This section contains information on the known limitations and workarounds for Oracle VM. Issues that are related to Oracle VM Server for x86 only are marked with x86 Only in the title. Issues that are related to Oracle VM Server for SPARC only are marked with SPARC Only in the title. Issues that are related to both x86 and SPARC platforms do not have any special marking in the title.
This section contains the known issues and workarounds related to the Oracle VM Manager user interface.
If you connect to the Oracle VM Manager UI in a web browser twice, in two different tabs or windows, unexpected display issues may occur.
Workaround: You should only connect to one instance of the Oracle VM Manager UI per web browser session.
Bug 13034728
Closing the web browser tab or the entire browser that has been used to access Oracle VM Manager results in the closure of the session, as validated by WebLogic server, within the specified timeout period. By default this is set to 30 minutes. Any open or pending jobs associated with the session are aborted after this period.
Workaround: Ensure that the browser window or the tab in which Oracle VM Manager is running is not closed before all pending jobs have completed.
Bug 15838028
Some jobs that spawn many sub-tasks or child jobs, such as refreshing NFS storage with many refresh servers, may take longer to complete than indicated in the Oracle VM Manager user interface. This can cause some actions to be blocked until all of the child jobs have completed.
The status of all spawned child jobs is correctly indicated in the Jobs View within the Oracle VM Manager user interface.
Bug 15946119
The Create New VM Wizard consists of a series of steps, some of which are optional. Once the required parameters have been entered within the wizard, and prior to performing any optional steps, the virtual machine is actually created within the environment. If you attempt to cancel out of the wizard once you reach the optional configuration steps within the wizard, you discover that the virtual machine has still been created.
Workaround: Simply delete the unwanted virtual machine when you exit the wizard.
Bug 15968687
The Guest VM Console accessible from Oracle VM Manager may not work properly for particular versions of the Java Runtime Environment (JRE) in conjunction with some web browsers and operating systems. The JRE that has been shown to work most consistently is Java SE6 Update 35. On some systems other versions of the JRE may function as well.
Workaround: While it is possible to run concurrent versions of the JRE, it is recommended that to use the Guest VM Console in Oracle VM Manager, you completely remove any previously installed JRE along with any downloaded certificates or temporary files before installing the Java SE6 Update 35 from http://www.oracle.com/technetwork/java/javase/downloads/index.html on the client desktop computer that you are using to access Oracle VM Manager.
Bug 14497293 and Bug 14276511
RPM Packages are available to ensure the correct functioning of the Guest Serial Console and the VNC Console within Oracle VM Manager when accessing virtual machines. These packages must be installed on the Oracle VM Manager host. The packages that must be installed are the TightVNC package which provides VNC access to the guest virtual machine, and the JTA package which is used to provide serial console access to the guest virtual machine.
VNC access is only available for x86 guests. Serial console access is available for both x86 and SPARC guests.
Packages are available at Oracle's public yum server:
TightVNC (Oracle Linux 5): http://public-yum.oracle.com/repo/EnterpriseLinux/EL5/addons/x86_64/tightvnc-java-1.3.9-4.noarch.rpm
TightVNC (Oracle Linux 6): http://public-yum.oracle.com/repo/OracleLinux/OL6/addons/x86_64/tightvnc-java-1.3.9-4.noarch.rpm
JTA (Oracle Linux 5): http://public-yum.oracle.com/repo/EnterpriseLinux/EL5/addons/x86_64/jta-2.6-1.noarch.rpm
JTA (Oracle Linux 6): http://public-yum.oracle.com/repo/OracleLinux/OL6/addons/x86_64/jta-2.6-1.noarch.rpm
Bug 14395338
Opening a guest serial console from Oracle VM Manager for a SPARC guest sometimes results in the console displaying information about the telnet connection but the console remains unresponsive. This tends to happen when the VM is running the installation configuration tool.
Workaround: Use ~w to get write permission to the console. Then refresh the console display using Ctl-L. If that does not work, type ~# for a break, then click r to reset the VM and watch it reboot in the console. Thereafter the console should work properly.
Bug 14803291
On Microsoft Windows guests prior to Microsoft Windows Vista, if the IPv6 address is changed and the IPv4 address remains constant, Oracle VM Manager does not reflect the IP address change until the IPv4 address is changed as well.
Bug 14005821
If you drag and drop a resource, such as a virtual machine or server, to an invalid target it fails silently. Please see Drag and Drop in the Oracle VM User's Guide for more information.
Bug 15957926
After acknowledging an error event on a virtual machine, the text in the Event Severity column of the Virtual Machines perspective in the Servers and VMs tab is displayed in red, instead of black. The error event has been acknowledged, and the incorrect text color can be ignored.
Workaround: Navigate to another page in the UI, then return to the Virtual Machines perspective in the Servers and VMs tab. The text is the correct color, black.
Bug 14027051
kdump.conf
Ignored/var/crash
This section contains the known issues and workarounds related to Oracle VM Servers and server pools.
If a server pool has been created within Oracle VM Manager and has been configured to use the Xen hypervisor, but all servers have been removed from the server pool, it is possible to incorrectly add SPARC servers to the server pool in the Edit Server Pool dialog. Server pools must contain a homogenous collection of either SPARC or x86 servers. If you are managing both, make certain that you do not mix SPARC and x86 servers within the same server pool.
In cases where a server pool already contains servers, the list of available servers that can be assigned to it is automatically filtered to match servers that are able to run on the specified hypervisor.
Bug 14360725
When a server pool has finished being created, if an incorrect IP address was entered into the Virtual IP field, it is not possible to change the virtual IP address.
The only way to resolve this is to remove all the servers from the server pool, delete the server pool and create a new one.
Bug 12821548
The following server BIOS settings may be required to use Oracle VM Server:
AHCI mode may be necessary to recognize the CDROM device to perform an installation from CDROM.
Disable I/O MMU Virtualization Settings; for Intel-based servers this is VT-d; for AMD based servers this is AMD-Vi or IOMMU. I/O MMU is not supported in this release.
When using a PXE boot kickstart file to perform an Oracle VM
Server installation, make sure you specify the network interface
to be used for the management interface first. If you have more
than one network interfaces specified in a kickstart file, the
first interface is used as the management interface. The
management interface is the only live
interface after a PXE install of Oracle VM Server. You should
manually configure any other network interfaces on the Oracle VM
Server to start on boot in the
/etc/sysconfig/network-scripts/ifcfg-* files. Set the
ONBOOT=NO
parameter to ONBOOT=YES.
Bug 12557470
When using a PXE boot kickstart file to perform an Oracle VM Server installation, there is no option available to enable the installer setting to make the server boot from a multipath device on a SAN. The installer option to boot from a multipath SAN device can only be set manually during the installation procedure.
Workaround: Some success with a multipath kickstart installation may be acheived by following these instructions:
Create an
/etc/sysconfig/mkinitrd/boot-multipath
configuration file that contains the 'BOOT_MULTIPATH=yes'
directive.
Run mkinitrd.
Ensure that the /etc/blkid/blkid.tab
cache file that is created by mkinitrd is
removed at the end of your %post.
Bug 13967964
When installing Oracle VM Server on a machine with more than 128GB RAM, the memory allocated to Dom0 may turn out to be too restrictive for some configurations. For example, you may need to increase Dom0 memory in case the Oracle VM Server runs a large number of virtual machines and connects to a large number of LUNs.
Dom0 memory is calculated as follows: 2% of server RAM + 512MB. However, calculations are cut off at a maximum of 2% of 128GB RAM, meaning that Dom0 has a maximum of around 3GB RAM available. This should be sufficient for most scenarios, but an increase in Dom0 memory may be required to accommodate the demand of a particular environment.
Bug 13922885
There are cases where it may become necessary to increase the Dom0 memory size to meet the demand of running applications. For example, presenting one iSCSI LUN takes approximately 3.5MB of system memory. Consequently, a system that uses many LUNs quickly requires a larger amount of memory in accordance with the storage configuration.
Workaround: Change the amount of memory allocated to Dom0. See Changing the Dom0 Memory Size for information on how to change the Dom0 memory allocation.
Editing the kdump.conf
file on an Oracle VM Server
to add the "default shell" parameter does not have the intended
effect. The server reboots, but does not drop to a shell prompt.
Bug 15938733
When using kdump on an Oracle VM Server instance that crashes, the
vmcore that is generated is not dumped to the default path
/var/crash
, if the filesystem type and
mountpoint are not specified in
/etc/kdump.conf
. These options are usually
commented out by default when you install kdump. In order for
kdump to function correctly, your
/etc/kdump.conf
file should include lines
similar to the following:
ext3 /dev/sda1 path /var/crash
It is also important that the path configured for kdump has enough disk space to contain the vmcore file. By default, the root filesystem installed with Oracle VM Server is too small to contain the vmcore file. Therefore, you should either specify a different partition where there is disk space to match the machine memory or you should increase the root partition size according to the requirements of the given system during the installation stage.
Bug 15844134
When kdump on an Oracle VM Server is configured to use sparse files for vmcore dump files to reduce the dump file size, kdump fails with an error similar to:
Scanning and configuring dmraid supported devices ... cp: unrecognized option `--sparse=always' ...
Workaround: Add the following to the kdump.conf on the Oracle VM Server:
core_collector /bin/cp --sparse=always
Bug 15959684
Using kdump on an Oracle VM Server over a network using a XEN bridged interface fails and is not supported in this release.
Bug 16085278
Installing Oracle VM Server on a system such as the Dell 380, with a Broadcom Gigabit Ethernet Controller fails with an error similar to the following example:
Traceback (most recent call first): File "/usr/lib/anaconda/network.py", line 685, in write if dev.get('BOOTPROTO').lower() in ['dhcp', 'ibft']: File "/usr/lib/anaconda/yuminstall.py", line 1394, in doPreInstall anaconda.id.network.write(anaconda.rootPath) File "/usr/lib/anaconda/backend.py", line 184, in doPreInstall anaconda.backend.doPreInstall(anaconda) File "/usr/lib/anaconda/dispatch.py", line 207, in moveStep
The installer cannot detect the network adapter, so fails to complete the installation.
Bug 13387076
The Oracle VM Server installer is unable to install to a disk that is using a GUID partition table (GPT) or a Unified Extensible Firmware Interface (EUFI). Currently Oracle VM only uses the standard BIOS to support MBR disk installation. Attempting to install on a GPT disk results in an error and the installation fails.
Workaround: Use Alt-F2 to access a different terminal screen to the installer. Use fdisk from the command line to manually repartition the disk using a dos partition table. Alternatively repartition the disk prior to installation, ensuring that the partition table is not using GPT.
Bugs 14674445, 15856811
If you are installing Oracle VM Server on a Sun Fire X4800, you must provide extra parameters when booting from the installation media (CDROM or ISO file), or when using a kickstart installation. These parameters allow the megaraid_sas driver to load correctly.
If booting from the installation media, press F2 when the initial boot screen is displayed and provide the following additional parameters as part of the boot command:
mboot.c32 xen.gz extra_guest_irqs=64,2048 nr_irqs=2048 --- vmlinuz --- initrd.img
If using a kickstart installation, add the additional kernel parameters to the PXE configuration file.
If you want to make these changes permanent, edit the /boot/grub/grub.conf file in your Oracle VM Server after the installation has completed.
Bug 12657272
Installation of Oracle VM Server may fail on some HP server hardware when the tape drive is attached, rendering error messages describing a problem with a call to "makeDriveDeviceNodes". To avoid this issue, disconnect the tape drive and begin the installation again.
Bug 14667345
When creating a SPARC-based server pool, the Storage Location field in the Create a Server Pool dialog box is grayed out and unavailable. As HA of SPARC-based servers is not available in this release, a server pool file system is not required. All SPARC-based server pools must be unclustered.
Workaround: Create an unclustered server pool.
Starting or restarting Oracle VM Servers fails if the Oracle VM Servers in a server pool are on different subnets.
Workaround: Use IPMI (Intelligent Platform Management Interface) to start or restart Oracle VM Servers in a server pool that are on different subnets.
Bug 12410458
If Oracle VM Manager is redeployed to a new computer, you should rediscover any file servers. If you do not rediscover the file servers, and the server pool file system is on the file server, you cannot remove Oracle VM Servers from the server pool.
Bug 12333132
If you have server pools in your environment and you clear the Oracle VM Manager database, you cannot rediscover and rebuild your previous Oracle VM environment. The following error is in the job:
OVMRU_000021E Cannot perform operation on pool: Unknown pool found for discovered Pool FS. The pool does not have a Virtual IP Address.
Workaround: Follow these steps to rediscover the server pools:
Discover one Oracle VM Server from the server pool.
Register, and refresh your storage server.
Refresh the file system that contains the server pool file system.
Refresh the file systems that contains the repositories.
Refresh the repositories.
Refresh all Oracle VM Servers in the server pool to discover the virtual machines.
Bug 12724969
If the master Oracle VM Server in a clustered server pool becomes unavailable or loses its connection to the storage containing the server pool file system, another Oracle VM Server takes over the master role and the server pool virtual IP. When the unavailable server comes back online, it rejoins the cluster, unless its access to the server pool file system is not (yet) restored. This is where the inconsistency occurs. The original master server contains information indicating it has the master role, while another server in the cluster may have assumed the master role in the meantime. If the entire cluster has been down, the original master server can continue to fulfill that role. If the cluster remained operational, however, changes may have occurred that the original master server has no information about. As a result, two Oracle VM Servers in the same clustered server pool may claim the master role and the virtual IP, and Oracle VM Manager may not be able to resolve that conflict.
Workaround: Manually assign the master role either to the original master or to another server that is still active within the cluster. Follow these steps:
In Oracle VM Manager, open the Servers and VMs tab.
In the navigation pane, select the server pool and click Edit Server Pool.
Select the appropriate server as master server and click OK to save your changes.
Bugs 13875603
The OCFS2 heartbeating function can be disturbed by I/O-intensive operations on the same physical storage. For example: importing a template or cloning a virtual machine in a storage repository on the same NFS server where the server pool file system resides may cause a time-out in the heartbeat communication, which in turn leads to server fencing and reboot.
Workaround: To avoid unwanted rebooting, it is recommended that you choose a server pool file system location with sufficient and stable I/O bandwidth. Place server pool file systems on a separate NFS server or use a small LUN, if possible.
Bug 12813694
If OCFS2 file systems are still mounted on an Oracle VM Server you want to remove from a cluster, the remove operation may fail. This is due to the fact that the OCFS2 mount is an active pool file system or storage repository.
Workaround: If a storage repository is still presented, unpresent it from the Oracle VM Server before attempting to remove the Oracle VM Server from the cluster. If a pool file system is causing the remove operation to fail, other processes might be working on the pool file system during the unmount. Try removing the Oracle VM Server at a later time.
An Oracle VM Server is in heartbeat configured mode if the server pool file system is not mounted. Either the file system failed to mount or the mount was lost because of hardware issues such as unavailability of the LUN of NFS server containing the file system to be mounted.
Workaround: Mount the file system at the Oracle VM Server command line, or, as a last resort, reboot the Oracle VM Server and allow it to join the cluster automatically at boot time.
In some situations where a Sun Fire X4170 server has a hardware failure or fault/error, the Oracle VM Server may not be fenced and removed from the cluster, even though it is not available to participate in the cluster.
Workaround: Fix the hardware fault, and/or clear the error messages/faults using the ILOM CLI.
Bug 14463948
To use netconsole you must specify a non-bridged ethx device, in the /etc/sysconfig/netconsole file on an Oracle VM Server, for example:
# The ethernet device to send console messages out of (only set this if it # can't be automatically determined) # DEV= DEV=eth2
Bug 12861134
On some Intel-based systems, the following error may occur after installing Oracle VM Server when the computer is started:
[ 0.674742] ACPI Error: Field [CPB3] at 96 exceeds Buffer [NULL] size 64 (bits) (20090903/dsopcode-596) [ 0.675167] ACPI Error (psparse-0537): Method parse/execution failed [\_SB_._OSC] (Node ffff88002e4fba50), AE_AML_BUFFER_LIMIT
This has been observed on systems with the following BIOS information, but may also occur in other BIOS versions:
Vendor: Intel Corp. Version: S5500.86B.01.00.0036-191.061320091126 Release Date: 06/13/2009 BIOS Revision: 17.18 Firmware Revision: 0.0
HP ProLiant BL685c G6 HP BIOS A17 12/09/2009 Backup Version 12/01/2008 Bootblock 10/02/2008
This error can safely be ignored.
Bugs 12865298 and 12990146
Cisco Blade servers cannot be configured to boot with random UUIDs. Each Oracle VM Server must have a fixed UUID to allow the Oracle VM Server to move between Blades. To configure a fixed UUID, see:
Bug 13000392
When booting Oracle VM Server on a system using a Radeon graphics adapter, and error may appear in the dmesg output similar to the following:
[drm:r100_ring_test] *ERROR* radeon: ring test failed (scratch(0x15E4)=0xCAFEDEAD) [drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22). radeon 0000:01:03.0: failed initializing CP (-22). radeon 0000:01:03.0: Disabling GPU acceleration
Workaround: Ensure that the
radeon module is loaded with the modeset parameter set to zero.
To do this, add the following line in
/etc/modprobe.conf
:
options radeon modeset=0
In order for the change to take effect, you should reboot the server.
Alternatively, the parameter radeon.modeset=0
can be appended to the boot options within
/boot/grub/grub.conf
. For example:
title Oracle VM Server (2.6.39-300.20.1.el5uek) root (hd0,0) kernel /xen.gz dom0_mem=987M module /vmlinuz-2.6.39-300.20.1.el5uek ro root=UUID=8d7687eb-fa2b-42ef-b411-044ba66cca6c radeon.modeset=0 module /initrd-2.6.39-300.20.1.el5uek.img
Bug 14376597
This section contains the known issues and workarounds related to virtual machines.
Any virtual machine which has resources in ocfs2 repository cannot be migrated to Unassigned Virtual Machines folder, the only exception is if the ownership of this OCFS2 repository has already been released.
Bug 16005634
Any changes to the virtual machine configuration file should be performed using Oracle VM Manager. The only exception to this is if advised to manually edit the vm.cfg file by Oracle Support Services, or as explicitly specified in these Release Notes.
Manual changes made to the virtual machine configuration file (vm.cfg) are not reflected in Oracle VM Manager. Any manual changes to the vm.cfg file may result in unexpected and undesirable behaviour. For example, if you edit the HA setting in the vm.cfg to disable HA, and the virtual machine is stopped by any method other than using Oracle VM Manager, the virtual machine is restarted. Oracle VM Manager is not aware of the HA change in the virtual machine's configuration file.
Bugs 12654125 and 13391811
When importing and deleting several templates concurrently, or when an Oracle VM Server is removed during the import of a template, a lock exception error may appear. However, the template upload often completes successfully despite the error message, but it does not appear in the list of available templates in the storage repository.
To resolve this problem, refresh the storage repository. The uploaded template file should then appear in the list. In some cases the imported template turns out to be incomplete. In that case, you should delete the template and import the template again.
Running more than one clone job at a time causes the subsequent clone job(s) to silently fail, without error. This only occurs when you use the cloning wizard multiple times, not when cloning an object multiple times.
Workaround: Run only one clone job at a time, then wait for the cloning job to complete before running the cloning wizard again.
Bug 16013472
When live migration of a virtual machine fails, Oracle VM Manager reports the migration job as failed. However, the migrated virtual machine may appear under the correct target Oracle VM Server in running status, as if the migration had completed successfully. At the same time, the virtual machine remains in running status on the Oracle VM Server it was supposed to be migrated away from, but Oracle VM Manager reports it as stopped.
Workaround: Kill the virtual machine affected by the failed migration operation on the target Oracle VM Server; then restart the server. Rediscover the Oracle VM Server on which the virtual machine was originally running, and it should appear again in running status.
Bugs 13939895, and 13939802
A virtual machine may fail to restart on another server within the cluster when the original server becomes unavailable and ACPI power management is enabled within the BIOS for either of the Oracle VM Servers within the cluster.
Workaround: Disable ACPI within the BIOS for each of the Oracle VM Servers within the cluster.
Bug 14747722
When performing a graceful shutdown from within a virtual machine running Oracle Solaris 10, or when attempting to stop the virtual machine from within Oracle VM Manager, or when performing a migration of the virtual machine, the virtual machine hangs during the shutdown process.
This bug does not affect virtual machines running Oracle Solaris 11.
Workaround: Kill the virtual machine affected using the Kill VM option within Oracle VM Manager.
Bug 15863719
When you update an Oracle VM Server, the following jobs are submitted:
Server Yum update check operation
Server Maintenance Mode
Queued Server Yum upgrade
The Yum Update on Server
server_name
job displayed
in the Oracle VM Manager Job
Summary pane indicates that these three jobs have
been successfully submitted, but may not mean that the Oracle VM Server
has actually been updated. To see which version the Oracle VM Server is
running before and after an upgrade, click the
Servers and VMs tab,
select the Oracle VM Server in the navigation tree and then select
Control Domains in the
Perspective
drop-down list.
If the Yum update fails due to the ServerEvacuateFailedEvent event (use the Events perspective when selecting an Oracle VM Server to check for this event), a virtual machine is running that cannot be automatically migrated. To remedy this situation, either shut down the running virtual machine on the Oracle VM Server, or provide more resources in the server pool so that the running virtual machine can be migrated to another Oracle VM Server.
Bug 16080382
Creating a virtual machine using the Network method (PXE) does not proceed beyond pre boot, so the virtual machine is not created. This occurs for Oracle Linux 5.x virtual machines.
Bug 12905120
Creating hardware virtualized guests requires the Oracle VM Server has an Intel-VT (code named as Vanderpool) or AMD-V (code named as Pacifica) CPU. See the Oracle VM Installation and Upgrade Guide for a list of supported hardware.
Creating a PVM guest using local storage may take a very long time. This may be caused by write caching being turned off for the local disk.
Workaround: Enable write caching for the disk using the hdparam utility.
Bug 12922626
If the configuration file of a PVHVM virtual machine or template lists a virtual CDROM drive as an IDE device, that is as /dev/hda, dev/hdb, and so on, the virtual CDROM is not available inside the guest. To enable the CDROM drive inside the guest, it must be defined in the vm.cfg file as a paravirtual device, that is /dev/xvda, /dev/xvdb, and so on.
Bug 14000249
When creating or editing a virtual machine, you can set both the
number of CPUs it uses and the maximum number of CPUs it is
allowed to use. However, if the guest is configured with a large
number of virtual CPUs, the respective vcpus
and maxvcpus
values are not passed correctly
to the hypervisor. As a result, the creation of the virtual
machine fails, or if the virtual machine already exists, it
fails to start.
Workaround: When creating or editing a virtual machine with 32 or more virtual CPUs, do not enter a value for the maximum number of CPUs.
Bug 13823522
If a guest is configured with a large number of virtual CPUs, the virtual machine may hang at some point in the boot sequence. This is likely caused by the kernel or the hypervisor, and there is no workaround if you need more than 32 virtual CPUs.
Bug 12913287
Linux UEK x86_64 2.6.32-100.26.2.el5 as guest kernel does not work well for live migration. Please update to the latest UEK.
Bug 14276606
For both hardware virtualized (HVM) guests and hardware virtualized guests with paravirtualized drivers (PVHVM), the possibility to change the number of virtual CPUs is limited by the kernel of the virtual machine. The table below provides an overview of guest kernel support; it applies to both x86 and x86_64 guest architectures.
Table 10 Guest kernel support for hot-changing virtual CPUs
Oracle Linux Version | Type | Kernel Version | Hot-Add | Hot-Remove |
---|---|---|---|---|
Oracle Linux R5U5 | PVM | 2.6.18-194.0.0.0.3.el5xen | Yes | Yes |
Oracle Linux R5U5 | PVHVM | 2.6.18-194.0.0.0.3.el5 | No | No |
Oracle Linux R5U6 | PVM | 2.6.18-238.0.0.0.1.el5xen | Yes | Yes |
Oracle Linux R5U6 | PVHVM | 2.6.18-238.0.0.0.1.el5 | No | No |
Oracle Linux R5U7 | PVM | 2.6.32-200.13.1.el5uek | No | Yes |
Oracle Linux R5U7 | PVHVM | 2.6.32-200.13.1.el5uek | No | No |
Oracle Linux R5U8 | PVM | 2.6.32-300.10.1.el5uek | No | Yes |
Oracle Linux R5U8 | PVHVM | 2.6.32-300.10.1.el5uek | No | No |
Oracle Linux R6U1 | PVM | 2.6.32-100.34.1.el6uek | Yes | Yes |
Oracle Linux R6U1 | PVHVM | 2.6.32-100.34.1.el6uek | No | No |
Oracle Linux R6U2 | PVM | 2.6.32-300.3.1.el6uek | Yes | Yes |
Oracle Linux R6U2 | PVHVM | 2.6.32-300.3.1.el6uek | No | No |
Oracle Linux R5U8 with UEK2 | PVM | 2.6.39-100.5.1.el5uek | No | Yes |
Oracle Linux R5U8 with UEK2 | PVHVM | 2.6.39-100.5.1.el5uek | No | No |
Oracle Linux R6U2 with UEK2 | PVM | 2.6.39-100.5.1.el6uek | Yes | Yes |
Oracle Linux R6U2 with UEK2 | PVHVM | 2.6.39-100.5.1.el6uek | Yes | No |
Bugs 12913287, 13905845, 13823853, and 13898210
When using the virtual machine console, the following error may be displayed:
JAR sources in JNLP file are not signed by the same
certificate
This occurs if the client JRE version is 1.6.0_14.
Workaround: Upgrade the client JRE version to 1.6.0_30 or above.
Bug 13621606
When you launch the virtual machine console from Oracle VM Manager you may experience that the mouse pointer on your local machine and the mouse pointer in the virtual machine travel across the screen at different speeds.
If your guest virtual machine's operating system is Linux-based, the following workaround may reduce the mouse control issue. Enter the following on the guest's command line:
# xset m 1 1
Virtual machines with the guest operating system Solaris 10 may experience kernel panic on systems with AMD processors. Kernel panic has been seen in Solaris 10 9/10 (Update 9) and Solaris 10 8/11 (Update 10).
Workaround: To work around this issue:
During the installation or first time boot up, edit the grub
menu and append the -kd
kernel boot
parameter.
Continue with the boot to run the Solaris kmdb. When the following screen prompt is displayed:
Welcome to kmdb [0]>
Enter the command:
cmi_no_init/W 1
Enter the following to continue the installation or system boot:
:c
After Solaris is installed and booted up, append the following line to the /etc/system file to make this change persistent across system reboot.
set cmi_no_init = 1
Bug 13332538
When booting a virtual machine with Oracle Solaris 10 Release 8/11, the guest OS hangs when the copyright information screen appears. This is caused by a change in CPUID handling in dom0, which triggers a Solaris bug on platforms with CPUs of the Westmere-EP family.
Workaround: To make Solaris 10 run, kill the virtual machine from Oracle VM Manager and then apply the following manual fix:
At boot, edit the grub menu: append the
-kd
kernel boot parameter. This runs the
Solaris kernel debugger.
Continue the boot sequence up to Solaris
kmdb
.
At the kmdb
prompt, enter the following
command:
Welcome to kmdb
[0]> apix_enable/W 0
Enter :c
to continue the system boot
sequence.
When Solaris has been installed and has successfully booted,
make this fix persistent by adding the following line to
/etc/system
:
set apix_enable = 0
Bug 13876544
When booting a virtual machine with Oracle Solaris 11 Release 2011.11, the guest OS hangs. This is caused by a Solaris bug where interrupt storms occur on Intel systems based on Sandy bridge and Westmere CPUs. The issue has been fixed in Solaris 11 2011.11 SRU 2a.
Workaround: To make Solaris 11 2011.11 run on Oracle VM 3.2, use the SRU 2a, or apply the following manual fix to the GA release:
At boot, edit the grub menu: append the
-kd
kernel boot parameter. This runs the
Solaris kernel debugger.
In the kernel debugger, enter the following commands:
[0]>::bp pcplusmp`apic_clkinit
[0]>:c
kmdb: stop at pcplusmp`apic_clkinit kmdb: target stopped at: pcplusmp`apic_clkinit: pushq %rbp [0]>apic_timer_preferred_mode/W 0
pcplusmp`apic_timer_preferred_mode: 0x2 = 0x0 [0]>:c
Continue the system boot sequence.
When Solaris has been installed and has successfully booted,
make this fix persistent by adding the following line to
/etc/system
:
set pcplusmp:apic_timer_preferred_mode = 0x0
For more information, please consult the Support Note with
ID
1372094.1. You can also find this document by logging on
to My Oracle
Support and searching the knowledge base for
1372094.1
.
Bug 13885097
Windows Server 2008 Release 2 64-bit hardware virtualized guests fail to shut down cleanly. After the guest has been shut down, and started again, Windows reports that it was not shut down cleanly.
This is not an issue for Windows Server 2008 Release 2 32-bit hardware virtualized guests with paravirtualized drivers (PVHVM).
Bug 12658534
When you add a new disk to a virtual machine and refresh the device manager, the new disk is displayed with a yellow mark. This occurs in Microsoft Windows guests that have the Oracle VM Windows Paravirtual Drivers for Microsoft Windows Release 2.0.7 installed.
Bug 12837744
When you add a new disk to a virtual machine, the new disk is not automatically detected. This occurs in Microsoft Windows 2008 Release 2, 64-bit guests that have the Oracle VM Windows Paravirtual Drivers for Microsoft Windows Release 2.0.7 installed.
Workaround: After you add a new disk, scan for new hardware changes using Server Manager > Disk Drives > Scan for hardware changes.
Bug 12837004
Netconsole does not work on a virtual machine running SUSE Linux Enterprise Server 11 (SLES) guest, as polling is not supported by xennet.
Novell Bug ID 763858
It is not possible to migrate a guest VM that has been stopped or suspended using the Oracle VM Command Line Interface.
Workaround: If a VM is stopped or suspended, perform a remove VM and then and add VM operation to successfully perform a migration. Alternatively, use the Oracle VM Manager User Interface, which fully supports migration of suspended and stopped VMs, to perform the operation.
Bug 14501247
If a virtual machine is powered off from the OS running within the virtual machine, using the poweroff command for example, and if you subsequently try to remove the VM role from the network used by this virtual machine then this will fail with an error similar to the following:
command "remove-vswitch": failure (failure: Service is active and being used by bound client(s) Failed to remove virtual switch)
This is caused because the poweroff command stops the virtual machine, but it remains bound on the Oracle VM Server. When you attempt to remove the VM role, Oracle VM attempts to remove the corresponding bridge or vswitch on the Oracle VM Server, but this fails because the virtual machine is still bound.
Workaround: Do not use the poweroff command from within the virtual machine, but rather stop the virtual machine from within Oracle VM Manager. If the poweroff command has already been used, use Oracle VM Manager to kill the stopped virtual machine before attempting to remove the VM role from the network.
Bug 15942008
Although Oracle VM Manager provides the option to add a sparse-allocate virtual disk to a virtual machine, cloning the virtual machine results in a non-sparse file for the cloned virtual machine. This is because sparse copy is not supported by the Solaris dd command.
Bug 14851945
When adding or removing memory for a running virtual machine Oracle VM Server for SPARC, the request to add or remove memory may be partially fulfilled. In such a case, Oracle VM Manager does not report any error after the memory add or remove operation, and Oracle VM Manager indicates that the entire operation has succeeded. However the virtual machine is actually allocated an amount of memory different from the amount indicated by Oracle VM Manager.
The correct amount of memory will be restored when the virtual machine is stopped and then restarted.
Bug 14844937
If the virtual machine memory is less than the available memory on the Oracle VM Server that the virtual machine is to be migrated to, the migration may fail with an error message stating that there is not enough memory available to complete the request. This is due to memory fragmentation on the target server. The details of this are discussed in http://docs.oracle.com/cd/E35434_01/html/E23807/migratingactivedomain.html#memory.
Bug 14851322
For virtual machines not running Oracle Solaris 11.1 or greater, dynamic reconfiguration (DR) of memory is disabled after the virtual machine is migrated. This is a documented issue and is described at http://docs.oracle.com/cd/E37707_01/html/E29665/migratingactivedomain.html#memory. In order to re-enable DR, simply reboot the migrated virtual machine.
Bug 14844852
This section contains the known issues and workarounds related to networks and networking.
The vlangroup name is used as the ID when a vlangroup is created. Subsequently (since IDs must be unique) a vlangroup's name must be unique. This means that you cannot create two vlangroups with the same name.
Bug 14470245
If you have a management network set up and none of the servers on that network have the VM Role enabled, enabling the VM Role on one of the servers creates a bridge on that server; going on to discover another server on the network with the VM Role enabled, results in the server being discovered but the bridge not being created.
Workaround: This can be achieved in the Oracle VM Manager user interface by removing the VM Role from the management network and then adding it again. Unfortunately, this would require you to stop all running virtual machines. To effect the change without removing the VM Role first, you can set the role list for a network in the Oracle VM Command Line Interface. As long as there is at least one change in the list of roles, the CLI applies the set of roles for the network. This causes the bridge to be created. The simplest way to force a change in the list of roles would to be add or remove the Storage role at the same time that you add the VM role. After the VM role is added, you can then add back or remove the Storage role as needed to re-set the initial state. Adding and removing the storage role should have no impact while this change is occurring.
Bug 15888317
Oracle VM Manager supports eight network cards for each HVM virtual machine as outlined in Table 5 “Virtual machine maximums”. However, the system library does not allow users to add more than three network cards when creating a virtual machine from installation media.
Workaround: After the virtual machine is created, add up to five new network cards by editing the virtual machine in Oracle VM Manager.
QLogic currently does not provide a supported driver for the
QLE8242 CNA network card on Oracle VM Server Release 3.1 or Release 3.2.
Installation of the provided RedHat driver fails with the error
message, Cannot determine the Redhat release
.
Bug 14300919
On some systems, such as the HP ProLiant DL980 G7 server, the QLogic cLOM8214 1/10GbE NIC initializes with link state down.
Workaround: Disable the
use_msi_x option for the qlcnic module in
/etc/modprobe.conf
:
options qlcnic use_msi_x=0
In order for the change to take effect, you should reboot the server. To temporarily disable the option for the module without rebooting, you can perform the following steps:
# ifconfig ethx down # rmmod qlcnic # modprobe qlcnic use_msi_x=0
Bug 15865548
On some systems at initial install, udev may rename the QLogic cLOM8214 1/10GbE Ethernet port to __tmpxxxxxxxx. This is due to a slow initialization of the port.
Workaround: Determine which interface name the QLogic cLOM8214 1/10GbE Ethernet NIC should use, for example eth1. Find the MAC address for the QLogic cLOM8214 1/10GbE Ethernet NIC that you want to configure. You can do this by running the following command and studying its output to find the incorrectly named interface:
# cat /etc/issue IF : bond0 MAC : 1c:c1:de:74:b9:34 IP : 10.211.1.135 IF : eth0 MAC : 1c:c1:de:74:b9:34 IF : eth1 MAC : 1c:c1:de:74:b9:35 IF : __tmp713192749 MAC : 24:be:05:ef:5a:f4
In /etc/sysconfig/network-scripts
, create
an ifcfg-
configuration file for the interface name that you wish to use
and ensure that the file contains a line that expresses the MAC
address for the NIC that you want to assign to this interface
name. For example, using the example output from the previous
command, you could create a new file at
ethx
/etc/sysconfig/network-scripts/ifcfg-eth2
with the following contents:
DEVICE=eth2 BOOTPROTO=none HWADDR=24:BE:05:EF:5A:F4 ONBOOT=yes
Finally, bring up the interface using the ifconfig command. For example:
ifconfig eth2 up
Bug 15897549
To solve a networking bug where Broadcom 10 Gigabit ethernet NICs would throw crash information on HP ProLiant DL980 G7 servers when used for a virtual machine network, resulting in adverse performance, TPA is disabled for the bnx2x driver module.
If the port is used for Dom0 traffic only, you can enable TPA by
changing the entry for the bnx2x module in
/etc/modprobe.conf
to:
options bnx2x disable_tpa=0
Bug 15971208
The Broadcom BCM5754 Gigabit Ethernet network controller does not support Jumbo Frames.
VLANs over bond mode 6 (balance-alb) bridge interface are not supported as this mode is incompatible with VLAN bridge interfaces.
Workaround: There are two workarounds for this problem:
Use bond mode 6 as a bridge interface; do not use VLANs over bond mode 6.
Use VLANs over bond modes (1=active-backup or 4=802.3ad) as a bridge interface.
If you move the Cluster Heartbeat network role to another network, with a different IP address, the change is not reflected in the Oracle VM Servers.
Workaround: Edit the /etc/ocfs2/cluster.conf file on each Oracle VM Server in the network to reflect the new IP address, and restart each Oracle VM Server.
Bug 12870099
This section contains the known issues and workarounds related to storage.
If a server pool file system is not clean (contains existing files and server pool cluster information) and used to create a server pool, a number of errors may occur.
Cannot create a server pool using the file system. The following error is displayed:
OVMAPI_4010E Attempt to send command: create_pool_filesystem to server: server_name
failed.
OVMAPI_4004E Server Failed Command: create_pool_filesystem ... No such file or directory
Bug 12839313
An OCFS2-based storage repository becomes orphaned (the clusterId that was used when the OCFS2 file system was created no longer exists), you cannot mount or refresh the repository, and the following error is displayed:
"OVMRU_002037E Cannot present the Repository to server: server_name
. Both server and repository
need to be in the same cluster."
Workaround: Clean the file system of all files before it is used as a server pool file system.
Bug 12838839
When discovering a generic ISCSI SAN server using CHAP authentication, it is important to select the Use Chap check box in the Access Information step of the Discover SAN Server wizard before adding the access host. If you add an access host first, then select the Use Chap authentication box, an error is displayed at the end of the wizard:
OVMRU_007202 ... The access host value must be set for this type of storage server.
Workaround: In the Access Information step of the Discover SAN Server wizard, select the Use Chap check box, then add an access host. The error is no longer displayed.
Bug 16036915
When you resize a LUN and rescan the physical disks on the storage array, the new size is not reflected in Oracle VM Manager.
Bug 12772588
If a LUN belonging to an unmanaged Fibre Channel array is unmapped from a server and a new target is mapped to it using the same LUN, then after rescanning the physical disks on that server, the size of the new LUN is set to the size of the LUN that was unmapped.
Remapping LUNs is risky because it can data corruption since the targets have been switched outside of the server. An error should appear in the messages file:
Warning! Received an indication that the operating parameters on this target have changed. The Linux SCSI layer does not automatically adjust these parameters.
Workaround: Ensure that when exposing a new target, you do so with a LUN that has not previously been used since the last reboot.
Bug 14851579
When you attempt to create a storage repository on a LUN that was previously used by another clustered server pool, the operation will fail. This is due to a built-in mechanism that prevents the creation of a new OCFS2 file system if the disk or partition already contains cluster data.
Workaround: Clear all files and file system information on the LUN before placing a storage repository on it.
Bugs 13806344
Oracle VM Server cannot add the system disks for Legacy LSI MegaRAID (Dell PERC4) bus controllers to the /etc/blacklisted.wwids file, so the disks are not blacklisted in the multipath configuration. This occurs because the bus controllers are not capable of returning a unique hardware ID for each disk. Using system disks on Legacy LSI MegaRAID (Dell PERC4) bus controllers is therefore not supported.
Bug 12944281
Installing Oracle VM Server on an HP Smart Array (CCISS) fails to blacklist system disks (they are not included in the /etc/blacklisted.wwids file). Messages similar to the following are logged in the /var/log/messages file:
multipathd: /sbin/scsi_id exited with 1 last message repeated 3 times
Workaround: Configure multipathing to blacklist the CCISS system devices by adding a new line to the multipath.conf file:
# List of device names to discard as not multipath candidates
#
## IMPORTANT for OVS do not remove the black listed devices.
blacklist {
devnode "^(ram|raw|loop|fd|md|dm-|sr|scd|st|nbd)[0-9]*"
devnode "^hd[a-z][0-9]*"
devnode "^etherd"
+ devnode "^cciss!c[0-9]d[0-9]*" <<====
%include "/etc/blacklisted.wwids"
}
This workaround is not required for HP Smart Array disks backed by the HPSA module.
Bug 12722044
HP Smart Array disk devices backed by the older CCISS module are not supported for use as a local storage repository, or for use as raw disks for virtual machines, or for server pool file systems. HP Smart Array disk devices backed by the HPSA module are fully supported.
To verify whether the controller has support for the HPSA module, first obtain the PCI ID of the raid controller for the local disk using the following command:
lspci -vnn|grep "RAID bus controller"
Check the list of controllers supported by the HPSA module to confirm whether or not your controller's PCI ID appears within the list, by running the following command:
modinfo hpsa -F alias
If your controller is supported, ensure that you use the HPSA module instead of the CCISS module for your controller.
Bug 16072257
When an NFS file server has two IP addresses, it cannot expose the same file system over both interfaces. This situation would occur if you configure both IP addresses as separate access hosts; for example to provide access to different Oracle VM Servers via different paths. As a result, the same file system would correspond with two different storage object entries, each with a different path related to each of the IP addresses. As a storage server can only be represented by one object, this configuration is not supported in Oracle VM Release 3.2.
Workaround: Configure only one access host per storage server.
On a multipath SAN boot server with only one active path from server to storage, there is a potential risk of kernel panic, or the file system becoming read-only when doing storage rescans. This occurs because multipath SAN boot is only supported with full path redundancy. A minimal full path redundancy configuration can be illustrated as:
Server HBA1 --- FC switch 1 ---- Storage controller 1 port1 |-- Storage controller 2 port1 Server HBA2 --- FC switch 2 ---- Storage controller 1 port2 |-- Storage controller 2 port2
Workaround: Configure multiple paths to the file systems, instead of a single path.
Bug 13774291
When performing a fail-over (for example, by disabling one protocol such as Fibre Channel), the paths are grouped and the traffic flows through all protocols (for example, both Fibre Channel and ISCSI), which is not as expected. This also occurs with fail-back. This issue occurs in Oracle VMRelease 3.1.1 and Release 3.2.
Workaround: Run the following commands on the Axiom server:
# service multipathd restart # multipath -v3
Bugs 14329375 and 14314282
When a NAS-based file system is refreshed, it may produce invalid or overlapping exports. During a file system refresh job, all mount points defined in the NAS-based file server's exports file are refreshed, even file systems that are not intended to be used in Oracle VM environments.
Top level directories which also contain subdirectories in the exports file may also cause problems, for example, if an export file contains /xyz as an export location, and also contains /xyz/abc. In this case, the following error may be displayed during a refresh file system job:
OVMRU_002024E Cannot perform operation. File Server: server_name
, has invalid exports.
Workaround: For the second issue, to work around this problem, do not export top level file systems in the NAS-based file server's exports file.
Bug 12800760
Only local SAS storage is supported with Oracle VM Manager. Oracle VM Release 3.2 does not support shared SAS storage (SAS SAN), meaning SAS disks that use expanders to enable a SAN-like behavior can only be accessed as local storage devices. Oracle VM Manager recognizes local SAS disks during the discovery process and adds these as Local File Systems. SAS SAN disks are ignored during the discovery process and are not accessible for use by Oracle VM Manager.
It is possible to determine whether SAS devices are shared or local by running the following command:
# ls -l /sys/class/sas_end_device
Local SAS:
lrwxrwxrwx 1 root root 0 Dec 18 22:07 end_device-0:2 ->\ ../../devices/pci0000:00/0000:00:01.0/0000:0c:00.0/host0/port-0:2/end_device-0:2/sas_end_device/ end_device-0:2 lrwxrwxrwx 1 root root 0 Dec 18 22:07 end_device-0:3 ->\ ../../devices/pci0000:00/0000:00:01.0/0000:0c:00.0/host0/port-0:3/end_device-0:3/sas_end_device/ end_device-0:3
SAS SAN:
lrwxrwxrwx 1 root root 0 Dec 18 22:07 end_device-0:0:0 -> \ ../../devices/pci0000:00/0000:00:01.0/0000:0c:00.0/host0/port-0:0/expander-0:0/port-0:0:0/ end_device-0:0:0/sas_end_device/end_device-0:0:0 lrwxrwxrwx 1 root root 0 Dec 18 22:07 end_device-0:1:0 -> \ ../../devices/pci0000:00/0000:00:01.0/0000:0c:00.0/host0/port-0:1/expander-0:1/port-0:1:0/ end_device-0:1:0/sas_end_device/end_device-0:1:0
For SAS SAN storage, note the inclusion of the expander within the device entries.
Bug 13409094
Oracle Storage Connect plug-ins for generic as well as vendor-specific storage hardware exist in different versions that have been adapted for use with a particular release of Oracle VM Server. If storage operations in Oracle VM Manager fail consistently with your storage plug-in, verify that the correct plug-in version is installed on your Oracle VM Servers. The lists below show compatibility of Oracle Storage Connect plug-ins for Oracle VM Server Release 3.0.3, 3.1.1 and Release 3.2.
Oracle VM Server Release 3.2 (GA release) compatible plug-ins:
osc-plugin-manager-1.2.8-26.el5
osc-plugin-manager-devel-1.2.8-26.el5
osc-oracle-ocfs2-0.1.0-38.el5
osc-oracle-netapp-1.2.8-7.el5
osc-oracle-generic-1.1.0-74.el5
osc-oracle-s7k-0.1.2-50.el5
Oracle VM Server Release 3.1.1 (GA release) compatible plug-ins:
osc-plugin-manager-devel-1.2.8-19.el5
osc-oracle-netapp-1.2.8-6.el5
osc-plugin-manager-1.2.8-19.el5
osc-oracle-generic-1.1.0-55.el5
osc-oracle-s7k-0.1.2-45.el5
osc-oracle-ocfs2-0.1.0-36.el5
Oracle VM Server Release 3.0.3 (GA release) compatible plug-ins:
osc-plugin-manager-devel-1.2.8-9.el5
osc-oracle-netapp-1.2.8-1.el5
osc-plugin-manager-1.2.8-9.el5
osc-oracle-generic-1.1.0-44.el5
osc-oracle-s7k-0.1.2-31.el5
osc-oracle-ocfs2-0.1.0-31.el5
Bug 13938125
When you create a configuration with virtual machines sharing an OCFS2 cluster file system on a virtual disk, severe I/O interruptions may occur. These may affect the heartbeating function of a clustered server pool and even cause Oracle VM Servers to reboot. Therefore, a shared OCFS2 cluster file system on a virtual disk is not a supported configuration.
Workaround: Use a physical disk or LUN. Make sure that the virtual machines in your configuration have shared access to this physical disk or LUN, and create the shared OCFS2 cluster file system there.
Bug 13935496
If the partitions for a virtual disk are not properly aligned to the OCFS2 block size, OCFS2 serializes I/O to the virtual disk image resulting in degraded performance. This only affects virtual disk images stored on OCFS2 repositories and does not apply to NFS repositories.
Workaround: Check for misaligned partitions in the virtual disk images and, if any are found, recreate the disk images with properly aligned partitions. To determine whether partitions are correctly aligned, use fdisk -lu to find the starting sectors of the partitions. Ensure that these are a multiple of 8 (512 byte sectors), which aligns them at 4KB, the OCFS2 block size.
Many recent operating systems often start the first partition in sector 2048, aligning it to 1MB, which works well for most storage RAID stripe sizes.
Bug 14355588
This section contains the known issues and workarounds related to any other issue that does not fit in the previous categories, for example, withOracle VM Utilities or the Oracle VM Command Line Interface.
MySQL performance within Oracle VM Manager is severely degraded on Oracle Linux systems that have been formatted with an EXT4 filesystem due to default mount parameters.
Workaround: Change the mount parameters for the filesystem where Oracle VM Manager has been installed. This can be done by editing /etc/fstab and changing the mount options to (rw,noatime,data=writeback,barrier=0,nobh)..
Bug 14624617
It is not possible to delete a vmdiskMapping from a running virtual machine using the Oracle VM Manager CLI. This means that it is not possible to eject an ISO that is presented as a CDROM device to a virtual machine.
Bug 15901598
Using unicode or multibyte characters for appliance or server names within an assembly is not supported within Oracle VM.
Bug 14624338
Using the BasicFiles configuration for Large Objects (LOBs) can result in the corruption of the Oracle VM Manager database. The BasicFiles LOB configuration has been superceded by the SecureFiles LOB, which offers better performance and more stability. Usually when a corruption occurs as a result of this configuration, a SQL Exception is returned with the following text:
ORA-01555: snapshot too old...
It is recommended that the database configuration is updated to avoid this bug in the future. To start using SecureFiles, instead of BasicFiles, the LOBs within all tables must be converted. Details describing how LOB migration should be performed are covered, along with a description of the benefits of SecureFiles, in the whitepaper available from:
http://www.oracle.com/technetwork/database/securefilesmigrationpaper-130440.pdf
Always ensure that your database is backed up regularly, so that it is possible to recover from a corruption event.
Bug 13873367, 17080032