Important Operating Note

When using the Unbreakable Enterprise Kernel Release 4 Update 4 (UEK R4u4) with Oracle Linux 6 or Oracle Linux 7, perform a yum update to obtain the latest UEK R4 release updates. Oracle Server X7-8 requires Unbreakable Enterprise Kernel version 4.1.12-94.5.7 or later.

For more information and instructions for updating Unbreakable Enterprise Kernel Release 4, refer to the Unbreakable Enterprise Kernel web site at: http://docs.oracle.com/cd/E93554_01/ .

For OS installation information, see Oracle Linux OS Installation in Oracle Server X7-8 Operating Systems Installation Guide.

Bug ID: 25702796

Issue: When installing Oracle Linux 7.3 on a system with an NVMe storage device (either a storage drive or add-in PCIe card) with larger than 2 TB capacity, the installation might fail due to a Linux kernel panic. This might appear during install as a stack trace on the console. The install process hangs without warning or any other symptom, and the installation does not complete.

Affected Software: Oracle Linux 7.3

Workaround: To avoid this panic during the install, select manual partitioning when configuring the NVMe storage device and ensure that any partition larger than 2 TB capacity uses the ext4 file system instead of XFS. If you perform an automated install using a kickstart configuration file, ensure that your storage partitions on NVMe devices use the ext4 file system, if the partitions are larger than 2 TB.

Bug ID: 25233304, 26235050

Issue: On systems running Oracle Linux when all 16 PCIe cards are installed, kdump might fail.

Affected Software: Oracle Linux 7.3

Workaround: To ensure that kdump works properly, change the crashkernel settings.

In grub.cfg, replace crashkernel=auto with the following settings:

crashkernel=2048M,high crashkernel=256M,low

Change the nr_cpus setting to the following value:

/etc/sysconfig/kdump nr_cpus=4

Bug ID: 27045051, 27044222, 27044007

Issue: On systems configured with Oracle Flash Accelerator F640 PCIe cards, under heavy workloads, the system might display the following kernel warning and retry messages in the dmesg and /var/log/messages files that result from I/O processing delays:

• soft lockups

• nvme and SAS scmd timeouts

• blk_update_get failures

• rcu_sched: stalls

• block device driver error recovery messages, including command and controller aborts and resets, unassociated with underlying media errors

Affected Software: Oracle Linux 7.3

Workaround: Perform the following tasks.

• Increase the workload numa balancing

• Reduce the workload stress level

• Disable the Oracle Linux serial console and global logging of kernel messages:

  1. Remove the serial console boot options (such as console=ttyS0,9600) from /boot/efi/EFI/redhat/grub.cfg.

  2. Add kernel.printk=1 to /etc/sysctl.conf.

  3. Reboot the system.

In rare cases, I/O requests might return EIO or ETIMEDOUT errors. In these cases, retry the I/O by the user application.

Important Operating Note

On systems running Linux with Intel network interface cards or LAN-on-Motherboard (LOM) controllers, if MMIO memory access is set to strict access in the Linux kernel (iomem=strict or as part of the kernel build configuration) you will see the following message in syslog/dmeg when running the Oracle Hardware Management Pack fwupdate tool:

kernel: Program fwupdate tried to access /dev/mem between
c4a00000->c4a01000. (Address may vary)

This message is expected and should not cause an issue with the operation of the operating system. There will be one message each time fwupdate is run and the kernel is in strict MMIO access mode.

However, when the kernel is running in this mode, fwupdate will not be able to access Intel-based network controllers to either list information or update firmware.

For more information on this issue including a workaround, see the "Linux MMIO Access Settings Can Affect fwupdate Commands On Intel Network Controllers" section in the Oracle Hardware Management Pack 2.4 Server CLI Tools User's Guide at: https://www.oracle.com/goto/ohmp/docs .