2.1 Changes and Improvements in Release 2.3.3

This section describes functional changes, improvements and bug fixes compared to the previous release.

Support for Next-Generation Hardware

Oracle PCA Release 2.3.3 software adds provisioning and orchestration support for the Oracle ZFS Storage Appliance ZS5-ES. It is a standard component in new Release 2.3.3 systems, offering the same functionality and configuration as its predecessor. The internal components are more up-to-date and offer an increase in performance and disk space.

Server Operating System and Kernel Upgrade

The Oracle PCA Release 2.3.3 software includes the latest supported operating system and kernel for all management and compute nodes: Oracle Linux 6 Update 9 with UEK Release 4 Update 6. These contain important bug fixes to improve stability and compatibility, and eliminate a number of vulnerabilities.

Security Enhancements for Appliance and Virtual Machines

Oracle has developed patches related to the Spectre and Meltdown vulnerabilities ( CVE-2017-5715, CVE-2017-5753, and CVE-2017-5754). Please refer to the Addendum to the January 2018 Critical Patch Update Advisory for Spectre and Meltdown MOS note (Doc ID 2347948.1).

Additional information specific to Oracle PCA is also available on My Oracle Support. For details, refer to the support article with Doc ID 2432160.1.

For improved virtual machine security, Oracle PCA now includes a version of the Xen hypervisor that allows you to disable paravirtualized virtual machines, or PV guests, through a configuration file setting. For instructions, refer to the section Disabling Paravirtualized Virtual Machines to Avoid Security Issues in the Oracle Private Cloud Appliance Administrator's Guide.

Bugs Fixed in Release 2.3.3

The following table lists bugs that have been fixed in Oracle PCA Release 2.3.3.

Table 2.1 List of Fixed Bugs

Bug ID



Compute Nodes Reboot and VMs Crash Due to Low Free Memory Setting

During management node and compute node upgrade, the vm.min_free_kbytes kernel parameter is set to a higher value than the Oracle VM default. This reduces the risk of virtual machine crashes and eliminates a number of random compute node reboots.


Software Update Fails Due to Running Oracle VM Jobs

The mechanism to detect and remove running and stale jobs prior to upgrade has been enhanced. As a result, the upgrade process is impacted significantly less by existing Oracle VM jobs.


Manual Pre-Upgrade Script Required to Detect Database Password Issues

The functionality of the script named pca_precheck_mysql.sh, which was previously provided on My Oracle Support alongside the Oracle PCA software image, has now been incorporated into the automated pre-upgrade validation.


Virtual Machines Take Long to Restart After Oracle VM 3.4.x Upgrade

The algorithm to select the best possible compute node when a virtual machine is started, has been optimized for better performance. VMs now restart significantly faster.


Reprovisioning Fails, Compute Node Marked Dead

When tenant group settings are applied across all member compute nodes, the code checks for events indicating that a server has gone offline, and attempts to remove it. However, such events do not always imply that a server is unavailable. The detection mechanism has now been refined, so that previously acknowledged events can no longer cause reprovisioning issues for compute nodes belonging to a custom tenant group.


Post-Upgrade Tasks Do Not Complete If a Dead Compute Node Is Present

Previously, it was possible to execute a controller software update on a system containing a compute node marked dead in Oracle VM. However, this can lead to internal timeouts in the compute node upgrade process, and prevent the controller software update from finishing all post-upgrade tasks. As a result, the system can only be brought back online by manually cleaning up the errors.

From Release 2.3.3 onward, the pre-upgrade validation script returns a fatal error when dead compute nodes are discovered, and upgrading is no longer allowed. Dead compute nodes fall under the scenario described in the section Upgrading to Oracle Private Cloud Appliance Release 2.3.x with Non-Functional Compute Nodes in the Oracle Private Cloud Appliance Administrator's Guide.


Regression in Xen Reduces Performance

Due to a regression in Xen, the vCPU allocation applied within Oracle PCA was no longer effective, resulting in decreased VM performance. The internal mechanism has been adapted accordingly, and vCPU usage has been optimized.


OSWatcher Produces Large Number of InfiniBand Errors in Logs

OSWatcher is set up to collect metrics on all compute nodes. It uses the deprecated ifconfig -a command to retrieve network information, and this fills the log files with a large number of entries related to the InfiniBand devices in the compute nodes.

This issue has been addressed in a new version of OSWatcher, which is now included in the Oracle VM version that ships with Oracle PCA Release 2.3.3.


Compute Node Post-Upgrade Error: FM Not Installed

In certain controller software update scenarios, it occurs that the FM package cannot be installed because IPMI is not active on a given compute node. This results in an error during the post-upgrade tasks.

The compute node upgrade code has been improved to make sure that the IPMI service is always running before the FM package installation takes place.


Pre-Upgrade Validation Fails When No FC Modules Are Installed

On racks containing Fabric Interconnects with no Fibre Channel I/O modules installed, the pre-upgrade validation script fails, even though such a configuration is supported. The validation script has been updated to correctly execute all its tests on rack configurations with or without FC modules.