured.

Symptom

When troubleshooting system, usage of "alt + sysrq + t" with netconsole/tg3 configured to take a system stack trace may cause the system to panic . This does not affect the normal operation of the system.

Solution/Workaround

Fixed in 2.6.18-92.0.0.0.1.el5 kernel or higher

Bug No

5240928

Summary

warning messages in trace files during certain workloads with async io turned on

Affects

Performance under high i/o load

Symptom

Following warning in trace files WARNING: ORACLE PROCESS RUNNING OUT OF OS KERNEL I/O RESOURCES

Solution/Workaround

Fixed in 10.2.0.4. For 10g, versions lower than 10.2.0.4 ,apply patch 5452672.

Bug No

5215593

Summary

unable to start RAC with db_cache_size > 4gb

Affects

RAC instances with > 4gb db_cache_size

Symptom

nodes that started up correctly with 10.2.0.1 may fail to startup with ORA-4031 errors in 10.2.0.2

Solution/Workaround

Fixed in 10.2.0.4. For 10g, versions lower than 10.2.0.4, apply patch 5940792; or set _ksmg_granule_size=33554432 in the init.ora. Larger sga sizes may need a higher value.

Bug No

5526245

Summary

warning messages in trace files from qmnc process

Affects

10.2.0.2 and 10.2.0.3 RAC

Symptom

Following warning in trace files - KSXPLOOKUP: warning cnh incarnation number may be stale ...

Solution/Workaround

Fixed in 10.2.0.4. For 10g, versions lower than 10.2.0.4 ,apply patch 5526245.

Bug No

6057703

Summary

ORA-00600:[kclchkblkdma_3] in RAC instance

Affects

10.2.0.3 RAC

Symptom

the user process may terminate leaving the above mentioned traces in alert/trace files

Solution/Workaround

Fixed in 10.2.0.4. For 10g, versions lower than 10.2.0.4 ,apply patch 6614205.

Bug No

6682370

Summary

When configuring the cluster interconnect the interface name should not contain a "."

Affects

10.2.0.3 cluster interconnect configuration

Symptom

If the interface name being used as the private interconnect contains a "." then 1. oifcfg getif does not display the confguration; and 2. The public interface gets used as the cluster interconnect

Solution/Workaround

Fixed in 10.2.0.4. For 10g, versions lower than 10.2.0.4 avoid using interface names with a "."

Bug No

8451866

Summary

Incorrect multipath.conf setting affects RAC node failover. Settings(features = "1 queue_if_no_path") and/or (no_path_retry = queue) causes the node hang if io path fails. no_path_retry has to be set to the numeric value explicitly.

Affects

All applications using device mapper multipath devices which have contingency plan in case of io path failure.

Symptom

RAC node hangs if io path fails.

Solution/Workaround

In configuration file /etc/multipath.conf,set no_path_retry and polling_interval to the numeric values, for example, "no_path_retry=5" and "polling_interval=10" so as to disable queuing. These values depend on system workload.

Kernel

RHEL 5 x86_64

kernel-2.6.18-53.1.14.el5.x86_64.rpm or higher

Packages

oracleasmlib-2.0.3-1.el5.x86_64.rpm

oracleasm-support-2.0.4-1.el5.x86_64.rpm

oracleasm-2.6.18-8.el5-2.0.4-1.el5.x86_64.rpm

ocfs2-2.6.18-8.0.0.4.1.el5-1.2.8-2.el5.x86_64.rpm

ocfs2console-1.2.8-2.el5.x86_64.rpm

ocfs2-2.6.18-8.el5-1.2.8-2.el5.x86_64.rpm