Technical Notes

`Mono-spaced Bold Italic`

books        Desktop   documentation  drafts  mss    photos   stuff  svn
books_tests  Desktop1  downloads      images  notes  scripts  svgs
static int kvm_vm_ioctl_deassign_device(struct kvm *kvm,
                 struct kvm_assigned_pci_dev *assigned_dev)
{
         int r = 0;
         struct kvm_assigned_dev_kernel *match;

         mutex_lock(&kvm->lock);

         match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
                                       assigned_dev->assigned_dev_id);
         if (!match) {
                 printk(KERN_INFO "%s: device hasn't been assigned before, "
                   "so cannot be deassigned\n", __func__);
                 r = -EINVAL;
                 goto out;
         }

         kvm_deassign_device(kvm, match);

         kvm_free_assigned_device(kvm, match);

out:
         mutex_unlock(&kvm->lock);
         return r;
}
When trying to run puppet it would not start and the following error message was displayed:
'The field must contain an empty string or a valid URL.'
This was caused because the certname of puppet was using  valid.URL() instead of valid.FQDN(). This has been corrected so puppet will now run.

When a Node is being managed by, for example, by oVirt Engine, it should prevent the user from changing certain elements such as the Network configuration, as the network configuration is done by Engine/VDSM. Keys have been added to let the TUI know about:
- The management instance which is managing Node
- Important NICs on the system
- Pages to be locked
These changes help with Node management.

The hypervisor base image did not include the latest version of dracut (dracut-004-336.el6_5.2). When using auto install with fips=1 a kernel panic would occur. This has been fixed by adding the latest version of dracut.

Upgrading the hypervisor was failing with the following error message: "No module named puppet.puppet_page" This occurre because Specfile definitions were allowing autoinstall files for plugins to be placed onto the base image (where the plugins are not present). This is no longer possible, and autoinstall scripts for all current and future plugins should now stay in their appropriate packages where they belong, rather than the base image.

In previous upgrades cron was used to reboot the system. During the installation process the installer wrote a new crontab file, but the installer was not removing the crontab entry after an upgrade.This meant the system kept on rebooting. To fix this, the crontab entry responsible for rebooting the system is removed at boot, this ensures that the system is not kept in a reboot cycle.

When installing Red Hat Enterprise Virtualization 6.5 Hypervisor from a USB-DVD ROM, the installer would get a kernel panic. This was due to a missing xhci-hcd usb 3.0 module that affected certain machines in which their cd/dvd drive were connected through usb. This has now been added.

When upgrading from version 6.4 to 6.5, some of the existing configurations were not migrated across. The new codebase in 6.5 did not show previously configured values in the TUI because it relied on the presence of keys in /etc/default/ovirt to populate fields, and the old TUI did not set these values. Though previously configured services continued to work, their configuration could not be seen in the TUI. To fix this, a migration script has been added which checks the configuration files and adds keys to /etc/default/ovirt if necessary.

Previously the value of "Logrotate Max Log Size" could not be saved in "/etc/logrotate.d/ovirt-node". This was because the ovirt-node configuration file for logrotate had an
invalid syntax, which prevented augeas from being able to edit it. Now the syntax of the file has been fixed, and the value can be saved.

Previously, the storage volume space retrieved the default value for the data volume size from ovirtfunctions. This resulted in the data volume size being set as -1 on the first run and displaying properly while leaving the model size set to 0. Now, a default size of 0 is returned when running the TUI installer, so the volume is properly created.

When switching to the logging menu a blank page was being displayed. Previously the ovirt-node configuration file for logrotate had an invalid syntax, this prevented augeas from beeing able to edit it. Now the syntax of the file has been fixed.

The virt-who package, which provides an agent for reporting virtual guest IDs to subscription-manager, is now installed by default with the rhn-plugin for the hypervisor.

After the successful addition of the host to setup, the information for rhevm bridge was not being displayed in TUI. The TUI was displaying information for the em1 interface which was not relevant. To fix this the following keys were added to let the TUI know about:
- The management instance which is managing the Node
- Important NICs on the system
- Pages to be locked.

Previously, changing the iSCSI initiator name on the hypervisor console did not take effect until the hypervisor was rebooted. Now, the iSCSI daemon is automatically restarted when changes are made in the hypervisor console, so changes to the initiator name are persistent.

After registering a hypervisor to the Manager with the "automatically configure host firewall" option, the host's ICMP ports were blocked by the iptables rules. This caused the hypervisor to become non-responsive. host-deploy now configures iptables rules correctly after a host is registered to the Manager.

Previously when clicking the "Save" button to save the puppet configuration, it would clear /etc/puppet/puppet.conf. This was due to a problem in the code. When "Save" is clicked now, the puppet configuration information is saved in /etc/puppet/puppet.conf.

The kernel ordering of /dev/sdX devices is not necessarily deterministic. When booting the Hypervisor using PXE with the storage_init parameter set to a /dev/sdX device it was therefore possible to accidentally delete data from Red Hat Enterprise Virtualization storage domains.

Additional safe guards have been added to check that the device specified in the storage_init parameter does not contain a storage domain. Installation is blocked where this is the case. To use the storage_init parameter in a deterministic manner specify devices using the paths presented under /dev/mapper/ instead of paths of the form /dev/sdX.

When the hypervisor got approved on the manager it changed its state to Non Responsive. This was caused by the port being opened but the change not being saved saved. This lead to a situation where the iptables service got restarted and overwrote/closed the previously opened ports. Now the configuration is saved (and persisted) after a port is opened.

edit-rhevh could not edit an iso with SELinux enabled. This had already been fixed in a later build.

After configuring NIC with IPv6 DHCP and Vlan mode, a Vlan bridge could not be established and Vlan information was missing. This problem had already been fixed in a later build.

Previously on the network interfaces section of the hypervisor TUI, the boot protocol displayed as "none" when the IP configuration was static. Now, the boot protocol field displays as "static" when a static IP configuration is present.

autoinstallations with activation keys for sam failed to register properly, when using activation keys with a sam registration rather that just user/pass. This was caused because the --auto-attach option was being used. It is no longer used in conjunction with activation keys, so registration is successful.

The /usr/share/snmp/mibs directory was not writeable under the read only root, therefore custom snmp mibs were unable to be installed.The /usr/share/snmp/mibs directory is now read/write which means files can be placed under /usr/share/snmp/mibs now.

Previously, the RHN tab on the hypervisor setup screen continuously showed an error message even when the required CA and URL information was already provided. This has been fixed, and RHN information can be entered successfully.

When the hypervisor was booted in EFI mode, the expected behaviour was for a boot message to replace its previous line. Instead, the messages were appended to each line. This update shortens the GRUB EFI menu to under 54 characters, to prevent boot messages from overlapping.

The log rotation configuration stored in the /etc/logrotate.d/syslog file was not correctly set up to rotate the /var/log/messages file. As a result the /var/log/messages file was deleted instead of rotated when the log rotation cron job ran. The /etc/logrotate.d/syslog file has been updated to correctly configure log rotation for the /var/log/messages file.

Due to a bug in the upgrade code, grub was not correctly installed on upgrades performed from the manager. To fix this grub installation paths were updated and now properly handle the manager initiated upgrade and now upgrades successfully.

A problem was caused by two different hypervisor images existing and using the same CPE information.
This meant the CPE information could not be used to distinguish between the different hypervisor images
To fix this, the CPE information of the base image was adjusted.This means the hypervisor base image can now be distinguished from other hypervisor images because it has its own CPE name.

Previously for automatic installation, the link state of a device was determined in one pass. This could lead to problems when the hardware took several minutes to initialize, or set the link state to UP. Now five passes are allowed to find a device with a link before it falls back to the TUI installation.

Previously, EFI's kernel and initrd were symbolically linked. This caused installation on a UEFI machine to fail after rebuilding the hypervisor ISO with the edit-node command. Hard links are now used for the kernel and initrd in the EFI/BOOT directory, so this error no longer presents.

When the hypervisor was booted in EFI mode, the expected behavior was for a boot message to replace its previous line. Instead, the messages were appended to each line. This update shortens the GRUB EFI menu to under 54 characters, to prevent boot messages from overlapping.

The hypervisor was missing kernel modules and crashing kdump over a network. this was due aggressive stripping of file system contents while trying to limit the size of the live file system image. This meant the "igb" network card driver module was not functional in the kdump environment as the "ptp" and "pps_core" modules on which it depends were not available when the initial RAM disk for the kdump environment was generated.

To fix this the stripping process was adjusted so as not to strip out these modules. Any hypervisor kernel crash dumps can now be captured over the network on systems that require the "igb" network card driver.

The "nomodeset" kernel parameter, which was previously missing, has now been added to the list of Basic Video options for EFI boot.

RHEL6 supports a FIPS compliant mode enabled via fips=1 boot parameter and the installation of the dracut-fips package. When fips=1 is added to the cmdline Booting fails with a kernel panic
To fix this the logic was improved to handle the hypervisor case better, which means the hypervisor can now be booted in FIPS mode.

A previous fix which populated an empty /etc/hosts file with default entries also overwrote an /etc/hosts file populated with custom entries. Now, when ovirt-node detects that the /etc/hosts file contains custom entries, the file is persisted upon reboot.

This feature is a new plugin based RHN registration page. This happened because the TUI was switched to a new framework and a rework of the Red Hat Network registration page was necessary to use the new urwid framework.

Previously, installing the hypervisor with the vlan parameter set to vlan=vlanid resulted in a kernel 'panic'. Setting the vlan boot parameter to the format of vlan=vlanid: can be validated within dracut, so the hypervisor installation succeeds.

When booting the iso the kdump services failed to start which caused confusion. This has been resolved by adding an ovirt-kdump service wrapper to prevent confusing errors.

On the cim configuration page, there was a minor typing error "CIM Configuation", missing the character "r" in the word "configuration". This was fixed un the UI redesign.

The Nic vendor information was not being correctly retrieved. This happened because some of the assumptions made about a devpath were incorrect. Now the device PCI address are taken from the back of the devpath, which is a more robust method. Also, the parsing of the lspci output has been improved. The  Nic vendor information displays correctly.

The following exception occurred in the "other device" page, even if the correct device path was used: "Device' object has no attribute 'translate_device_name". This was fixed in a later build and now works with the correct device path.

If a user used ssh remotely to register the host once on Red Hat Network, they could not register the host again unless they logged out and logged back in. This was caused by overly strict validation procedures. The checks in the validation procedure have been relaxed to allow the user to register the host without having to logout.

iSCSI booting in UEFI mode is now supported.

A bug created bridges or other network devices with characters that could not be interpreted by the shell, causing debug messages to flood the Network page of the hypervisor installation TUI. This bug has been fixed, and the debug messages no longer appear.

A version of the hypervisor was showing incorrect CPU information when checked using the "View CPU Details" button. The errors included showing the incorrect number of sockets or cores. It was found that hosts with multiple numa nodes could end up reporting incorrect information in the libvirt capabilities' tag.

This has been fixed through a more in depth parsing, and the correct information is now displayed.

Previously, attempting to use the mem_overcommit kernel argument to set the swap ratio caused installation to fail. Now, specifying mem_overcommit calculates the swap size for installation. If the swap size is specified in the storage_vol parameter and the mem_overcommit argument is also provided, the hypervisor is installed using the swap size defined in the storage_vol parameter.

It was requested that the TUI could fill the whole screen on a monitor. The new TUI can fill the entire monitor screen in both basic mode and regular mode.

The manager is having support for ipv6 added, so there was a need to expose the ipv6 functionality into the TUI. The IPv6 function is enabled in the TUI, and the NIC can get IPv6 addresses - already available in version: rhev-hypervisor6-6.5-20131017.0.iso.

Dell PowerEdge Servers ship with an iDRAC Service Processor that is used to manage the system remotely rom the host operating system on Red Hat Enterprise Virtualization deployments. The IPMI drivers are not loaded during system startup like they do on Red Hat Enterprise Linux 6 deployments (by the /etc/init.d/ipmi script). Having the drivers load automatically on system startup is desireable where auto load of drivers on detection of an IPMI capable device would be ideal. The system now installs the ovirt-node-plugin-ipmi ipmitool OpenIPMI and OpenIPMI-libs package. Then clean installs rhevh-6.5-20131024.1.0.iso on DELL PowerEdge R510,login rhev-h.

The hypervisor was reporting "Low disk space for /var/log", and "No space left on the device" messages on TUI, even when there was sufficient space. This was caused by the the logger writing errors and warnings to stderr.

With the new TUI, warnings and errors are written to /var/log/ovirt-node.log but not to stderr, which means these message won't appear incorrectly anymore.

Previously the Hypervisor upgrade process used the already installed codebase. If there were issues in the existing upgrade script they would cause upgrade failures, and users would have to use installation media to correctly upgrade. The new process uses the new image's codebase to perform the upgrade.

The igb kernel module did not load after inserting the MegaRAID kernel module in a hypervisor iso image. This was caused by a missing i2c driver which has been added: /lib/modules/*/kernel/drivers/i2c. the kernel module now loads.

mcelogd is now included included in the hypervisor. This was needed to check and configure alerts regarding memory/hardware failure in the hypervisor. 

It is run using:
1. Run #rpm -qa|grep mcelog
2. Run #rpm -ql mcelog
3. It has the mcelog file in /var/log/ folder.

When running 'libguestfs-test-tool' the following error message was generated: "$ libguestfs-supermin-helper: failed to find a suitable kernel. I looked for kernels in /boot and modules in /lib/modules. If this is a Xen guest, and you only have Xen domU kernels installed, try installing a fullvirt kernel (only for libguestfs use, you shouldn't boot the Xen guest with it)". This occurred because the kernel and initrd are not located in /boot. To fix this the kernel has been symlinked into /boot. The libguestfs-test-tool now runs without error.

The dracut patch could be applied properly when trying to install/upgrade/downgrade the ovirt-node package. This was caused by an out of date patch. A new patch has been developed to fix this problem.

Previously, a problem occurred where Satellite could fail when registering Red Hat Network satellite 5.5. This was caused by trying to configure the 'Profile Name' using less than three characters. This has been resolved by providing validation that only allows an empty string or a string three characters or more to configure the 'Profile Name'.

Previously the installer accepted an empty password but failed to complete the installation. This happened due to an incorrect password check. Now the password is checked correctly and the user get's feedback about different problems such as password length. Additionally the installer only allows the user to carry on when a valid password is given.

Auto action for RHN/puppet could be executed even if didn't configure them in auto install progress. This was  caused by a combination of Specfile definitions which were allowing autoinstall files for plugins to be placed onto the base image, and the autoinstall script for the Red Hat Network plugin attempting to configure Red Hat Network, regardless of whether or not the requisite arguments were passed. The fix makes sure that Red Had Natwork is only configured if the necessary kernel arguments are present.

Users were able to configure SNMP without having to enter a password. Previously when users selected 'enable snmp' without using a password, it would run 'snmp transaction' and enable 'snmp service'. Now if the user selects 'enable snmp' without using a password, the system prompts "Unable to configure SNMP without a password!"

The cim '%post scriptlet' was failing. This was caused by the snippet used in the '%post scriptlet' to pick the UID erroneously closed an expression by using a backtick instead of an apostrophe. This has been fixed so the '%post scriptlet' works correctly.

Device information was going missing after moving the cursor to another device. This was because the the displayed device information wasn't limited in size,which could lead to a situation where one line occupied two lines This has been fixed in version 6.4 where the whole device information is shown in one line.

The hypervisor did not offer a way to show disk diagnostics in the user interface, and users were required to drop to a rescue shell in order to check disk status.

There is now a "Diagnostics" page which allows users to check the output of:
multipath -ll
fdisk -l
parted -s -l
lsblk

From within the user interface, and the output is displayed as part of normal user interaction, allowing for end-users to verify the status of Fiber Channel or iSCSI targets easily.

There was no way to stop the puppet service. A disable_puppet function has been added in puppet_page.py to stop puppet service and remove related configuration information from "/etc/puppet/puppet.conf".

A request was made to disable 'Apply' until there was a valid input. If nothing was selected or there was no data input, the user user could not use 'Apply'. This request has been addressed with the new TUI, when nothing is input or selected, and 'Apply' is selected, a pop-up a window is displayed with the message. "There were no changes, nothing to do".

The '%post scriptlet' was failing, which meant that the cim user did not get created, and the password could not be set in the TUI. This was caused by a code error and is now fixed, meaning the cim user is created and the password can be set.

The grub was missing the backup version after upgrading in uefi machines. This meant that backup boot entries were not created in the grub config. The backup entries are now correctly created.

A new feature has been developed, which is to create a new installer based on the new TUI code-base. This has been done because the new codebase is more flexible and supports features like response to screen resize request and mouse input. The installer has been migrated to the new TUI codebase and utilizes all it's features

The hypervisor-6.4 home page showed a blank screen when scrolling to the Red Hat Network option. Also the screen took a long time to load. The blank screen was caused by a missing "Loading page" routine, and the slow loading because the program was checking for a Red Hat Subscription Manager connection. A patch has been developed that shows the page loading rather than displaying a blank screen, and loads faster as it no longer checks for a Red Hat Subscription Manager connection.

--no tech note required

Previously we used certain bond options to setup "bond". The name "bond" was being recognized by dracut and causing kernel panic with
"dracut:FATAL:No netboot configured, bond is invalid". This was caused by having duplicate names in the system. To fix this "bond" has been renamed  as "bond_setup" to solve conflicting names. This information is used to setup bond during an auto-installation.

An exception has occurred when kdump_ssh to different host

The incorrect information for VLAN_ID was being displayed after nic was configured with "vlan". This was because the 'bootif' and 'vlanid' were only set correctly when IPv4 got configured. This is now longer the case, that way IPv6 can be configured independently.

When attempting to register to the Red Hat Network(rhn+satellite+sam) using auto install, registration failed when using Red Hat Network parameters. This was caused because Red Hat Network was not included in the new autoinstaller parameters. It has since been included, so registration using Red Hat Network Parameters is now possible.

The initial implementation of the IPMI plugin for the hypervisor did not require OpenIPMI in the RPM specfile or start the service by default. It now does so, and IPMI works as expected.

Some information in ovirt.log was being missed due to using an incorrect function, logging.info("") this has been rectified using the correct function logger.info("") to record all log information in ovirt.log.

Previously when selecting the disk that wasn't managed by multipath as 'Boot disk',Auto install failed with the error message "ERROR-install-Grub Installation Failed" A patch was written to fix this.

Previously, attempting to view the details of installed plugins returned an exception. This was a side effect of restructuring the manifests into directories. The edit-node script now removes manifests after performing all operations, so the plugins can be viewed without error.

When trying to auto-install the hypervisor with 
with "BOOTIF=eth0 storage_init=/dev/sda,/dev/sdc;/dev/sdb iscsi_init=/dev/sdc iscsi_target_name=iqn.2008-09.com.redhat:rhevh iscsi_install iscsi_server=xx.xx.xx.xx.x
storage_vol=::::::200:200 swap_encrypt=Swap;Swap2 firstboot" parameter, AppVG was not being created. A patch was written to fix this command and create AppVG when it is run.

Due to differences in file locations between Fedora and RHEL 6, the Puppet plugin patched an incorrect file location for Facter, and did not parse output properly, so a non-zero exit code was returned.

Previously, booting rhev-h failed with a kernel panic during an enabled-iSCSI boot on Dell smbios 2.6+ machines. 

Booting rhev-h no longer fails with a kernel panic during an enabled-iSCSI boot on Dell smbios 2.6+ machines.

Non-"basic video" boot entries were being booted in basic video mode. This was caused by the nomodeset kernel argument being  added to non "basic video" entries. This was fixed by ensuring the nomodeset kernel argument is only used in the "basic video" boot entries. This means that Basic Video boot entries boot with the nomodeset kernel argument, all others don't.

Incorrect boot entries were being displayed to the user. This was because incorrect boot entries remained in the grub configuration.This was fixed by removing the incorrect boot entries, so only valid boot entries are shown to the user.

Added a feature to show pages in the TUI
1. Press `F2`
2. [root@localhost admin]# cat /var/log/ovirt-node.log | grep 'Current page'
3. Output appears as follows:

2013-11-11 08:16:42,112       INFO Current page is 'Network'
2013-11-11 08:16:42,303       INFO Current page is 'Status'
2013-11-11 08:17:05,004       INFO Current page is 'Status'
2013-11-11 08:17:05,028       INFO Current page is 'Status'

Also, pressing F12 will reset the cursor to the original position.

When trying to run auto install ovirt-node-iso with the follow parameters:
BOOTIF=bond0 storage_init=/dev/sda bond=bond0:eth0,eth1 firstboot, the system was entering the TUI install process not the auto install process. This was because BOOTIF=bond* was not supported. This has been fixed so it is supported, and auto install ovirt-node-iso with the follow parameters:
BOOTIF=bond0 storage_init=/dev/sda bond=bond0:eth0,eth1 firstboot works as intended.

When moving to the device page and password page there was a noticeable delay in the pages loading. This problem was solved by doing the storage discovery before the installer UI is started, that way there is no delay when switching to storage related pages.

Using "livecd-iso-to-disk --format --efi *.iso /dev/sdx" to make usb media, then auto installing the hypervisor  into a uefi machine created the following error: "Installing Bootloader" with error info:
IOError:[Error 2] NO such file or directory: '/liveos/efi/EFI/redhat/grub.cfg'. This was caused by a missing directory and configuration fille, which have been fixed in a patch to allow this procedure to work.

There were two module imports that had name collisions.This meant that the keyboard layout would not persist because the keyboard config didn't get loaded during boot. There is now a different module import, which allows the keyboard config can now be loaded at boot.

After installing the iso created by 'edit-node', the system was reporting an error when entering the 'plugins' page in the TUI. The cause of this was 'edit-node' didn't check whether or not the paths passed for the repository and RPMs were actually repository files and RPMs, so it would attempt to add arbitrary files to the image. These checks are now in place, so the image is built, and can be installed correctly.

The 'manpage' of 'edit-node' said, multiple packages could be installed using a comma to separate each item. However, this format was not supported. This occurred because 'edit-node' was not parsing the correctly. The parsing has been fixed, so now multiple packages can be installed by listing them with a comma separating each item.

Previously the iSCSI initiator name (if given or random) was set after the installation, but in some cases the target only allowed discovery if the initiator had the correct name.
With this patch the iSCSI initiator name is set (if it is given) before the target discovery is initiated.

The run edit-rhevh and the part of "Creating Manifest" was taking more than 10 minutes to complete. This issue was fixed in rhev-h 6.5:ovirt-node-tools-3.0.0-6.1.el6.noarch

When creating "bond" 1 to 9, 'bond' zero would be generated automatically in TUI, which caused a problem as it had no member. This was fixed in a later build, onlybonds 1 to 9  are created in TUI.

Profiles on the kdump page were still being shown after "() Disable" was selected to disable the profile. This has been fixed so that when "() Disable" is selected now, the kdump profile is deleted from TUI and "/etc/kdump.conf".

When a user configures kdump over SSH, saving that configuration restarts the kdump service, which attempts to reach the target. In the process, kdump may prompt the user to enter a password to set up authentication.

Pressing Ctrl+C to abort entering a password generated an uncaught KeyboardInterrupt to the hypervisor user interface, which brought the user back to a login prompt without finishing kdump configuration.

Ctrl+C is now properly caught, and pressing it returns the user to the hypervisor kdump configuration page.

When trying to view the ovirt.debug.log from the support menu, the system would generate the following error: "No such file or directory". The log for /tmp/ovirt.debug.log has been removed from support menu, therefore it can no longer be accessed.

Previously every error message which appeared on the kdump page was not discarded after it was shown, so error messages were repeated even if when they did not apply. Now, kdump error messages are discarded after they are shown.

When switching profile from spindown-disk to any other profile,the system would generate a series of error messages, for example: "/etc/tune-profiles/functions: line22: echo: write error: Invalid argument". These were invalid messages as the profiles were still switched. This has been modified so these error messages no longer appear.

The Common Platform Enumeration (CPE) for the hypervisor is now determined based on the $VERSION parameter stated in the version.ks file.

Previously the command  "less" wasn't working correctly. It produced a sequence of garbled characters when viewing the  log or another file and it was hard to exit the current view page. This has been rectified using alias less "cat $1 | /usr/bin/less", so "less" works correctly now.

When using edit-node install ovirt-node-plugin-puppet into ovirt-node-iso, entering into the puppet page and enabling the puppet service, then configuring the puppet server
and puppet certificate name, clicking save and run, an error occurred while configuring puppet in the transaction puppet page:

"An error occurred while applying the changes:
__init__() takes exactly 2 arguments(3 given)"

This was occurring because ovirt.node.utils.fs.File() has changed arguments, this mode is no longer set, but the puppet_page was doing so. It has been changed to the new interface, and now works correctly.

After upgrading the hypervisor on a legacy (non-EFI) machine, the /liveos directory was unmounted, so users could not access the backed up version of the hypervisor. Now, a non-EFI installation reads grub.conf for changes before it unmounts the /liveos directory, so the hypervisor backups are still available after upgrade.

cim and snmp plugins are installed by default during the image generation, but the screens were not showing on the TUI. Previously, the Python files were defined as _SCRIPT in automake. This has been changed to _PYTHON so they automatically get compiled, and the compiled files added to the RPMs so image-minimizer will leave working plugins. The screens now display in the TUI.

The status page always showed the IPv6 link address after configure nic with dhcp/auto of the IPv6 setting. This is related to independently configuring IPv4 and IPv6. Previously the 'bootif' and 'vlanid' were only set correctly when IPv4 got configured. This is no longer the case, they have been changed so that IPv6 can be configured independently. The information for IPv4 and IPv6 is now displayed correctly.

This feature enables iSCSI software root support. By enabling a small footprint installation on the physical machine. Root/HostVG can be located on an iSCSI target enabling usage of small drives for installation.

When a ssh location was entered on the kdump page, there was no validation to check on the authenticity of the location. Validation for the authenticity of the ssh location
when configuring kdump with ssh has been added, and if the location is invalid, the following message is displayed: "The field must contain an empty string or a valid SSH Address".

There was a prompt that read "Hit return to select layout" on the  installer keyboard page. However there is no return key, only the enter key. The prompt has been changed to read:
"Hit Enter to select a layout", to make it less ambiguous.

Previously the 'progress page' allowed pressing the F2 key to drop to shell. If the F2 key was pressed repeatedly when starting the hypervisor installation, the action could cause the hypervisor installation to fail. The F2 key is now disabled when starting the hypervisor installation, after the installation is finished, the F2 key is enabled again.

The "PRODUCT_NAME" argument wasn't assigned in other device page, and should have been assigned as "RHEV Hypervisor". Also the prompt "please select the disk to use for booting PRODUCT_NAME" was inaccurate, it should have read as  as "please enter the disk to use for booting RHEV-Hypervisor".

The following changes were made:
1. "PRODUCT_NAME" was changed to RHEVH
2. The word "enter" has replaced "select"

The Keyboard Layout Selection page did not show the existing keyboard configuration.A KeywordLabel has been added to show the currently active keyboard layout.

It wasn't apparent which format of hostname is valid. The hostname as FQDN format was not accepted so clarification was needed. There is now added validation (to comply with RFC1035) to check the length of each field, as well as the entire hostname to make sure it's not too long

Incorrect information was being shown when performing transactions on the security page, for example, performing "Enable SSH password authentication" also showed "(2/2) configuring SSH AES NI" when this wasn't performed. Previously the transaction elements appearing in the  transactions used on the security page weren't connected to the changes done in the TUI. Now the same transaction elements are used in the transaction which has eliminated these errors.

Previously when configuring AES-NI, it would cover the original /etc/profile file. It also could not delete the values in /etc/profile when deleting the AES-NI bytes in UI. This has been fixed by setting AES-NI correctly when unset in the TUI and also fixing writing to /etc/profile.

Previously the footer was sometimes hidden due to missing attributes (which lead to the text being drawn in the same color as the background). This was sometimes misconstrued as missing prompts which wasn't the case. This has been fixed by making sure only the default text color is used.

Previously it could happen that servers which were removed in the TUI didn't get removed in the actual config files.
This is now fixed by removing all existing servers and writing the new ones.

Previously there was no indication to the user when they set a weak password in password_page. Now the migrated password_check function is used in the pasword
wigets to return informative messages to the user, for example: 

"You have provided a weak password! Storage passwords contain a mix of uppercase, lowercase, numeric and punctuation characters. They are six or more characters long and do not contain dictionary words. 
Reason: it is WAY too short"

The server port used to be able to begin with the 0, for example 0000514, which could cause confusion on the screen. This has been fixed, if the first number of a port is specified as 0 the following warning is displayed: "The field must contain a port in the bounds [1. 65535]".

On the password_page and upgrade_page the user needs to set   the  password for the admin user, but the prompts read  "Password for local console access" which was confusing. This prompt has been changed to "Require a password for the admin user" to get rid of conflicting terminology.

If multiple admin sessions were opened, changed values which
relied on Augeas (DNS and NTP settings principally) did not
reflect changes between them without quitting the UI. This has been fixed by forcing a reload of cached Augeas values so it properly updates.

When on the Remote Storage page and configuring "iscsi initiator name", an error was generated in the UI saying "Invalid Data" if it was configured with an empty string.
The iscsi initiator name doesn't allow an empty string to be used anymore, also setting "Rsyslog address" and "Netconsole address" with an empty string are not permissible and therefore no errors are produced.

Currently the hypervisor can support configure logrotate maxsize from the TUI, but it doesn't support it for automated installation. This is now supported. Support was added for setting logrotate_max_size in cmdline args:
logrotate_max_size=number. This information is then used to set logrotate_max_size during an auto installation..

The man page for 'edit-node' contained parameter information  that was redundant:
[-o output
[-k Kickstart file
[-s script.sh
[-t tmpdir
[-r releasefile
[-b builder
This has been removed from the man page.

Previously the IPv6 addresses were shown as ordinary text. To highlight that they are IPv6 addresses they are now shown according to RFC 2732 - contained  in [] brackets.

Previously lsblk diagnostic was not working, the following error message was being generated: "ascii' codec can't decode byte 0xe2 in position 169: ordinal not in range(128)" when selecting lsblk in the diagnostic screen. This was caused by the UI not handling the encoding of the console getting changed. This has been fixed by making sure the UI is always using UTF-8. lsblk now works correctly.

Previously when users selected 'enable cim' without a password, the cim transaction and enable cim service would still be run. Now if the user selects 'enable cim' without a password, it is not run and the system will prompt: "Unable to configure CIM without a password!".

Previously Node conuldn't discover all allowed network layouts. This is now fixed and additionally many unit tests were added to prevent regressions.

The reset button at the bottom of the Puppet plugin was not present. It is now shown.

Previously ovirt-node-upgrade needed to reload the 'install' and 'ovirtfunctions' modules from the new image to use new 'install.py' and 'ovirtfunctions.py' patches for upgrading the hypervisor. The ovirt-node-upgrade  was still using the old 'ovirtfunctions' module from "ovirtnode.ovirtfunctions".
This has been fixed with: ovirt-node-upgrade:reload ovirtfunctions modules from new image.

Previously the term "Netmask" was used incorrectly for the IPv6 prefix length. The name is now corrected and additionally only an integer is allowed as the prefix length.

Previously the system did not show a dialog before entering the rescue console. The following indication has been added:

Rescue Shell  
Making changes in the shell is unsupported. Do not use this without guidance from support representatives

After creating a bond using "Create Bond", for example bond0, moving to the network page and selecting the bond for configuration, the system would generate the following exception: "Unknown network interface: 'bond0'" This was caused because previously Node could not discover all the allowed network layouts. This is now fixed, so 'bond' can be created, selected, and configured.

Previously the bridged layout could be set in the NIC details dialog, but it couldn't be switched back to the direct (bridgeless) mode. Now the bridged mode can be turned on and off as expected.

Previously the scripts of /etc/ovirt-config-boot.d would  be executed no matter whether it was auto install or not. A new statement "if is_auto_install()"has been added to make sure these scripts can only be executed during auto install.

The old installer process hypervisor 6.4,displayed a "Cancel or Reboot" prompt when the "Quit" button was clicked.Cancel returned to the initial screen, and  reboot rebooted the system. This was missing in hypervisor 6.5. The "Cancel or Reboot prompt has been added back into the latest version.

This feature enhances the password tool usability. Previously the usability of the new password tool was suboptimal so some changes were made.  

It provides a command list in the initial text and also tells the user the current state of SSH passwd auth

Possible commands:

admin  Set the password of the user 'admin'
help   Show this help. [`help --all` to show all available functions]
q      Quit this tool
quit   Quit this tool
root   Set the password of the user 'root'
ssh    Enable or disable the SSH password authentication

[INFO] SSH password authentication is currently enabled

It also provides overrides for some commands:
admin to set admin password
root for set root password
ssh for set ssh password authentication
q for quit

Previously, specifying None at either end of the bounds for a
number validator would return a message which showed that it must be within [0, None] instead of infinity. This has been changed to show [0, Infinity].

The system was allowing ssh/NFS to be used as a KDump target without a configured network

This meant that the KDump configuration failed because no network was available when the user tried to configure a SSH or NFS location.

This has been fixed by only allowing the KDump ssh/nfs targets to be configured when the network has been configured beforehand.

This means that the user does not run into a situation where the kdump configuration fails because of an unconfigured network

Previously a NIC was only assumed to be configured if a bootprotocol was given. But this is wrong for the static IP cases when no bootprotocol but an IP address are set. The symptoms were that the NIC appeared as unconfigured in the network page, when a static IP was set.The logic has been updated and a NIC is assumed to be configured when a
bootprotocol or an IP address is set, so the NIC appears as configured in the network page.

It was possible to run setup in rescue mode, invoking the TUI menu. This has been fixed by adding the following prompt in rescue mode, after run "setup" is entered: "The TUI cannot be used in rescue mode, Please reboot without rescue to configure/install".

An error in the UI was being reported when selecting Keyboard and Performance through ssh selecting wireless network. This happened because the UI encountered difficulties when the encoding of the console got changed.Now the UI always uses UTF-8, so this error no longer occurs.

Previously when configuring the NFSv4 domain (for example domain1), it would report an error "No such file or directory" The code has been modified and this no longer happens.

Auto installing the hypervisor failed when using the "storage_init=ata:ALL" parameter.Previously both ata-* and scsi-* were parsed as /dev/sda. So when executing "find_disk ata ALL", it would return "/dev/sda,/dev/sda", when it was only meant to return "/dev/sda". This was fixed by making the"for loop" stricter so only "/dec/sda" is returned. It now correctly installs using the "storage_init=ata:ALL" parameter.

The netconsole and lldpad services were set to start automatically on hypervisor boot, but both failed to do so.

lldpad failed to start due to a read-only directory lldpad expected to be read/write.

netconsole failed to start because the configuration was missing or invalid.

Now, netconsole only starts if it has been configured, and lldpad is able to write /var/lib/lldpad, so it starts as expected.

When values for kdump are entered which cause restarting the service to fail, the hypervisor should revert to the previous configuration.

Previously when configuring the admin password on the Security page, the system would confirm mismatching passwords. For example, entering 'test one' in the Password field and 'test' in the Confirm field would configure the password as correct when Save was clicked. The validation has been changed so that mismatched passwords are no longer accepted and an error message is displayed: "Passwords Do Not Match", also, the Save button is disabled.

Previously, pinging while the network wasn't available or pinging an unreachable host lead to the ping page freezing.
The code has now been refactored and simplified to prevent this happening.

An exception occurred when viewing "lsblk" diagnostic utilities in the Diagnostic page: 'ascii' codec can't decode byte 0xe2 in position 142: ordinal not in range(128)' This issue was fixed in a different build and ovirt can output the correct information when viewing all diagnostic utilities in the Diagnostic page.

The monitoring page should not have been in the git repository and has been dropped from rhev-hypervisor6-6.5-20130725.0.el6.

Libvirt was "reading" an incorrect number of sockets/cores.A version of the hypervisor was showing incorrect CPU information when checked using the "View CPU Details" button. The errors included showing the incorrect number of sockets or cores. It was found that hosts with multiple numa nodes could end up reporting incorrect information in the libvirt capabilities' tag.

This has been fixed through a more in depth parsing, and the correct information is now displayed.

A warning message used to appear indicating that the field must contain a valid URL, even if a valid URL had been entered. Previously the validation for the URL required the path of the URL. This was incorrect, it needed validation for the scheme/netloc of the URL. The validation for the scheme/netloc of the URL is now active, validation for the path of the URL has been removed, eliminating the warning message.

When performing an Auto Install of the hypervisor with the following parameters:

BOOTIF=eth0 storage_init=/dev/sda,/dev/sdb iscsi_init=/dev/sdb iscsi_target_name=iqn.2013-02.com:hadong iscsi_install iscsi_server=10.66.11.102:3260
firstboot

The Auto Install succeeded,but when booting the hypervisor, then entering the grub menu and pressing the "e" key to edit the kernel the command line failed . This has now been resolved, previously, splash.xpm was copied to /liveos in all cases. Now it is put on the proper boot volume if it's an iSCSI install.

The system showed 'configure netconsole' successful when it was not. This has been rewritten and now works correctly, and new messages have been added to provide the user with more information when performing this process:
-  If the netconsole hasn't been restarted, the user is notified.
- In the case where netconsole doesn't restart, a message is displayed reminding the user to check whether the host is resolvable or not.

If the NFSv4 Domain configuration was not disabled when the network was not up, error messages were reported, for example -An error occured while applying the changes - Command '['service','rpcidmapd','restart']' returned non-zero exit status 6". To stop these error messages being generated, the NFSv4 Domain configuration is now disabled when the network is not up.

If one NIC was configured as a bridge, the other NICs also appeared to be configured as a bridge. This was caused by a configuration entry being used incorrectly.

The configuration entry is now interpreted correctly, so the bridge configuration is only shown for the bridge configured NIC and not for all of them.

If 'edit-node' was used without an option it generated failed with a string of Trace-backs. Now if 'edit-node' is used without an option the following message is generated:
"A manifest or editing option must be specified, use --help for more information"

Previously, clicking the "Flash Lights to Identify" button to identify which nics belong to "bond" didn't work. The code has been rewritten and now if the "Flash Lights to Identify" button is clicked, the nic "em1" will flash lights for a short time to show which nics belong to "bond".

Previously the storage related pages in the installer loaded slowly. This was caused by repeated calls to a function that performed slowly. This still happens, but the value, which does not change, is cached to speed up operations, meaning the pages load faster.

An IPv6 address could be entered for a netconsole server in the hypervisor, but netconsole does not support IPv6. IPv6 addresses entered are no longer valid.

The hypervisor allowed IPv6 addresses to be entered as kdump targets, but the kdump service does not support IPv6. Now, the hypervisor marks IPv6 addresses as kdump targets as invalid.

Incorrect plugin information was being displayed on the plugins page when multiple packages were installed. This was caused by a flaw in the program logic which has been rectified, so that the plugin information is now correctly displayed.

Previously, when using edit-node to install multiple packages resulted in the iso file name being overwritten with a new, longer, complex name. This has been fixed so that when using edit-node to install multiple packages, the iso file retains a simple name.

No confirmation dialog appeared when exiting the hypervisor TUI. If the user had made a mistake and didn't want to exit, they had no opportunity to stop the exit process. A dialog box now appears when the user is exiting, and prompts them "Are you sure you want to quit? Yes/No" allowing the user to confirm their choice.

The device name of the second hard disk wouldn't display in the TUI installation process. This has been fixed so that All device names(local disk/iSCSI/fc/usb) show in the TUI installation process.

lldpad would not start because a path required by lldpad was not writeable. This has been fixed by making the path writeable and lldpad now starts.

An error message "Failed to read image" was shown in grub if installing the hypervisor via livecd. This problem was solved in version 6.4 and no longer occurs.

The 'Man' page for 'edit-node' has typing errors which meant
certain options were incorrectly documented. The errors in the 'edit-node' 'man' page have been corrected.

The title of the 'performance' page was inconsistent with the other pages, as the first letter of 'tuned' wasn't capitalized - 'tuned Configuration'. This has now been changed to a capital letter, so the titles of all pages now have a consistent appearance.

An error code was displaying in the ping test results page when the input host length was more than 64 characters. This no longer occurs as there is now a check for the length of each field as well as the entire hostname to make sure it is not too long. It adds a validator for length, which checks for total length of less than 256 octets and field lengths of less than 64 octets.

Configuration of Puppet in the hypervisor did not persist /etc/puppet/puppet.conf, and the entered configuration was not saved between reboots. Now, /etc/puppet/puppet.conf correctly persists between reboots.

The hypervisor was hanging for a long time if simultaneously operating the logging menu on local host and serial console. This has been fixed using a transaction lock (python-lockfile) so the hypervisor no longer hangs.

A shell option is now available in edit-node to perform manual operations on the hypervisor image while it is open for editing.This was done so end users can script any modifications necessary to their environment.

Previously  the profile name was not displayed in the Red Hat Network page. As the profile name is used to find the hypervisor in Red Hat Network, it would be beneficial to the user to have the profile name displayed. The profile name is now displayed in TUI as a reminder.

AES-NI and NFSv4 could show on the TUI in the hypervisor, but were missing in automated installation parameters. AES-NI and NFSv4 are now included in automated installation parameters.

When F8 was pressed and the journal(systemd) was selected, the system produced an error message: "An error appeared in the UI: CalledProcessError()". To fix this, the journal(systemd) has been removed from the F8 Dialog.

The validation process for the "Other Device" item still accepted a live device in ROOT/HOSTVG_STORAGE_PAGE.
Previously, there was no check as to whether "Other Device" input matched against a known live device or not, for example, USB boot. Now, when a live device is input into the device path of the ROOT/HOSTVG_STORAGE_PAGE, the following error message appears. "Can't be the same as the live device"

The edit-node 'man' page for uid/gid options was not very clear. This has been rewritten to make it clearer, and to make it easier for users to use these options.

Previously when changing uid/gid elements, the edit-node tools only supported changes to an individual user.
Now multiple users can be specified by using a colon delimited list.(--uidmod=USER1,
UID1:USER2,UID2 or --gidmod=USERGROUP1,GID1:USERGROUP2,GID2).

The hypervisor installer uses a size of -1 to fill all remaining free space for the data partition. This value is automatically set, but if a user changed it, they were unable to change it back to -1 and proceed with the installation. This is now possible.

The Red Hat Network page loads very slowly, so it was decided to move it to the end of the setup menu. This means that moving between the other pages is now a lot smoother.

There was an error on TUI upgrade page. It displayed "leftblank" instead  of"left blank". This has now been fixed so the proper text is displayed.

There was a validation problem configuring the iSCSI qualified name (IQN) . The name could be entered in an incorrect format, for example, with special characters included, and it would still be accepted. Validation has been added, so if the name is entered in the wrong format, the following error message is displayed: "The field must contain a valid IQN" and save is disabled.

The '--shell' option was missing from the 'edit-node' man page. It has been added now.

When installing the hypervisor on a local machine, and then using multipath -ll, there was redundant text appearing on the screen. This was confusing as it appeared to be an error message when it wasn't. It appeared to only occur when using the mkdir/etc command. The text no longer appears on the screen.

Previously, when using edit-node with "-n" or "--name", it was changing the iso file name. It would append a plugin or package name to the existing iso file name. A routine has been added to stop this, and now the iso file retains its original name.

Previously, if you set the domain to some value that is found
elsewhere in the idmapd.conf file, and then updated again, it would replace all occurences of the value in the file. This has been fixed by making sure to only edit the Domain line in idmapd.conf.

Install and install-plugin options are now enabled in edit-node.They are now enabled to provide the ability to install plugins or other necessary packages by end users

Some users had a problem updating the dns server IPs once the hypervisor has been joined to the hypervisor manager. The problem lay in the prompt generated by cat /etc/resolv.conf: 
"Please make changes through the TUI. Manual edits to this file will be lost on reboot". It failed to say that sometimes the Node is managed by the management server. To fix this the prompt has been rewritten: "Please make changes through the TUI or management server. Manual edits to this file will be  lost on reboot".

The hypervisor failed to install when multiple devices were used. The cause of this was that sometimes, the canonical name of a device wasn't used. The canonical name of a device is now used everywhere, so the hypervisor can now be installed successfully with multiple devices.

Previously the edit-node function was not properly cleaning up temporary directories when finished.It failed to remove dd-isoXXX dir during cleanup.This has now been fixed so these files are removed.

The hypervisor could support 'configure keyboard layout' from the TUI, but it didn't support it for automated installation.
This has been changed so the keyboard layout can now be configured during auto installation.

The Reboot of the hypervisor is now performed async. This is 
to eliminate any processes from killing themselves and aborting the reboot. A single process daemon is created to handle the restart instead of using cron. There is now a 
consistent reboot process which no longer relies on cron.

Previously when selecting two disks as HOSTVG, storage_vol_page, the system would only
calculate one disk size(installation.device.current). This is now fixed so that the system displays the size of all the disks selected by the user HOSTVG(installation.devices).

Using auto install hypervisor registration showed package updates. This was because when registering the hypervisor to the Red Hat Network, the package list would be sent and possible updates would be shown in the Red Hat Network page, rather than leaving those details out. Affected hypervisor nodes would show as having possible single rpm package updates when they would not be applicable. The hypervisor Red Hat Network registration now skips the package update in the profile. The registration now works without package lists being compared for updates.

Previously when users selected 'enable cim' without entering a password it would still allow the 'cim transaction' and 'enable cim' services to be run. Now if the user selects 'enable cim' without entering a password, the following prompt appears "Unable to configure CIM without a password!".

Previously it was possible to get into a state where the storage volume page was still flagged as invalid, even when all the partition sizes were valid. This has been fixed by 
re-triggering the validation of every invalid field when free
space is not negative, so it behaves as expected.

The Red Hat Enterprise Virtualization Hypervisor can now be configured through Puppet for ease of use. It also provides an alternative to basic PXE to deployment and configuration.

When disabling puppet by unchecking "Enable Puppet []", the following exception was generated:"TypeError("The type of OVIRT_PUPPET_ENABLED" This was caused by a logic problem in the code which is now fixed. puppet is disabled when the check bock is unchecked.

When viewing the Plugins > View Details page, and a manifest was selected to view, the hypervisor title and manifest details of installed plugins were overlapped. The location of the information popup has been moved so the title bar is no longer overlapped.

The "help Show this help. [`help --all` to show all available functions]" was being displayed twice. This was caused by a coding issue and has been rectified so it is now only displayed once.

When disabling puppet by unchecking "Enable Puppet []", the following exception was generated:"TypeError("The type of OVIRT_PUPPET_ENABLED" This was caused by a logic problem in the code which is now fixed. puppet is disabled when the check bock is unchecked.

Previously, using multipath-capable devices for the install
sometimes panicked the kernel. This was a problem caused by multipath.conf. With the addition of an option to 'dracut' to disable it, multipath support is removed during the install phase, but is left in the bootloader arguments for normal booting. This allows the processes to work as they are meant to.

Previously options that did not work appeared on the ssh restart page. The new TUI has fixed this problem.

Previously when users set the 'setup data_size' as "0" to disable the data partition, it would create a data partition with the remaining free space. Now when the 'setup data_size' is set to "0" the data partition is disabled.

The lock page remained static when an incorrect password was entered with no indication that the password was incorrect. This has been changed so a dialog appears with the message "The provided password was incorrect." and resets the password field when the dialog is closed.

A NIC was displayed as "unconfigured" even if it had a valid IPv6 configuration. This was caused by the IPv6-only configuration not being identified as a valid IP configuration

The IPv6-only configuration is now identified as a valid configuration and a NIC appears as "configured" when it has an IPv6 configuration only (and no IPv4 configuration).

Previously a comma was used to divide Rsyslog and Netconsole information on the status page. While functional, it was cramped so a separate line is now being used to display them to assist in readability.

Previously, output from the IPMI plugin was being displayed,
and gave the appearance of 'hanging' if enter was pressed. The output is no longer displayed.

When using SSH to access the hypervisor host, locking the system and focusing on the lock screen, The unlock text isn't in the centre of the screen. This bug has been addressed with new TUI rewrite. The unlock text is centred in the screen.

When installing RHEV-H via the TUI on iSCSI storage,there was no description of the storage details. With the new TUI rebuild it is now possible to select the configure storage tab to display storage details.

The system was hanging when entering the 'setup' menu in 'rescue' mode. This was because the TUI could not be used in rescue mode. Now when in rescue mode, after running 'setup', the following error message is displayed. "The TUI cannot be used in rescue mode, Please reboot without rescue to configure/install". Then, in the local host Press F2 and run setup.

The prompt in the puppet page "Puppet Certificate Name" was missing a colon. It now appears as "Puppet Certificate Name:"

Previously a redundant character H appeared on the log view page when the TUI installation failed. This has been fixed with the new TUI build.

When you set "Logrotate Max Size" to five digits, it was only displaying four digits. The new TUI displays more than five chars at a time.

Certain network cards would not display their vendor ids properly on the network page. A fallback mathod was created to make another attempt at determining the appropriate vendor for display purposes. The vendor name is now displayed correctly on the network page.

When uninstalling the hypervisor in a uefi mode machine, efi entries weren't being removed. Now after uninstalling the hypevisor in a uefi mode machine it removes the existing efi entries This can be checked by rebooting from uefi mode.

The man page of 'node-creator' contained a minor typographical error: 
man(8) node-creator man page man(8)
NAME
node-creator - prepare the kickstart recipe and build the LiveCD image
SYNOPSIS
node-creator  
The angle bracket symbols surrounding  are superfluous and have been removed.

A red screen was observed while clicking on the "Flash Lights to Identify" button in the 'network detail page'. The system now operates correctly and ends up on a setting process page.

The validation process for 'current password' is still displaying an error message 'Current Password Invalid' when  the cursor is placed in the current password field
and nothing is input. With the new TUI, when the cursor is put in the password field and nothing is input, the message 'You have not provided a new password, current admin password will be used.' displays below the input fields.

The installer thread could hardly be debugged because in some cases no backtrace was getting logged. Now a traceback is logged to help debugging.

When an operator didn't enter the current administration password in the installer, no indication was given that they needed to enter it to continue. A hint "You have not provided a new password, current admin password will be used" is now shown on the password page during an upgrade or re-installation.

A new upgrade utility is provided to use the python modules from the newer media. This is to prevent work-flow upgrade issues where problems are solved in a newer version. These issues were overcome by booting install media in previous versions.

When trying to rotate files manually, an error was generated on logrotate.status:

# logrotate /etc/logrotate.d/vdsm
error: error creating state file /var/lib/logrotate.status: Read-only file system
error: could not read state file, will not attempt to write into it.

A bug in the do_filp_open() function caused it to exit early if any write access was requested on a read-only file system. This prevented the opening of device nodes on a read-only file system. With this update, the do_filp_open() has been fixed to no longer exit if a write request is made on a read-only file system.

Users could not find help for uid/gid as the uid/gid editing options were missing from edit-node man page. These options  have been added to the man page.

Previously rpcbind raised erroneous notices about an non writebale /var/cache/rpcbind dir. even though the services were working correctly. A patch was raised to fix this and the messages no longer appear.

When entering the TUI of a hypervisor  and  attempting to move to the tab 'Plugins', an error appeared in the UI: "IndexError('list index out of range',)
Press ENTER to logout ...or enter 's' to drop to shell" The system now continues to the TUI on the plugins tab without any error.

When boot copy_dir_if_not_exist was called, it raised an exception which is logged in /var/log/ovirt.log. This was fixed in rhevh-6.5-20131031.1.0, ovirt-node-3.0.1-7.el6.noarch.

Files used by yum are currently not being cleaned up, leaving redundant files on the system. Some of the areas affected were:

/var/lib/yum/history
/var/lib/yum/yumdb
/etc/yum.repos.d

This has been fixed by disabling all yum repos in /etc/yum.repos.d rather than just fedora.repo and fedora-updates.repo, and then remove the yum database information.

When configuring a network on ovirt-node, the created network was implemented as a bridge type with no option to control it otherwise. NICs can now be configured without being the slave of a bridge.

VDSM now uses tuned profiles for hypervisor hosts. Upon hypervisor installation, the installation script changes the host profile to 'virtual-host', which optimizes host performance. Note that the tuning is done only once, so it is not suitable for dynamic systems.

A new re-sizable UI has been created that fixes a number of the limitations of the previous UI while maintaining a similar look and feel. The recursive elements of the old ui (no more max depth reached) have been removed, and some other features are - it is re-sizable dynamically, it has common infrastructure code base to ease new page development, and it retains a similar look and feel to the old UI.

When running an auto install, the temporary password was logged in the ovirt.log. This has now been altered so temporary password no longer appears.

Virt-viewer now allows the user to do arbitrary monitor arrangement when going fullscreen. This allows for greater flexibility to permit each individual display to go fullscreen on its current monitor.

Pressing the Win+L key over the guest console to lock the Windows client session could be intercepted by the guest and repeated with no release. Now, when the client receives a grab-broken event, it releases all pressed keys. As a result, when pressing Win+L over the guest console, the 'L' key is not repeated.

An issue with sticky Windows or super keys after pressing Win+L has been fixed.

Previously, virt-viewer did not respect the automatic configuration of monitors as set by the client machine. Now, monitor settings are respected by virt-viewer and no longer need to be manually changed every time the virtual machine is restarted.

Previously, "AltGr" keys, such as @ or |, did not work with certain keyboard layouts, such as de-CH when using a Windows client with a Red Hat Enterprise Linux guest. Now, "AltGr" keys are passed correctly from the client to the guest so that these keyboard layouts work as expected.

When pressing Alt-Tab on a Windows virt-viewer client, the command was passed to the client instead of the guest. This update fixes this special keys grab regression, so Alt-Tab is caught by the guest only.

When running a mouse over a PuTTY session with Windows guests, the mouse pointer appears "black". This update improves the cursor contrast, so the cursor is visible on PuTTY sessions with Windows guests.

Hot Key combinations are now automatically added to the "Send Key" menu when remote-viewer is launched. These hot key menu entries can be used to pass key combinations to the guest.

virt-viewer has been rebased to version 0.5.6, which provides dedicated MIME file registration support, a new MSI build, and various bug fixes.

virt-viewer now translates end of line sequences to target different operating systems during copy and paste operations, because Linux and Windows systems use different end of line sequences.

virt-viewer did not accept proxy URLs ending with a '/' symbol. If proxy was specified in such a way, virt-viewer would not be able to use it. To fix this issue, virt-viewer now removes trailing / from proxy URLs, so proxy URLs with or without trailing / can be used.

virt-viewer now installs only one of 32-bit or 64-bit virt-viewer binaries on a 64-bit Windows client machine. If a version of virt-viewer is already installed, the other version will not be installed.

When disabling the first display of a multi-display setup, remote-viewer would hide it, but not indicate to the guest that it was disabled. If the first display was the primary one for the guest, the start menu and windows on that display were hidden away. 

To fix this, the monitor for the guest is disabled in addition to hiding the window. This will cause the guest to rearrange the first display content in the other visible displays.

Previously, trying to use smartcard passthrough in Enhanced Protected Mode of Internet Explorer 10 caused virt-viewer to crash. Now, the code has been updated so that smartcards can be successfully used in Enhanced Protected Mode of Internet Explorer 10.

This update provides a new virt-viewer MSI installer which is more administrator-friendly, robust, and follows Windows Installer practices.

In autoconf mode, remote-viewer opened as many windows as configured guest's displays. If the guest has more configured displays than the number of client's monitors, extra windows will be opened. Now, the extra guest displays are hidden. In autoconf mode, the number of fullscreen windows matches the number of monitors.

Invalid internal SPICE 'multimedia time' after a seamless migration while a video is playing caused the video to freeze for an indeterminate amount of time. Now, virt-viewer validates that the multimedia time is reset to a known value when migrating. so video playback is not affected by migration.

Two remote-viewer instances running at the same time could cause one of the instances to try to handle a USB device redirected by the second instance. Consequently, error messages popped up in remote-viewer when trying to use USB redirection. Now, only the remote-viewer instance which is redirecting a given USB device will try to interact with it.

When the Win key was pressed outside of virt-viewer, it was captured by both client and guest. Now virt-viewer does not handle special keys with a global hook when the keyboard is not grabbed. As a result, the Win key is captured only by Windows or by virt-viewer, and not both.

Previously Windows always captured the Ctrl+Alt+Del key sequence and did not let applications intercept it. It was not possible for mingw-virt-viewer to catch Ctrl+Alt+Del and pass it to the guest when the guest was in focus. This key combination can now be passed to the guest using the Ctrl+Alt+End key combination.

When opening a console to a guest configured with two monitors or more, not all guest windows were shown on their respective client monitors. Now, the guest windows are moved to the respective client monitors.

Previously virt-viewer would not properly detect agent disconnections, so it crashed when the guest SPICE agent terminated. virt-viewer now properly detects and handles agent disconnections, and no longer crashes on such events.

On a guest with multiple monitors, changing one display from fullscreen mode to windowed mode also affected all other displays. Now, full screen mode is left on the current display only, allowing greater flexibility in display placement.

Previously when a user set a window in fullscreen and then changed the resolution in the guest, the guest resolution reverted to the resolution of the client. Fullscreen windows now allow user to set the resolution of the guest to a resolution higher than the native client resolution.

Previously, keyboard shortcuts were not working properly on virtual machines using SpiceX. This update adds a missing "secure-attention" hotkey configuration in the setup file, and removes the difference between hotkey configuration behavior set with the controller or with the setup file.

Previously, installing rhevm-reports on a remote database failed when it could not connect to the database. Now when the database connection could not be established, it raises a runtime error instead of killing the installation process.

When upgrading rhevm-reports, the installation directory owner and permission changed, therefore the reports setup failed, and the Manager could not run. Now, the database and web archive (WAR) files are backed up and restored upon upgrade, and the correct permissions are restored.

The reports database schema has been updated to be compatible with JasperServer 5.5, so now the import and export functionality operates correctly.

Previously, setup would fail when upgrading rhevm-reports from version 3.2 or 3.1 using the postgres user as the remote user. This was due to an inability to drop cascade on cleaning the database as part of the upgrade. Now, rhevm-reports-setup uses cleandb.sh from dbscripts, which correctly drops all objects one-by-one.

Previously, upgrading from 3.2 to 3.3 with a local database only backed up the database after changing ownership of all objects. If a failure occurred before the ownership was updated, no backup would be created, or the restoration would be performed from a different backup. Now, the backup is performed before any changes are made to the database.

Previously, uninstalling the reports package caused reinstallations to fail while setting up rhevm-reports. This was because no password was supplied to connect to the reports database. The execExternalCmd command has been updated with the correct PGPASSFILE environment value, allowing execution of database commands.

When upgrading from 3.2 to 3.3 with a local database, ownership of all database objects is changed from 'postgres' to 'engine_reports'. This is done by running the pg_dump command to search for the postgres owner, which can fill up the disk with logs if the database size is large. Furthermore, the SQL commands for changing the ownership are run individually, which can be slow. Now, pg_dump only retrieves only the schema, which is sufficient for this purpose, and all SQL commands are run in a single PSQL command.

In some custom environments reports were not deployed correctly, which caused JBoss to try and start all deployed applications, and Manager initialization failed. The setup utility has been updated to run with correct environment settings. As a result, reports are now deployed correctly and the Manager starts successfully.

rhevm-reports-setup now supports unattended installation when an answerfile is supplied.

rhevm-reports can only be installed by the root user, and a warning appears if the setup process is run by an unprivileged user.

Instructions for creating a remote database did not appear during a user session, which could cause difficulty for users installing reports with a remote database server. Now, the database creation instructions are shown during the interactive setup session.

Upgrading the rhevm-reports from version 3.1 failed with the error "database "ovirtenginereports" does not exist" because the database name had been changed to rhevmreports. This has now been fixed, so installing rhevm-reports succeeds.

This new implementation requires the Manager and DWH to be run and set up correctly before the reports package can be set up. If the settings for the Manager or DWH are not found, the reports setup will exit with a corresponding message. The database configuration is now defined in a spec file.

rhevm-reports now validates existing configuration files, to prevent the installation from failing due to missing configuration files.

When entering different passwords on the administrative user prompt, the setup proceeds with the remote database flow, even when the local database is used. Now, the password prompt provides a suitable error message if the passwords do not match, and the setup does not proceed.

The instructions to troubleshoot a failed rhevm-reports installation were incorrect. They have been fixed, and now users are given the correct command to remove the database.

rhevm-reports now correctly reports the operating system versions on actively running virtual machine.

After installing the Manager, the instructions to install the database warehouse package were for "ovirt-engine-dwh-setup". This has now been changed to "rhevm-dwh-setup".

The input control in Trends dashboard read a variable as string when it should be an integer, causing errors. The variable is now read in the correct format, so the dashboard can be used without errors.

Several configuration files and values have been renamed to match the Manager conventions, in order to simplify the upgrade process.

Previously, rhevm-reports could not be installed after using otopi to set up the manager. This has now been fixed, and rhevm-reports can be set up as expected.

Previously, rhevm-reports-setup would fail on a clean installation because there was a mismatch in the reports user and ownership. This update fixes the instances when ownership was not correctly recognized, and fixes the change to the database credentials dictionary that could incorrectly cause the setup to try to use the wrong credentials.

After single sign on to the engine rhevm-reports-setup failed, because the externally defined password was listed with an encryption value of null. Removing the word null from the password field resolves this issue, and the reports setup does not fail.

Previously when rhevm-reports was removed, the Dashboard tab would still be visible in the web administration portal, and clicking on it would cause error. Removing the reports package now removes the tab as well.

A problem with starting the oVirt-ETL (extract, transform, load) service caused rhevm-dwh installation to fail. Now, rhevm-dwh uses the engine's script /usr/share/ovirt-engine/bin/java-home to detect the JAVA_HOME location, so this error no longer occurs.

The oVirt-ETL (extract, transform, load) service could not start after a successful rhevm-dwh installation. To fix this issue, dwhd service has been added to init levels using chkconfig call, so it is started or stopped in a similar way to the engine and postgres services.

Previously, upgrading from 3.2 to 3.3 only backed up the database after changing ownership of all objects. If failure occurred before the ownership was updated, no backup will be created, or the restoration will be performed from a different backup. Now, the backup is performed before any changes are made to the database, and all ownership changes are run in a single psql command.

Virtual machine client IP is now collected in user sessions tables and in virtual machine statistics ad hoc domains.

Previously, rhevm-dwh-setup used an incorrect execution of 'su' and failed. The 'su' utility location has been updated, so installation succeeds as expected.

ENUM values were loaded incorrectly into the database causing mis-rendered text. This issue is resolved and ENUMs appear normally for all locales in this release.

rhevm-dwh-setup now supports unattended installation when an answerfile is supplied.

During an unattended installation, if the CREATE_READONLY_USER parameter was specified in an answer file, the readonly user was not created. Now if CREATE_READONLY_USER is set to yes or true it will be created using validated values from READONLY_SECURE, READONLY_USER, and READONLY_PASS.

Previously, the installer passed a relative path to the database schema creation script, which created the log file under the current directory. Now, the installer passes the full log path, so the logs are created in the expected directory.

rhevm-dwh now checks the available space and the size of the database and prompts the user to create a backup of the database before performing an upgrade.

Boolean values have been added to the virtual machine user's usage statistics tables to indicate if a user is logged in to the guest.

Running rhevm-dwh-setup with a non-root user now exits with a notification message that the setup has to be executed by the root user.

The rhevm-dwh service was previously running with the 'root' user. It is now run with the 'ovirt' user, which is the same as the engine service.

The database size is now reported in human-readable format, using megabytes or gigabytes.

postgresql.conf was truncated during setup, which corrupted the database. The configuration file is no longer truncated, so database corruption does not occur.

The rhevm-dwh database was not correctly detected when upgrading from version 3.2 to 3.3 because it was owned by the postgres user. Now the database owner is correctly updated and detected during upgrade.

Previously when all hosts were non-responsive, report collection stopped. A heartbeat has been added to the Manager to make sure statistics are updated and collection continues even when all hosts are non responsive.

The created role 'engine_history' user and password were only saved to the configuration file at the end of the setup, and could be lost if the setup failed halfway. This could block reinstallation due to a missing password. rhevm-dwh-setup now saves these entities right after they are created.

Upgrading rhevm-dwh failed because a new user was created, ignoring the existing user, causing user and password mismatch. Now, a read-only user is created upon upgrade only if no user had been created previously.

A warning message misled users to think a database backup would take hours when the process only took several seconds. The message is now more specific, and only warns users of a long backup duration in specific cases, such as if the database size is large.

Clearer instructions are now provided for installing a remote database. Administrators are now instructed to create a database, a database user, and to configure the database encoding type.

rhevm-dwh-setup now validates a provided user name before it creates a read-only user. The user provided cannot be one of the reserved users (postgres, engine, engine_history and engine_reports), it cannot be empty and it should not exist in the database.

Custom virtual machine operating system information is now automatically synced to history database.

When configuring a remote database, the remote database user was incorrectly configured, so the rhevm-dwh service could not start. The setup now correctly configures the remote database user in the rhevm-dwh properties file, so this error no longer occurs.

To set up rhevm-dwh after the first attempt is unsuccessful, follow the instructions here: https://access.redhat.com/site/solutions/291973

rhevm-dwh now tests the postgres server to verify that it responds to SQL queries. This is useful for cases where the PostgreSQL service reports status Up but the server is not ready to receive connections. In such cases the setup fails on database operations.

After a system was upgraded from 3.1 to 3.3, the ovirt user could not write to the log file because the file was previously owned by the root user. Consequently, the rhevm-dwh service failed to start. The log permissions are now correctly updated so the ovirt user can write to the log file.

rhevm-dwh-setup now checks for valid names when creating a read-only user. The user name can only include alphanumeric characters, otherwise an error message appears.

Several configuration files and values have been renamed to match the Manager conventions, in order to simplify the upgrade process.

Previously, a clean install with a remote database, wrongly behaved as an upgrade and failed. Now it correctly does a clean install.

rhevm-dwh setup failed during upgrade, because the installer connected to the database using the engine_history user while the owner is actually postgres. Now, the database owner is set as engine_history during installation, so this error does not occur.

When upgrading from 3.2 to 3.3 with a local database, ownership of all database objects are changed from 'postgres' to 'engine history'. This is done by running the pg_dump command to search for the postgres owner, which can fill up the disk with logs if the database size is large. Furthermore, the SQL commands for changing the ownership are run individually, which can be slow. Now, pg_dump only retrieves the schema which is sufficient for this purpose, and all SQL commands are run in a single PSQL command.

Virtual machine pool names and ids are now saved to virtual machine configuration tables and can be reported on via ad hoc domains.

rhevm-dwh used to run as root, but now it runs as ovirt. Condequently there are two different files with database access credentials. This update makes these files owned by root:ovirt and perms 0640, read-only for ovirt and unreadable for everyone else.

Previously a wrong locale used in database creation could cause issues in runtime. Now when the database is created, the correct locale is always used.

Users can now change fields to column or group within a table in the ad hoc editor, which allows users to change the ad hoc report design more easily.

When an Ad Hoc report is modified, mails are not stopped from being sent, and no errors are sent via email. The existing scheduled job just sends the report with the latest modifications, and after the report is deleted the scheduled job stops emails from being sent.

In an ad hoc report, the domain incorrectly generated a query with left outer join when a right outer join was defined in the domain. The update fixes the way the query is generated, so now the result set is displayed correctly.

Context sensitive help is now available for reports. Tooltips have been added to input controls.

Ad hoc reports filters can now be reordered, making it easier for users to edit ad hoc reports filters.

When exporting to .odt format many report elements were missing. These missing elements are now added to the .odt export, so the exported report is more complete.

A report's description that had line breaks and carriage returns caused incorrect parsing by JasperReports Server in ad-hoc designer, and was locked in a read-only state. This issue was fixed in ad hoc designer, as a result reports are no longer locked.

Previously large data sets results in input controls caused an "Unresponsive script" error and reduced performance. The input controls have been redesigned to allow larger data sets, so these errors do not happen now.

During the ad hoc design phase, the column's summary overrode the group's headline. The user interface design has been improved, and this design phase now displays correctly.

Several unnecessary filters have been removed, so the input controls display has fewer rows, which streamlines the report running process.

It is now possible to add a group within a group in the Ad Hoc designer, making ad hoc reporting more flexible and powerful.

The Jasper reports server could not be installed when using a database password that contained quotes. A password containing quotes is now supported, and installation succeeds.

Previously a report in .ods format contained text which was cut off to fit the column size. This resulted in missing or incomprehensible data. There is now an option to export full text in xls and ods formats.

After saving a newly created Ad Hoc report, the report now displays the name by which it was saved.

When trying to edit filters or pressing Apply or OK on a report that had been opened for long time, the report did not load as the connection had timed out. Now when timeout occurs, the user is automatically logged out of the reports portal.

It is now possible to save filters, and select from a listbox. In addition to the Repository view, it is possible to expand and view all the filters that were saved, and launch them. This makes it easier for users to quickly run reports.

In previous releases, after a JasperReports Server upgrade the ad hoc creation date would change to the installation date. Now, the ad hoc creation date is now saved, so ad hoc reports can be sorted according to creation date.

When exporting a report to .ods format, the report's images went missing. Now, images can be exported to .ods as well.

Previously, when accessing reports from the administration portal, the browser opened a new tab with the following message: "To display this page, Firefox must send information that will repeat any action (such as search or order confirmation) that was performed earlier."
This has been fixed by removing the Javascript code that checks authentication and resubmits the request, so this confirmation message no longer appears.

JasperReports Server Pro has been rebased to version 5.5. A list of features and fixed issues from version 4.7.1 is available in /usr/share/jasperreports-server-pro/releaseNotes.txt

Large reports repository folders could cause slow performance and an "Unresponsive script" message to appear. The user interface has been redesigned to work in this case, and is now much more responsive on a large scale.

When a host was put into maintenance mode, not all virtual machines on the host were migrated. Now, if a virtual machine encounters an input/output error during migration, the migration is aborted by default.

When virt-v2v converted a virtual machine from a foreign hypervisor such as Xen or VMWare, it set the vm_snapshot_id of all the virtual machine's disks to an empty guid. This caused erroneous disk IDs to be assigned to virtual machines when snapshots were created. Now, the engine checks if the snapshot ID is empty and provides an error message, preventing snapshot mismatch errors.

Previously, the getdisksvmguid() procedure would consume a large quantity of available CPU resources due to the inclusion of a large volume of irrelevant data in database queries, causing those queries to be slow. With this update, queries now only retrieve relevant data, increasing the speed of queries.

This update introduces a new PrepareForHibernation virtual machine status, which denotes that a virtual machine is about to hibernate but VDSM had not yet saved its memory state. The status changes to SavingState when the 'hibernate' verb is called in VDSM and VDSM reports back that it has started migrating to the file.

Previously users could edit the USB policy while the virtual machine was running, which could lead to errors in running the virtual machine. Now, the USB policy can only be edited when the virtual machine is down, preventing users from attaching a new USB device without detaching the old one.

When creating a virtual machine from a template with a disk image, the virtual machine's disk inherited the template disk's quota setting, even if the user creating the virtual machine did not have access to that quota. Now, when the template disk quota is enforced, virtual machines cannot be created from the template with disks. If this is attempted a null pointer exception is raised.

Previously, sound devices would not be found when restoring the configuration of a stateless virtual machine, resulting in a new sound device being added. This would result in new sound devices being added to stateless virtual machines each time those machines were started, eventually preventing them from starting. With this update, the code that searches for sound devices has been updated so that sound devices are now correctly discovered when restoring stateless virtual machines, and no new sound devices are created. Stateless virtual machines now only have a single sound device and will not experience any issues due to sound devices after being started multiple times.

Retrieving a list of virtual machine templates required a long time when there was a large number of templates, which also resulted in a long time to create virtual machines from templates. This update improves the performance of the search query for templates by a factor of 5, so creating virtual machines from templates is now faster.

Previously, pre-started Windows virtual machines in a virtual machine pool would not run sysprep when they start for the first time. With this update, pre-started Windows virtual machines in virtual machine pools now run sysprep correctly when they start for the first time.

A failure in the virtual machine hibernation process (for example, a VDSM hook error) caused the virtual machine to be stuck in a 'Saving' state, even though it is actually 'Up'. Only restarting the Manager could resolve this issue. A new internal status has been added to flag different phases of the hibernation process, allowing better recovery. Now, the virtual machine returns to the 'Up' state when the hibernation process fails, allowing users to continue using the virtual machine as normal.

To prevent the risk of split brain incidents on Red Hat Storage domains, the use of shareable disks on Red Hat Storage domains is disabled. Attempting to create a shareable disk brings up a warning in the administration portal which recommends the use of Quorum on the Red Hat Storage server to ensure data integrity.

Note that this policy is only applicable for Red Hat Storage domains created as a Red Hat Storage domain type. This policy is not enforced on Red Hat Storage domains created on a POSIX domain with GlusterFS specified as the virtual file system type.

A highly available virtual machine with an attached CD could not restart after the host on which it was running was fenced, and if there was no Storage Pool Manager. This was because the ISO prefix, which is needed for the CD-ROM path setting to run the virtual machine, was queried from the SPM. Now, the ISO path is queried from the destination host, and determined using priority values in descending order, which makes the process more efficient. Virtual machines can now migrate successfully after host fencing.

The is_initialized flag was ignored on virtual machines belonging to a pool, so custom sysprep files could not be applied on pool virtual machines if they were created through the user portal. This fix sets the is_initialized flag to false, so custom sysprep files work as expected on pool virtual machines.

After attempting to cancel multiple live migrations, some virtual machines were killed. To fix this, when the migration is cancelled, libvirt raises an error to prevent the operation from proceeding, which also avoids calling the destination VDSM to create the virtual machine instance.

Previously, pre-started virtual machines in a virtual machine pool would sometimes remain in an image locked state after being shut down. This was caused by a race condition in asynchronous task handling that would occur when the Red Hat Enterprise Manager attempted to start a pre-started virtual machine while that virtual machine was being shut down. With this update, shared resources are now locked correctly, allowing pre-started virtual machines in virtual machine pools to be stopped and started without issue.

Previously, OpenSSL would used special characters to build the subject field of certificates for Hypervisors being attached to the Red Hat Enterprise Virtualization Manager, resulting in an invalid subject when such characters were present. With this update, special characters are now escaped, allowing the subject to be produced as required.

Previously, the vm-devices feature included when upgrading from Red Hat Enterprise Virtualization 3.0 would cause the 'boot' flag on virtual machine disks not to be taken into account during the upgrade process. On virtual machines with more than one disk, this would sometimes result in a different disk being used as the boot disk after the upgrade, preventing those virtual machines from starting. With this update, the upgrade script now takes the 'boot' flag into account and retains the same boot disk after the upgrade is complete, allowing virtual machines with more than one disk to start without issue.

The CPU pinning option is now available for virtual machines running on the Local on Host type data center, and is enabled by default even if the virtual machine is not pinned to the host.

Previously, if importing a virtual machine failed for any reason, then trying to re-import the virtual machine again failed, stating "the Storage Domain already contains the target disk(s)". Only manual removal of the disks from the storage domain would allow the virtual machine to be re-imported properly. Now, if a virtual machine fails to import correctly, the disks are removed from the storage domain automatically and the virtual machine can be re-imported.

Previously, the MaxVmNameLengthWindows configuration option would be ignored, resulting in a hard-coded limit of 15 characters being imposed on the names of Windows virtual machines. With this update, the value of the MaxVmNameLengthWindows configuration option is now read correctly, allowing administrators to define Windows virtual machines with names longer than 15 characters if the configuration option allows it.

Previously, users needed import-export permissions on a virtual machine to remove it from the export domain. This was problematic if the virtual machine to be removed did not exist on the system. Now, the 'delete vm' permissions have to be conferred on the user for the export domain, so users can now delete virtual machines from the export domain even if they do not exist in the database.

Scanning an export domain marked as Up while the share is not available executes an SPM command which will attempt to perform failover in case of failure. This command was issued every 5 seconds, causing the requests to pile up and perform multiple failovers. Now, the automatic refresh is replaced with a manual refresh button, which decreases the number of failover attempts.

This feature provides support for a branding configuration file that allows administrators to configure the look and feel of Red Hat Enterprise Virtualization, including branding, and override certain text displayed to end users. With this feature, administrators can now install an external package that will override logos, titles and certain colors in the welcome page, Administration Portal and User Portal.

When a virtual machine disappeared or crashed during a Live Storage Migration, the newly created image on the destination storage domain was not cleaned up, preventing future migrations. Now the images on the relevant storage domain are properly removed.

A desktop virtual machine imported from version 3.0 did not have any sound devices. This update adds a check to the virtual machine and adds sound devices if it is a desktop machine.

Failure to remove volumes from an import domain marked the imported volumes on the data domain as illegal. With this update, the engine now checks if the parent command is an image or template removal command before setting the volume as illegal.

This update introduces two parameters that control the time required to react when there are connection issues. The vdsConnectionTimeout parameter sets the amount of time allowed for establishing connections for hosts before timing out. The vdsRetries paramter sets the number of times to retry host operations when there are I/O errors.

Previously the snapshot manager attempted to remove disks that are not part of a snapshot. Now, if the device is not a disk known by the engine, it is not removed, similar to the behavior of non-snapshot disks.

Previously a non-initialized array caused a null pointer exception when trying to create a virtual machine from a template without disks. This exception is now handled correctly, so virtual machines can be created from a template without disks.

This update introduces a search field that allows a user to search only for users (type = user) or groups (type = group).

Previously, engine-setup would not enforce umask or explicitly apply file attributes, resulting in an unusable installation if an administrator set up a custom umask. With this update, the expected umask is now enforced at the beginning of the setup process, resulting in a usable installation regardless of the umask.

When extending a LUN-based storage domain, the Manager attempted to connect all the hosts to the LUN. If one or more hosts fail to access the chosen LUN for the extension, the Manager attempted to disconnect all hosts from the target. Now, if the connection ID is null, the Manager loads it from the database, to prevent disconnection from targets that are used by other LUNs.

When creating a template from a virtual machine which has disks in multiple storage domains, the disk allocation table selected the first storage domain as the default storage domain, causing all disks in the template to be created on the single storage domain regardless of where the original disks resided. Now, the disks are created according to where the original disks resided.

Previously, certain disks and permissions would be left in the database after forced removal of a storage domain or storage pool, resulting in partial titles remaining visible in the Permissions tab of the details pane for the User tab. With this update, the forced removal procedure has been updated to include the deletion of all disks and permissions from the database.

The 'Refresh Capabilities' button is now available on the 'Hosts' tab of the administration portal. This allows the administrator to obtain updated network configuration (such as available NICs) from the host.

When deleting a snapshot, the volume was extended by 10% to allow for additional qcow2 metadata. This was not necessary in cases where the volume was very large, for example a 1TB volume would be allocated 100GB when it only needed 10GB. This update adds an autoshrink capability which shrinks the resulting merged volume after deletion of a snapshot.

Previously, the value of the SpiceReleaseCursorKeys configuration key would not be localized in the front end if changed by engine-config. This would result in an exception that prevented users from accessing the console of virtual machines. With this update, front-end support for the SpiceReleaseCursorKeys configuration key has been added, making it possible to change the value of this configuration key using engine-config and continue to access the console of virtual machines.

The configuration of the management bridge on the host was moved from ovirt-host-deploy phase into the engine for the 3.3 cluster level. Once the host is installed, the engine will configure the management network according to its logical definition on the data center level.

Previously, the ovirt-host-deploy created the management bridge using more primitive methods compared to the engine capabilities. This lead to higher chance of fail, failure to revert or having incorrect network settings.

As a side effect and by trusting the engine to be able to recover from invalid network configuration, the host reboot is no longer required as the last step of the host installation.

Previously, attempting to import a raw sparse disk to a block domain would result in a generic error message being displayed. The content of the error message displayed when attempting an unsupported import action has been updated to provide additional information on disk configuration and storage domain compatibility.

The engine performed reverse DNS lookups of IP addresses when it failed to connect to the database. This includes looking up its own name and delaying the rest of the operations, including the deployment of the connection pool, which raises a null pointer exception in the administration and user portals. This update modifies the engine so that it locates the data source in a loop and retries until it is successful .

A description field has been added for storage domains. Users can describe storage domains using the field in the administration portal and REST API.

After setting UserSessionTimeOutInterval value, automatic logout did not always happen after the specified period for inactivity, it took much longer than the time set. This update adds the UserSessionTimeOutInvalidationInterval parameter which invalidates the session timeout after a specified amount of time has elapsed and the logout has not occurred. The actual timeout value will be between the times set for both parameters.

The audit logs now distinguish between regular and force removal of data centers, which mirrors existing behavior for storage domain removal.

Previously, it was possible for separate users to attempt the installation, upgrade or approval of a single virtualization host at the same time, or for a single user to to fire such events multiple times by clicking a button more than once in succession. This would result in the same action being performed concurrently on the virtualization host, causing conflicts in processing. This feature locks hosts during the installation, upgrade and approval of virtualization hosts, preventing these actions from being run concurrently.

All virtual machine operating system properties (from vdc_options and VmOsType.java enum) are now merged in one file-based repository which is external to the code. This makes it easier for administrators to extend and configure operating system information.

Previously, a race condition would occur while loading the Import Template(s) window, preventing the window from loading and making it impossible to import templates via the Administration Portal. With this update, server calls are now serialized correctly, making it possible to import templates using the Import Template(s) window.

Previously, the data warehouse would assume that the engine was not running when all hosts registered in the engine were non-resposive because the engine would not update their data. This update introduces a data warehouse heartbeat job that lets the data warehouse know that the engine is active even when all hosts are in a non-responsive state. The heartbeat job functions by periodically updating the status in the database to notify the data warehouse that the engine is active. The interval for updating the heartbeat can be configured via engine-config using the DwhHeartBeatInterval variable.

When an SSH session is established with a host, the Manager extracts the destination host key fingerprint before authentication, and validates this fingerprint against the expected host fingerprint. If the fingerprints do not match, the Manager disconnects the host and fails host installation. This implementation prevents credentials or sensitive data from being exposed during host validation.

Previously, only user name and password authentication was supported for host installation and upgrade. With this update, public key authentication is now supported. This feature provides a higher level of security, streamlines subsequent authentication, and reduces the possibility of man-in-the-middle attacks by displaying and retaining the SSH public key fingerprint of remote hosts.

Previously, live storage migration of unplugged disks on a running virtual machine would fail. With this feature, administrators can now migrate unplugged disks on a running virtual machine as though the virtual machine was offline. Furthermore, errors messages are now displayed when attempting to migrate a mixture of plugged and unplugged disks to notify administrators that such operations are not permitted.

Previously, it was not possible to set the engine time zone using engine-config. With this update, engine-config can now be used to configure the time zone via two keys: DefaultWindowsTimeZone for Windows virtual machines, and DefaultGeneralTimeZone for all other virtual machines.

Red Hat Enterprise Virtualization Manager now supports a virtual watchdog device which can be used to pause, reset, dump a guest. The watchdog model and action options can be set from the New Virtual Machine -> High Availability tab in the administration and user portals.

Note that in order to use the watchdog device you need the relevant drivers depending on the guest operating system.

With this feature, the Edit Virtual Machine window now uses the default time zone specified in engine-config. The default time zone used in the window corresponds to the DefaultWindowsTimeZone key for Windows-based virtual machines and the DefaultGeneralTimeZone key for all other virtual machines.

Previously, deactivating the master domain and concurrently putting all hosts in maintenance left the hosts non-operational upon activation. With this update, a host that runs through InitVdsOnUp does not attempt to reconstruct. In case of failure during ConnectStoragePool, the host fails in initializeStorage only if the master domain is not in an inactive or unknown status and if the exception was not an XmlRpcRunTimeException.

Red Hat Enterprise Virtualization Manager now supports OpenLDAP as a domain provider, with the following conditions:

1. Kerberos must be installed and co-hosted with the OpenLDAP server.

2. The ldif files must be installed. 

Run the following commands to add the cosine and inetorgperson schemas:

# ldapadd -H ldapi:/// -Y EXTERNAL -f /etc/openldap/schema/cosine.ldif
# ldapadd -H ldapi:/// -Y EXTERNAL -f /etc/openldap/schema/inetorgperson.ldif

Add the memberof overlay: 

# cat > memberof.ldif <<'.'
dn: cn={0}module,cn=config
objectClass: olcModuleList
cn: {0}module
olcModulePath: /usr/lib64/openldap
olcModuleLoad: {0}memberof.la

dn: olcOverlay={0}memberof,olcDatabase={2}hdb,cn=config
objectClass: olcConfig
objectClass: olcMemberOf
objectClass: olcOverlayConfig
objectClass: top
olcOverlay: {0}memberof
.

# ldapadd -H ldapi:/// -Y EXTERNAL -f overlay.ldif

3. The field used to query for groups is groupOfNames, and the field for users is uid. 

4. The field used to map returned results to engine entities are:
- For group: entryuuid , memberof
- For user: entryuuid, givenname, title, mail, sn, memberof, uid

External systems and plug-ins can now log their events in the application Audit Log. This feature allows administrators to view events linked to external tasks using the same mechanism as application events.

This update enhances logging of the JndiAction class in debug mode, which enables users to view more detailed information on the query path, URI and user for performing the authentication and ticket creation.

When the host reported the ISO or Export domain as problematic during the InitVdsOnUp flow, it did not move to status Up. When ISO or Export domains are reported as problematic by some of the hosts those hosts remain Up and do not move to NonOperational. The behaviour between these two flows have been unified, so when hosts report ISO or Export domain as a problem, it does not stop the host from moving to Up.

With this update, the ovirt-engine configuration file located in /etc/sysconfig/ has been migrated to /etc/ovirt-engine/engine.conf. This is because the configuration file is not only used by the ovirt-engine service, but by other tools as well, and because the /etc/sysconfig/ directory is not common in distributions other than Red Hat Enterprise Linux.

This feature adds a new step to the flow of automatic fencing whereby non-responsive hosts are made responsive again faster and without having to perform real fencing. With this feature, VDSM is restarted using an SSH connection when a host is non-responsive, and real fencing is executed if this restart does not make the host responsive again. For more information, see http://www.ovirt.org/Automatic_Fencing#Automatic_Fencing_in_oVirt_3.3

This update changes the way asynchronous task placeholders are cleared on server restart for commands that are partially submitted to VDSM. For example, if a task has 4 placeholders and only 2 have been submitted to VDSM, the database has two tasks with VDSM IDs respectively and 2 tasks with empty IDs. When the server is restarted the task placeholders with empty VDSM IDs are removed from the database, and the tasks which have VDSM IDs are added to a partially completed tasks list. Once polling is done they are removed from the database.

Previously, administrators were not able to use engine-config to define whether communication between the Red Hat Enterprise Virtualization Manager and hosts was secure. With this update, administrators can now use engine-config to define whether communication is secure with the EncryptHostCommunication configuration key.

Previously, using the rhevm-manage-domains action=edit command for LDAP servers did not work as the changes were not stored at the proper location at the vdc_options table. This fix specifies the correct location for storing the changes, so edits to the LDAP server can be made using the rhevm-manage-domains tool.

Previously in the REST API, it was not possible to view a list of clusters for a data center. With this update, a clusters sub-collection has been added to data centers, making it possible for administrators to retrieve a list of clusters belonging to a data center using the REST API.

It is now possible for the REST API to retrieve users and disk resources by pagination, so the results can be sorted accordingly to the specified category.

Previously all engine failures resulted in 400 BAD_REQEST. Infrastructure has been added to return proper error messages, depending on the exact failure.

Previously users could not update power management options via the command line, while it was possible to do so through the administration portal. With this update, all the fields available from the administration portal can be used via the command line.

Previously in the REST API, there was no disk sub-collection for virtual machines or templates in an export domain. Due to this, users would not be able to use the REST API to view the disks owned by these objects. With this update, a disk sub-collection has been added to virtual machines and templates in export domains, making it possible for administrators to view the disks that belong to those virtual machines or templates.

/api/capabilities now lists the state of snapshots, for example "locked" or "ok".

Previously in the REST API, there was no way for administrators to view a list of the applications installed on a virtual machine. With this update, administrators can now retrieve a list of the applications installed on a virtual machine by using the applications sub-collection of a vm object.

Previously in the REST API, the keyboard layout used by virtual machines could only be set via a single, global configuration option. With this update, administrators can now set the keyboard layout for individual virtual machines via the keyboard_layout option in the display subtype of a vm object.

Previously, using a Java runtime environment other than OpenJDK as the default Java runtime environment would sometimes result in an unusable engine keystore when upgrading to Red Hat Enterprise Virtualization Manager 3.2. This was caused by PKCS#12 being output in a format unusable by OpenJDK. With this update, the OpenJDK keytool utility is now explicitly used, resulting in a usable keystore.

The Red Hat Enterprise Virtualization Manager 3.0 to 3.1 upgrade did not copy all pki files that were generated during the 3.0 setup. This resulted in issuing certificates without extensions in upgraded environments, where attempting to install new hypervisors would fail with the error message "Certificate enrollment failure". This update restores the pki artifacts missing from the 3.0 upgrade, so new hypervisors can be added without certificate errors.

Previously, migrating from Red Hat Enterprise Virtualization version 2.2 to 3.0 would update the key store with the CA certificate and not the trust store, resulting in an unusable trust store when later upgrading to Red Hat Enterprise Virtualization version 3.2. With this update, the trust store is now re-created during the upgrade to Red Hat Enterprise Virtualization 3.2, resulting in a valid trust store.

Previously, local replication of a yum repository would sometimes fail to correctly replicate groups depending on the tool used, resulting in no groups being replicated. This would result in yum issuing an exception when attempting to search for groups instead of returning an empty set, causing engine-setup to fail. With this update, such exceptions are now ignored and an empty set is returned, making it possible for engine-setup to proceed.

The java-1.7.0-openjdk-1.7.0.45-2.4.3.3.el6 package modified the default configuration of Java to use NSS as a default security provider, due to a bug in the PKCS#11 provider the JVM leaked memory. This issue is resolved by adding a conflict with java-1.7.0-openjdk-1.7.0.45-2.4.3.3.el6 so that older or newer packages will be pulled. There is now no memory leak as NSS is not the default security provider in these packages.

Previously, when installing Red Hat Enterprise Virtualization Manager, the system would not allow installation to proceed unless a fully qualified domain name (FQDN) was specified. This caused problems under some conditions, such attempting to install the Red Hat Enterprise Virtualization Manager in a lab where an FQDN may not be available. With this update, the Red Hat Enterprise Virtualization Manager can now be installed without specifying an FQDN unless in all-in-one configuration.

The backup and restore procedure can now be used to move the database from one location to another, for example from local to remote.

Previously, there was no supported method for renaming the Red Hat Enterprise Virtualization Manager. Due to this, the only methods for renaming the Red Hat Enterprise Virtualization Manager were difficult to perform and prone to error. With this update, a utility (/usr/share/ovirt-engine/setup/bin/ovirt-engine-rename) has been introduced that provides administrators with a supported method for changing the name of the Red Hat Enterprise Virtualization Manager and updating related configuration files and PKI certificates.

Previously, backup files created while upgrading the Red Hat Enterprise Virtualization Manager were stored in different locations. With this update, these backup files are now all stored in /var/lib/ovirt-engine/backups.

The taskcleaner.sh -l option now correctly produces a log file.

Previously, all packages except the rhev-guest-tools-iso package would be downloaded and installed when upgrading from Red Hat Enterprise Virtualization version 3.0 to 3.1. With this update, the rhev-guest-tools-iso package is now a dependency of rhevm, and older versions of rhev-guest-tools-iso in the ISO domain are updated when running engine-setup.

The Red Hat Support plug-in for Red Hat Enterprise Virtualization enables users to create and manage Red Hat support cases and access Red Hat documentation from the Red Hat Enterprise Virtualization Manager administration portal. It offers easy and instant access to Red Hat knowledge, resources, engagement, and diagnostic features.

This update enforces the use of an Apache proxy during a new installation of Red Hat Enterprise Virtualization version 3.3 if it is not enabled already, and forces migration to an Apache proxy during an upgrade from earlier versions of Red Hat Enterprise Virtualization Manager. Redirection of the ports for http and https to ports 80 and 443 respectively is also recommended. However, while the user is prompted to use ports 80 and 443 for http and https during the engine-setup process, existing ports are preserved when this process is performed without user interaction.

Conflicts with freeipa-server and ipa-server are now also detected, and installation or upgrade of Red Hat Enterprise Virtualization Manager on systems running these servers is disallowed.

The user portal can now be deployed on a standalone machine separate from the Red Hat Enterprise Virtualization Manager. This is done by configuring a web server to function as a proxy to the user portal. Further instructions can be found on https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Virtualization/3.3/html/Installation_Guide/index.html

Previously, Red Hat Enterprise Virtualization Manager would use multiple log rotation mechanisms, including the JBoss rotation mechanism for Java. Because the JBoss rotation mechanism does not support the compression of rotated logs, a proprietary cron script was introduced to compress and remove logs. However, the implementation of this script is application-specific and difficult for administrators to access. With this update, the standard logrotate package is now used to rotate logs for Red Hat Enterprise Virtualization Manager, resulting in a log rotation mechanism that is easier for administrators to access and allows configuration of log rotation capabilities via /etc/logrotate.d/ovirt*.conf such as interval, size, compression.

Previously, the disk table of the Basic tab in the User Portal would be automatically refreshed every five seconds, causing the Red Hat Enterprise Virtualization Manager to consume a high level of CPU resources to refresh the list. With this update, automatic refreshing has been removed and replaced by a button for manually refreshing disks, greatly reducing CPU resource consumption.

On the Resources tab in the Power User Portal, the user interface appeared unable to display all virtual machine disks. The layout has been fixed so all items are now correctly displayed.

Previously, user input for disk quotas would be ignored when a user created a virtual machine based on a template and quota mode is not disabled. This would cause the disk quota of the new virtual machine to be inherited from the template. With this update, the disk quota of the new virtual machine is set in accordance with user input, while the default disk quota remains that of the template. This allows users to change the disk quota for new virtual machines created based on a template.

This feature integrates the SPICE-HTML5 console client with Red Hat Enterprise Virtualization Manager, allowing users to connect to the SPICE console from a browser via the SPICE-HTML5 client. For this feature to work, the websocket proxy must be set up and running in the virtualization environment, and the Red Hat Enterprise Virtualization Manager must be aware of the websocket proxy by correctly configuring the WebSocketProxy configuration key using engine-config. Moreover, the client machine must have a browser that supports websockets and postMessage, and the Certificate Authority of the Red Hat Enterprise Virtualization Manager must be imported into the client browser is SSL is enabled.

Clipboard sharing is now enabled by default between a virtual machine client and guest using an RDP console.

Previously, the performance of displaying and updating sub-tabs in the General tab of virtual machines and templates in the User Portal was low when using older browsers such as Internet Explorer 8, reducing the overall performance of the User Portal. With this update, the implementation of these sub-tabs has been optimized to avoid unnecessary DOM operations, resulting in greater performance for sub-tabs in the General tab for virtual machines and templates both in the User Portal and Administration Portal.

Previously, when a user was logged in to the User Portal, they would never time out even when idle, preventing the user from being automatically logged out. With this update, the automatic refresh queries have been updated to correctly pass the flag that indicates that the user session should not be refreshed, automatically logging users out of the User Portal after the pre-set idle time.

The lexo-numeric comparator was using the Integer class which is limited to parsing numbers, and would cause an exception if the sequence of numbers was too long. Consequently, it was not possible to assign any network to any cluster or any virtual machine. This update changes the class to BigInteger which is unlimited in size, and has the required functionality. Now networks can be correctly assigned to clusters and virtual machines.

Previously, cloning a virtual machine based on the snapshot of a virtual machine would clone the virtual machine devices to the original virtual machine instead of the new virtual machine. This would cause the original virtual machine to become corrupted. With this update, virtual machine devices are now correctly copied to the new virtual machine, preventing corruption in the original virtual machine and making it possible to clone new virtual machines based on the snapshot of a virtual machine without issue.

Previously, the Import Domain operation executed from the Administration Portal would sometimes fail to attach the relevant Storage Domain. With this update, this issue has been resolved by eliminating a race condition in the execution of the steps required to complete the Import Domain operation.

Previously, when a host was put into maintenance, its icon changed immediately from a green Up arrow to a wrench which signifies that it is already in maintenance, even when it was still preparing for maintenance. This update introduces a new hourglass icon to denote the preparing for maintenance stage.

Several virtual machine status icons have been changed, including different icons for when a virtual machine is running or in a run once operation, new icons for migration, paused, suspended, not responding, and powering down.

Previously, the Manager prompted users to upgrade the hypervisor even when there were no updates available. This fix adds a validation to compare the hypervisor version against ISOs available on the Manager, and only prompts users to upgrade when there are new versions available.

The miimon=100 parameter is now set by default for preset bonding options. This allows the kernel to poll for the status of bonded NICs.

Users can now manage Red Hat Storage hooks (volume lifecycle extensions) from the Red Hat Enterprise Virtualization Manager. The hooks are deployed during installation of Red Hat Storage servers in a cluster. The functions available to users are enabling or disabling hooks, viewing hook content and resolving content conflict.

There is now a button to refresh the list of ISO images in the ISO domain 'Images' tab, and in the virtual machine Boot Options and Run Once windows. This enables users to select images from an updated list.

Previously, the Load column in the Host tab only reported the virtual machines in the Up status, without taking into account virtual machines which were in the process of migration. This has been fixed, and now virtual machines which are being migrated are also counted.

Previously, certain disks and permissions would be left in the database after forced removal of a storage domain or storage pool, resulting in partial titles remaining visible in the Permissions tab of the details pane for the User tab. With this update, the forced removal procedure has been updated to include the deletion of all disks and permissions from the database.

Previously, when an administrator attached nested tags with three or more levels of nesting to a resource, it was not possible to detach tags on the third level of nesting and below. With this update, administrators can now attach unlimited levels of tags, and detach tags on all levels without issue.

On Red Hat Storage nodes, gluster processes were blocked by iptables rules, so volumes could not be mounted. This update adds the ports required for GlusterFS brick processes (49152 to 49251) and for accessing Gluster volumes (port 111) in the iptables configuration file. It also leaves the old ports (24009 to 24108) open for compatibility with older servers.

Previously, upgrading a Data Center would sometimes involve upgrading the meta data of the storage pool in an operation that would take a long time to complete, during which no warning was displayed to the administrator. This would create the impression that an error may have occurred. With this update, administrators are now notified when a Data Center upgrade involves upgrading the storage pool meta data. This notification is displayed when upgrading from Red Hat Enterprise Virtualization version 2.2 to 3.* and from version 3.1 to 3.2 or version 3.2 to 3.3.

Previously, when virtual machines and templates were exported, some virtual machine devices would be marked with a resource type of OTHER. On import, it was not possible to tell which of the devices with a resource type of OTHER needed to be treated as managed devices, resulting in incorrect or duplicate device addresses for these devices. This would prevent virtual machines created based on imported templates from starting. With this update, special devices with a resource type of OTHER are marked as managed when importing templates, allowing virtual machines created from those templates to be used without issue.

Previously, when creating a new virtual machine based on a template, it was only possible to see the name of the template in the Based on Template drop-down list in the New Virtual Machine window. With this update, the description of the template is also included in the Based on Template drop-down list, providing additional information about the selected template and improving the user experience.

Previously, when an administrator would change the width of columns in the Administration Portal, those widths would not be saved when the page was refreshed or after the administrator logged out. With this update, column widths are now saved for individual browsers and are retained across page refreshes and subsequent logins.

Previously, when performing a search by pressing the Enter key, the entries in the search results would be duplicated, and selecting a duplicated entry would select both results. With this update, the code has been revised to prevent duplicate entries from appearing in the results of searches performed by pressing the Enter key.

Previously, it was not possible to specify the host on which a virtual machine would start running when starting the virtual machine using the Run Once option in the Administration Portal. This feature adds a field to the Run Once window that allows administrators to specify the host on which the virtual machine will start running.

This feature makes it possible to sort virtual machines in the Administration Portal in accordance with the host on which the virtual machines are running. With this feature, administrators can now access 'host' under 'Vm: sortby' in the search bar.

Previously, it appeared that users could edit the quota of a virtual machine pool using the Edit Pool window, because the quota field was enabled. The quota field is now disabled when editing a pool.

This feature allows users to access RDP consoles without the Microsoft RDP ActiveX plug-in. However, users can still choose the method of invocation via the Console Options dialog: Plugin, which requires Microsoft Internet Explorer and the RDP ActiveX plug-in; Native, which is supported in all browsers independent of whether the RDP ActiveX plug-in is present, and Auto, in which the Plugin method is used if available and the Native method is used if not. Furthermore, administrators can configure the default invocation method via the ClientModeRdpDefault configuration key using engine-config. Note that when using the Native invocation method, single sign on is disabled due to technical limitations of the RDP descriptor file.

Previously cloning a quota dropped previously-assigned consumers and permissions. Now users can select the option to copy consumers and permissions when cloning a quota, preventing manual reassignment.

Previously, when switching between sub-tabs of a main tab in the Administration Portal (for example, switching between the Network Interfaces and Disks sub-tabs when a virtual machine is selected) the details of the old sub-tab would display for a second before being replaced by a loading indicator. This created a source of confusion as to the currently selected sub-tab until the loading indicator displayed. With this update, the loading indicator is now shown when switching sub-tabs rather than when displaying the old data, providing a better indication of the current sub-tab and improving the user experience.

Previously, it was not possible to set the keyboard layout for individual virtual machines when connecting using the VNC protocol. Moreover, changing the global keyboard layout applied to all virtual machines was inflexible and required the Red Hat Enterprise Virtualization Manager to be restarted. With this update, the keyboard layout of individual virtual machines when connecting using the VNC protocol can be specified via the Edit Virtual Machine window. If not keyboard layout is specified, the default, global value is used.

The drag and drop logic for bonding in the Host Setup Networks window has been improved for ease of use. Now, bonds and interfaces can be joined by dragging one onto another without manual detaching and reattaching. The original network assignments are not overwritten unless specified.

Older versions of the Manager set the virtual machine's default time zone to an empty string, but this field is now set to a null value. This update adds a database migration script to handle this change, so users can update the time zone on running virtual machines.

Previously when a host was moved from a Virt cluster to a Gluster cluster, there was no check for Gluster capabilities on the host. Now, during a cluster change the host is checked for Gluster capabilities, and if it fails the host is not activated. The VDSM dictionary is also updated, so service monitoring strategies are updated.

Previously, the Red Hat Enterprise Virtualization user interface technology stack used older releases of dependent third party software such as software development kits, frameworks, and libraries. This prevented the user interface from incorporating the latest features of these components such as new features, bug fixes and other enhancements. With this update, all components of the Red Hat Enterprise Virtualization user interface technology have been upgraded to the newest release, making it possible to take advantage of the latest improvements that each component offers.

In the General tab under the Cluster tab, the Maximum Memory Over Commitment unit has been changed from MB to % for better readability.

Previously, when an administrator would look at the sub-tabs of the Templates tab, such as Network Interfaces, Disks or Storage, the data grid would appear blank if no network interfaces, disks or storage were defined. This caused the appearance of these sub-tabs to be inconsistent with those of other sub-tabs, which would display 'No items to display' when there was no data. With this update, the sub-tabs on the Templates tab now display this message when there is no data to display.

This feature provides support for a branding configuration file that allows administrators to configure the look and feel of Red Hat Enterprise Virtualization, including branding, and override certain text displayed to end users. With this feature, administrators can now install an external package that will override logos, titles and certain colors in the welcome page, Administration Portal and User Portal.

Users can now limit the inbound and outbound network traffic on a virtual NIC level by applying profiles which define attributes such as port mirroring, quality of service (QoS) or custom properties.

The General tab of the Hosts tab now displays the SPM priority of each host, for easy reference.

virtio-SCSI is a new para-virtualized SCSI controller device which provides similar performance as the virtio-blk device, while improving scalability, supporting standard SCSI command sets and device naming, allowing for SCSI device passthrough.

Attaching a storage domain to a pool now automatically activates the domain, so users no longer have to manually do it.

Previously, it was not possible to change the size of a virtual disk after that disk had been created. This feature adds a new field 'Extend size by (GB)' to the Edit Virtual Disk window that allows users to increase the size of a virtual disk in increments of single GB. With this feature, users can now extend disks even while those disks are in use.

This feature adds a new Comment field to data centers, logical networks, storage domains, clusters, hosts and virtual machines that allows users to add human-readable, free-text descriptions to these resources.

This features introduces the noVNC console client to Red Hat Enterprise Virtualization Manager. With this feature, users can connect to VNC consoles from within their browser using the noVNC client. For this feature to work, the websocket proxy must be set up and running in the virtualization environment, and the Red Hat Enterprise Virtualization Manager must be aware of the websocket proxy by correctly configuring the WebSocketProxy configuration key using engine-config. Moreover, the client machine must have a browser that supports websockets and postMessage, and the Certificate Authority of the Red Hat Enterprise Virtualization Manager must be imported into the client browser is SSL is enabled.

The setup scripts of the engine did not keep track of the packages used during setup, upgrade, or cleanup. Conseqeuently it was sometimes hard to analyse and find the root cause for problems if a user did not know the exact version of the packages used. Now, the setup, upgrade, and cleanup scripts have been changed to log an event when started and successfully ended, including the list of packages installed. As a result, when problems are encountered, it is easier to know the exact set of actions taken, whether they succeeded or not, and what versions of packages were installed during these actions.

The Storage Pool Manager role can be manually assigned or re-assigned to hosts, using the administration portal or the "forceselectspm" action on the REST API.

Custom properties can now be defined for virtual network interface cards (NIC) on virtual machines. This enables a range of connection options, including:
* Create a host NIC via Mellanox UFM and connect it directly to a virtual NIC.
* Use OpenStack's Quantum to connect a virtual NIC to one of its defined networks.
* Pass non-standard quality of service (QoS) settings for a virtual NIC.

Initial support for OpenStack Neutron as a network provider on Red Hat Enterprise Virtualization Manager is offered as a technology preview. 

Description of the Technology Preview: 
http://www.ovirt.org/Features/Detailed_Quantum_Integration

A cluster administrator can now grant the "migration" role to a cluster network. This feature separates migration data to the designated migration network, to prevent migration traffic from choking other networks. The Manager uses the migration network's IP address on the destination host when it requests a virtual machine migration, replacing the previous implementation when a virtual machine's details were sent in plain text to the destination host via the management network.

Red Hat Enterprise Virtualization now supports native GlusterFS-based storage domains and data center types.

Previously if the hypervisor was not directly visible to the SPICE client, users could not connect to a console of a virtual machine running on it. Now, it is possible to specify an FQDN for a host to return instead of its IP address, allowing users to connect to the virtual machine console.

Red Hat Enterprise Virtualization now provides an API set for Independent Software Vendors (ISVs) to backup and restore virtual machines. For backup, a snapshot of a virtual machine's disk is created then attached to a virtual appliance. For restore, disks are attached to a virtual appliance, the data is restored to the disks, then the disks are attached to a virtual machine

The Storage General tab now has a new column showing the total, free and used sizes of each storage domain.

The virt-alignment-scan application scans the partition alignment in virtual machines and reports to the Manager if a partition is not aligned.

A virtual machine's memory limit can be edited using the VM64BitMaxMemorySizeInMB, VM32BitMaxMemorySizeInMB and VMMinMemorySizeInMB options in rhevm-config. Likewise, a virtual machine's maximum number of CPUs can be edited using the MaxNumOfCpuPerSocket, MaxNumOfVmCpus, and MaxNumOfVmSockets options.

Virtual machines are now automatically unpaused after recovering from storage errors and the storage domain is active, decreasing downtime.

When trying to add a Storage Domain to an uninitialized Storage Pool, there is now a warning message that the Pool is uninitialized and that the user should first add a data domain to initialize the pool.

A new utility called engine-backup has been introduced to backup and restore a complete ovirt-engine environment. This utility is provided to simplify the backup and restore process.

It is now possible for administrators to copy permissions of a virtual machine to a template, and from a template to a virtual machine. This feature enhances usability so administrators do not have to manually add permissions to all virtual machines.

This update changes a move disk operation to a copy and delete disk operation. This enables the virtual machine to start in the target storage domain before removing the disk from the source storage domain, which makes the process more efficient and easier to troubleshoot.

It is now possible to save the state of a virtual machine's memory when creating a live snapshot, and restore a virtual machine's memory to the state captured in the snapshot.

Users can now search for disks by their ID. This feature is useful as a disk's name is not unique throughout the system, for example when several virtual machines are created from the same template it creates duplicate disks. The disk ID can be used to distinguish between disks.

The Storage tab now has a new column showing the total size of each storage domain.

The CPU Shares option allows users to specify the priority of CPU utilization between virtual machines with shared CPU resources. This option can be found in the Resource Allocation tab of the New / Edit Virtual Machine window in the administration portal.

CPU shares are evaluated by the hypervisor's kernel, for currently running virtual machines. The share's value can be a predefined one, or a custom number between 0 and 262144. By default this is disabled. A virtual machine marked with high priority will receive twice the amount of CPU resources allocated to a medium priority machines, and four times the amount allocated to a low priority machine.

The 'In Use' column indicated the LUN is a part of a VG (which contains data), but it was confused with being used by the engine (the LUN is a part of an existing storage domain). This column has been removed to decrease ambiguity.

The 'Allocated' field has been added to the General tab under the Storage tab, which displays the total space that has been assigned to virtual machines in the selected domain.

It is now possible to select multiple virtual machines and open all their consoles at once. Note for "native" console invocation this opens multiple browser windows.

Users can now add new bonds from the manager administration portal, in addition to the five predefined bonds for each host.

Previously when creating a virtual machine based on a template, the template was locked in a way that prevented creating additional virtual machines from that template at the same time. This lock has been removed, and users can now create multiple virtual machines from a template in parallel.

A pattern-based solution has been introduced for changing the name of virtual machines created from a pool. Users can now specify a mask inside the pool name for the indexes that are included in the generated names of the virtual machines in the pool.

There is now a "Start in Pause Mode" checkbox in the regular Edit VM window instead of only in the Run Once window. One can define a virtual machine as always started in paused mode for opening the console at the beginning of virtual machine's boot process.

It is now possible to assign multiple virtual machines from the same pool to a user, which is useful when a user needs multiple instances of the same system. To do so, specify the maximum value of assigned virtual machines per user in the Edit VM window in the web portals, or use the REST API to check out virtual machines up to the specified limit.

Users can now define a gateway for each logical network on a host. This feature is beneficial for deployments in which a host has more than one network device. Any traffic that has to be returned to a network outside the host's subnets will be routed back via the device through which the traffic came, instead of the host's default gateway.

It is now possible to edit and manage iSCSI storage domain connections via the REST API. This means that users can now edit the logical domain to point to a different physical storage, which is useful if the underlying LUNs are replicated for backup purposes.

The host's hardware tab now displays information about host bus adapters (HBAs), including type, model, manufacturer of the device, world wide name and world wide port name device and port identifiers. This enables identification of hosts with specific Fiber Channel or iSCSI devices.

Red Hat Enterprise Virtualization now supports cloud-init, which facilitates the provisioning of virtual machines by performing the initial setup of networking, SSH keys, timezones, user data injection, and more.

Previously, the Virtual Machines tab under the Storage tab displayed only the top ten virtual machines which consumed the most storage resources. This limitation has now been removed, so all virtual machines that reside on the storage domain can be retrieved.

Quota management, which was offered as a technology preview in version 3.2, is now officially supported Note that quota in version 3.3 does not include REST API support.

This feature makes it possible to view the content of the welcome page for the Red Hat Enterprise Virtualization Manager in accordance with the desired locale. With this feature, the welcome page now automatically determines the locale of the browser and presents content in the language of that locale. A button has also been added to the welcome page that allows administrators to select their desired locale. This preference is stored across sessions.

Previously, it was not possible to alter the parameters of connections to storage domains (NFS, POSIX, and local) in Red Hat Enterprise Virtualization. This update provides administrators with greater management capabilities that allow them to alter connections to storage domains after they have been created. These capabilities are available both in the Administration Portal and via the REST API.

VirtIO Console devices can now be attached to or detached from a virtual machine. The feature is exposed as a checkbox in the VM/Template/Pool popup dialog in the Console tab and in the REST API.

Red Hat Enterprise Virtualization Manager now supports parallel NFS (pNFS) storage. pNFS as a specification supports blocks, objects, and files. A hypervisor that uses pNFS to access a virtual machine data store can take advantage of multiple connections per session, which can be distributed across multiple interfaces.

This feature provides administrators with four additional VDSM hooks: before_disk_hotplug, after_disk_hotplug, before_disk_hotunplug, and after_disk_hotunplug. With these hooks, administrators can now add user actions before and after a disk is hot-plugged or hot-unplugged.

A new 'Create Snapshot' button has been added to the action panel of the Virtual Machines tab, and as a context menu item when a virtual machine is selected.

Previously, it was not possible to use encrypted channels as a policy for live migration with virtual machines and clusters. This feature introduces the ability to perform a 'tunnel migration' of clusters and virtual machines over encrypted channels, improving security.

Red Hat Enterprise Virtualization Manager includes a new scheduler to handle the placement of virtual machines, allowing users to create new scheduling policies and write their own logic in Python and include that logic in a policy.

The new oVirt scheduler serves virtual machine scheduling requests during migration or while virtual machine are running. Scheduling is performed by applying hard constraints and soft constraints to get the optimal host for that request at a given point of time.

Scheduling Policy Elements

* Filter: a basic logic unit that filters out hypervisors that do not satisfy the hard constraints for placing a given virtual machine.

* Weight function: a function that calculates a score for a given host based on the internal logic of the function. Weight functions are a method for implementing soft constraints in the scheduling process. Since these are weights, a lower score is considered to be better.

* Load balancing module: code implementing logic for distributing loads. To date, the definition of a load was mostly CPU related, which could be resolved by migrating a virtual machine. The new scheduler allows users to write their own logic to handle other load types such as network and I/O by other means such as integrating with third-party systems.

Scheduling Process Outline

Every cluster has a scheduling policy. To date, there were three main policies - None, Even distribution, and Power saving - and now administrators can create their own policies or use the built-in policies. Each policy contains a list of filters, one or more weight functions and a single load-balancing module. The scheduling process takes all relevant hosts and runs them through the relevant filters of a specific policy. Note that filter order is meaningless. The filtered host list is then used as an input to the relevant weight functions of that policy, which creates a cost table. The cost table indicates the host with the lowest weight (cost), which is the optimal solution for the given request. Multiple weight functions may be prioritized using a factor.

Adding User Code

See Bug #912059 for more information on adding code.

Important Notes

- New scheduling policies created by administrators are not validated by the system. This may end up with unexpected results, so it is highly important to verify that a new policy is not introducing issues or instability to the system.
- User-provided code is unsupported.
- Using user-provided code may affect performance, so administrators are advised to carefully test their code and the general performance changes.

Previously, multiple monitor support could misbehave on Red Hat Enterprise Linux guests. 

This occurred when a virtual machine used a single QXL device (SPICE virtual graphics card) which did not provide sufficient memory for the virtual machine.

Now, more VRAM is available for QXL devices, so multi-monitor is supported on Linux machines using a single QXL device. This is the default setting for Linux guests using the SPICE display type.

When a new storage domain is created, the host nominated as the Storage Pool Manager of the data center is automatically set as the default host, preventing performance issues with random selection and operations running on incorrect hosts.

Red Hat Enterprise Virtualization Manager now supports Trusted Compute Pools based on the OpenAttestation project. This feature allows administrators to build trusted clusters based on Intel's hardware-based security features. A trusted cluster includes only hosts verified by Intel's OpenAttestation, securing virtual machines infrastructure and establishing the foundation for a secure enterprise stack.
Note that in order to be able to use this feature in RHEV, the admin needs to install Intel's OpenAttestation RPMs.

Previously, it was not possible to view the available space on a storage domain in the storage domain drop-down lists of the Add Virtual Disk, Copy Disk(s), Move Disk(s), New Template and Import Virtual Machine windows. With this feature, administrators can now see the available space on a storage domain alongside the name of that storage domain in the above drop-down lists.

The login page in the user portal can now display a message of the day, which allows administrators to show users important system messages. To set this message, use the UserMessageOfTheDay option on engine-config.

Previously, the user interface plug-in API did not support integration with the main Events tab and its individual sub-tabs, both of which are standard tab types in the Administration Portal. This resulted in user interface plug-ins being unable to use 'add action button' functionality for these tab types, including context-sensitive buttons located in the action panel of the main tab and sub-tabs and context-sensitive menu items triggered by right-clicking the main tab or sub-tabs. With this update, the plug-in API has been updated to enable 'add action button' for the above tabs, allowing user interface plug-ins to integrate with the Events tab and its sub-tabs via the action button concept.

Previously, it was possible to both pin a virtual machine to a specific host and enable high availability for that virtual machine at the same time. This resulted in highly available virtual machines not being migrated when required due to being pinned to a host. With this update, high availability and host pinning are now mutually exclusive options and cannot be enabled at the same time for a virtual machine.

Red Hat Enterprise Virtualization now provides support for more guest operating systems, including Ubuntu 12.04 LTS, Ubuntu 12.10, Ubuntu 13.04, Ubuntu 13.10, and SUSE Linux Enterprise Server 11.

AuditLogDirector.log(*) methods now update engine.log, so every event that appears in the administration portal's event viewer is logged permanently to the log file.

Initial support for the Foreman host provider is offered as a technology preview. You can add the new Foreman provider in the administration portal, and use the "Add Hosts" window to select a host provided by Foreman on Red Hat Enterprise Virtualization Manager.

This feature makes it possible to use virtio balloons for memory optimization.

Every virtual machine in clusters with a compatibility level of 3.2 and higher includes a balloon device unless that device is specifically removed. This balloon device requires guest drivers and a guest agent to control the balloon size.

Ballooning optimization is a cluster-level policy attribute, which is disabled by default. To have a balloon running, the virtual machine must have a balloon device with the relevant drivers, and the cluster it belongs to must enable ballooning optimization. Each host in the cluster receives a balloon policy update when the status of that host changes to 'Up'. A manual option is also available for emergencies that makes it possible to force an update for a specific host.

Once this is set, MoM will start ballooning where and when possible to allow memory over-commitment, with a limitation of the guaranteed memory size that every virtual machine has.

Note that in some scenarios ballooning may collide with KSM. In such cases, MoM will attempt to adjust the balloon size to minimize collisions. Additionally, in some scenarios ballooning may cause sub-optimal performance for a virtual machine. Administrators are advised to use ballooning optimization with caution.

When ballooning is enabled for a cluster, appropriate messages now appear in the Events tab of the administration portal when the following errors occur:
* The balloon is requested but the balloon driver on the virtual machine is not responding.
* The balloon is inflated, but the Memory Overcommit Manager (MOM) has lost control over the balloon device for various reasons such as failure of the guest agent.

When a virtual machine is running on a host which cannot provide the amount of guaranteed memory specified in the virtual machine configuration, RHEV engine issues an alert. In a future release, this issue will be handled by migrating the virtual machine to a host which has sufficient memory.

Previously, all options in the New Virtual Machine and Edit Virtual Machine windows, including advanced options, were visible by default. This resulted in it being difficult to navigate these windows when only the basic options were required. With this update, advanced options are now hidden by default, and a new 'Show Advanced Options' button has been added to allow users to display advanced options as required.

Red Hat Enterprise Virtualization now provides technology preview support for importing and exporting images and disks from the OpenStack Glance domain provider.

The previous separation of flows for creating servers and desktop has been replaced with only one "New VM" window, which also provides a checkbox to optimize it for either server or desktop. This replaces the previous behavior where it was not possible to do mixed configurations (for example, a server with a soundcard).

ISO domains can now be created on any supported file-based storage, including NFS, POSIX, GlusterFS and local storage. This feature decreases dependency on NFS which was previously the only supported format, and allows creating ISO domains on local storage which is ideal for small-scale deployments or evaluations.

When a variable was defined in a recursive way in more than one configuration file, the parser tried to resolve the value of the variable in an infinite loop. The method used for parsing configuration files has been changed to rhevm-lib, so now configuration files are correctly parsed.

When running rhevm-image-uploader on a system upgraded from Red Hat Enterprise Virtualization 3.1 a warning was issued that a deprecated configuration key had been found. The legacy keyword 'rhevm' has been removed from the tool's configuration files, so the warning is not issued anymore.

On some hypervisors the sosreport collection could take more than 900 seconds, which is the default duration to keep the ssh client connection alive. Now, calling ssh with the -oServerAliveInterval=600 parameter prevents the connection timeout for the number of times specified using ServerAliveCountMax (default is 3 times).

The log collector was limited to 1000 hypervisors, now that limit has been removed.

Previously when a hypervisor was added to the manager using its FQDN, specifying its IP address using the -H option caused an error "The following hosts are not listed as hypervisors". Now, providing an FQDN that resolves to IP vice versa, either one can be specified to add a hypervisor to the manager.

When rhevm-log-collector attempted to collect logs from Red Hat Enterprise Linux hosts, the process was suspended indefinitely while processing the SELinux sos plugin. This update adds a "-t" flag to the ssh command used in the log collector, which provides a controlling TTY to finish the rhevm-log-collector process after the host sosreport is collected.

On systems running SELinux, all processes and files are labeled with a label that contains security-relevant information. This information is called the SELinux context. Previously the log collector provided information whether the SELinux is enabled or not, but did not provide SELinux context for files. The fix adds a new output file of the SELinux context in specific folders including /rhev, /etc, and /var).

Previously rhevm-log-collector did not appear to respect the --insecure option. It was impossible to work with the insecure option, except with https. Now, specifying --insecure mode will work with http.

Previously it was possible to access the database through a UNIX socket as the PostgreSQL sosreport plugin did not handle identity authentication correctly. As a result, the rhevm-log-collector did not correctly gather an engine database dump. Now the access for the 'engine' user is limited to TCP connections, and the rhevm-log-collector always uses the database host.

This update adds JBoss stack trace support to the rhevm-log-collector. The "jbosstrace" option sends a SIGQUIT signal to the JBoss process launched by ovirt-engine, causing a dump of the stack trace to /var/log/ovirt-engine/console.log. Other JBoss instances will not receive that signal.

Previously the logging system wrote all messages to the standard error. Now only errors are logged to the standard error, all other message are logged to standard output.

rhevm-log-collector ignores the "--ticket-number" option, which should create a tarball with a ticket number. This has now been fixed, so specifying --ticket-number option will generate a correct tarball name.

The --local-tmp option should create a tarball in the specified directory, however the tarball was always created under /tmp. Now, the --local-tmp option behaves as expected, and generates a tarball in the specified directory.

The openssh-clients package was missing on a Red Hat Enterprise Virtualization Manager system installed using the minimal-installation option. As the rhevm-log-collector utility has a dependency on the openssh-clients package, it could not collect logs from hypervisors when openssh-clients was absent. This update adds the dependency to the installation spec file, so openssh-clients is installed together with the Manager, and the log-collector can gather hosts sosreports.

Each log collection overwrote the existing log file. This made debugging issues difficult, especially when multiple attempts were executed. Now, log messages are written into timestamped files by default, preventing the overwrite of existing files. The default location of the logs has been changed from /var/log/ovirt-engine to /var/log/ovirt-engine/ovirt-log-collector. In addition, a log rotation was introduced to compress old logs.

The hookLevel parameter must be in lower case for creating a correct path for the verb hookAdd. This parameter has been changed to lower case, so adding a hook works as expected.

Red Hat Enterprise Virtualization does not support static (unmanaged) iSCSI connections presented to the system by iSCSI HBA cards. To work around this issue, VDSM exposes iSCSI HBA cards as Fiber Channel devices.

When an LVM clustered VG was visible on the SPM host, it was impossible to create a new storage domain. This limitation has been removed and LVM clustered VG are now ignored by VDSM.

QEMU sometimes returned a value for the highest allocated extent of a volume that was greater than the capacity of the qcow2 volume. In such a case VDSM attempted to extend the volume with every run of _highWrite. It did not ensure that the highest allocated extent was within the capacity of the volume before proceeding. Both _highWrite and _onAbnormalStop now share the same logic about the volume extension.

Previously ovirt-hosted-engine setup failed to mount a storage domain because VDSM used the wrong mount path. The mount path is now correct, so the setup executes as expected.

Occasionally, an incorrect or missing BOOTPROTO line in the ifcfg file of a vlan device would propagate to the underlying device. Consequently, adding a vlan network on top of a non-vlan network erased the dhcp setting on the latter.

VDSM now maintains the defined boot protocol when adding a new network on top of an existing one.

Previously storage operations (such as creating virtual machines from templates) were slow due to various validation checks and refreshes which were redundant. These checks have been removed, so storage performance has improved.

Previously, if sanlock was already running before installing vdsm, the daemon kept running with the same privileges it had before the service started. This prevented some operations on File Domains to be executed (sanlock lockspace add failure). 

Now, when vdsm is installed, sanlock is restarted so the daemon can pick up the required privileges.

An unhandled EPERM exception while attempting to unregister a file descriptor led to the virtual machine Channel Listener thread dying. VDSM now uses the capability of epoll to detect and unregister closed file descriptors, which prevents unregistering file descriptors which are already assigned to different objects.

Attempting to call the shutdown() method on a virtual machine which was in a 'Down' state resulted in an exception if called through xmlrpclib. This fix adopts the convention used by similar APIs such as migrate() and returns errCode['noVM'] instead. Virtual machines can be successfully shut down with no exceptions.

LvmCache did not invalidate stale filters, so after adding a new FC or iSCSI LUN to a volume group, hosts could not access the storage domains and became non-operational. Now, all filters are validated after a new device is added and before the storage domain is extended, so hosts can access storage domains which have been extended.

Previously, moving one storage domain to maintenance mode triggered an SPM contention between hosts, and caused the entire data center to become unusable. This was due to a regression which caused a KeyError on isoprefix for inactive domains. This update fixes the KeyError, so now when a storage domain is moved to maintenance the rest of the storage pool is still usable.

VDSM did not start properly upon clean installation, which could lead to failure when deploying the hosted engine. This update changes the return method from the vdsmd start function, so VDSM starts properly.

On some overloaded systems the Storage Domains' latency timings were reported with inaccurate values. Now VDSM reports the most accurate latency values in all cases.

Without manually configuring libvirt during an upgrade, VDSM maintained an older configuration of libvirt. Now, automatic libvirt reconfiguration is added to the %postun scope of vdsm.spec, so VDSM reconfigures libvirt if necessary during an upgrade.

For each storage domain, VDSM initialized a process pool of 10 processes, and each pool was related to the storage domain by its UUID. The parsing of the UUID was wrong, and caused a wrong number of process pools to be created. This update fixes the parsing of the storage pool path and uses the right UUID for the pool's key, enabling continual creation of process pools. This fix limits the initialized process pools to one for each storage domain as required.

When a hypervisor was rebooted, all logical volumes which were part of an FC storage domain were automatically activated. This caused some issues as logical volumes should by activated only on request of the engine, and deactivated immediately when they are not needed. These logical volumes did not pick changes done by the SPM on the storage, which could lead to data corruption when a virtual machine wrote to the logical volume with stale metadata. The fix checks all VDSM logical volumes during LVM bootstrap and deactivates them if possible. Special logical volumes are refreshed, since they are accessed early when
connecting to storage pool, before LVM bootstrap is done. Open logical volumes are skipped because they use correct metadata when opened.

A previous commit changed multipath names to be returned as full paths. Consequently, expanding an iSCSI storage domain produced a device identifier mismatch error. Now, device GUIDs are converted into /dev/mapper paths, so extending block storage domains succeeds.

When the connectivity to the storage was restored and the host was recovered, no virtual machine could be run. All of them failed to run with the error "Image path does not exist or cannot be accessed/created". This was caused by the run VM command being issued before the connect storage pool was completed. This was fixed by allowing the virtual machine to run without being connected to the pool.

Live migration was not backwards compatible, so virtual machines could not be migrated between hosts with older VDSM versions and 3.3 beta hosts. This has been fixed, so live migration works as expected.

A new SELinux context which allows qemu-kvm to generate coredumps has been added. This is a temporary solution until coredump is replaced by automatic bug reporting tool.

An application with non-ASCII characters in its name, installed inside a virtual machine, triggered an error when using the 'vdsClient getVmsInfo' command to list installed applications. This update allows proper manipulation of UTF-8 characters out of ASCII range. Now 'vdsClient getVmsInfo' shows the expected information even when non-ASCII characters are present in applications' names.

Cancelling a migration returned an error code in the wrong format. It has now been fixed, the error message now appears in the format of {'status': {'message': 'Migration canceled', 'code': 47}, 'progress': 1}

After a storage outage, the master domain became locked and could not be reactivated. With this update, VDSM prevents the master domain from being stuck in a locked status if the reconstruct process fails, so hosts can be activated successfully.

Several regressions were introduced as part of adding a daemonAdapter process that starts VDSM respawn with more management control, including not handling crashes of the adapter, and leaving the lock file on the system. This prevented new VDSM instances from starting, which caused host installation to fail. The code has been rewritten to handle these issues, and now VDSM starts correctly.

vds_bootstrap, that is still deployed on version 3.0 setups, expects netinfo.getVlanID to return strings, when it now returns integers. This update uses vdsm_compat/netinfo to convert the integers to strings, so vds_bootstrap can successfully add hosts when a VLAN is defined on top of its NIC.

The domain monitor thread did not provide notifications when going from an unknown state to a known state. The thread would start with a bogus valid state, before checking the domain. If the first domain check found the domain as valid, no notification was emitted. If there were paused virtual machines while connecting to the pool, they were not unpaused as they should be. 

Now, the monitor thread detects the first state change and emit notifiction with the domain state. Domain state changes are now registered before the monitor threads are started. Paused virtual machines are now unpaused when VDSM starts, and the domain is valid.

A misleading python traceback related to corrupted or missing images has been converted into a more meaningful error message ("Volume is corrupted or missing").

VDSM now ensures that the rpcbind service is present and available in order to handle NFS (version 3) Storage Domains.

Before a host took the SPM lock it did not connect to a storage pool, and VDSM returned an empty task list, which caused the engine and hosts to become out of sync. Now, VDSM only retrieves a task list when an SPM exists and has been recovered.

When trying to remove a vm that had a broken symlink, the system displayed a key error from VDMS. The root cause of this was live operations and deleteImage operations sent by the engine, when some (top) LV's were still open and failed to be removed. This was fixed in previous code changes.

The LVM filter has been updated to only access physical volumes by full /dev/mapper paths in order to improve performance. This replaces the previous behavior of scanning all devices including logical volumes on physical volumes.

Previously, accents in the virtual machine name caused migration to fail. This update adds UTF-8 encoding of the configuration XML on the destination to handle non-ASCII characters, so virtual machines with accents can now migrate successfully.

getImagePath did not work on Gluster storage domains because its domain directory's base paths for Gluster domains and other file-based domains were different. Consequently, disks could not be deleted from a GlusterFS storage domain. This update uses the correct domain directories to get the images, so disks can be successfully deleted.

Network hierarchies were processed in isolation, so if a network was added over a device in use by another network, adding the new network would momentarily take down the device to reconfigure it, regardless of whether reconfiguration was necessary. This update introduces configuration matching to check if a device needs reconfiguring and only interrupts its connectivity in such cases. Now, networks can be added over devices without affecting other networks as long as they do not alter the common devices' configuration.

The Lease Timeout and IOOP Timeout values were not maintained when the Master Domain Role was migrated from one Storage Domain to another. The issue has been resolved and it is now possible to deactivate the current Master Storage Domain without losing the configured timeout values.

Non-numeric characters were not allowed for network names, but some VLAN interfaces with the string "-fcoe" are required by the bnx2fc driver. Consequently, VDSM threw an exception when trying to activate hosts with these VLAN names. This update introduces "hidden_vlans" key configuration, allowing VDSM to skip VLAN devices with names which are not in the expected format.

Editing a storage domain while there were running tasks caused a timeout for other tasks. The setStorageDomainDescription now holds a shared lock instead of an exclusive one, so other tasks can run while the storage domain is being edited.

A previous commit to automatically unpause virtual machines when a storage domain became active unexpectedly caused an exception on removeVmFromMonitoredDomains during virtual machine shutdown. Now, the list of virtual machine domains is maintained inside the virtual machine object, so virtual machines can be powered off without error.

Previously VDSM bonding handled adding, editing and removing bonds in a non-deterministic order. Consequently, the bond to be created with the new name could be processed before the one with the old name was destroyed. The bond slaves would not have been freed yet and the new name bond would fail to be created. Now, VDSM first process all bonds to delete, then all the bonds to edit, and finally all the bonds to add. As a result, bond names can be modified and slaves can be taken from bonds to be removed in the same setupNetworks command.

In some cases, long tasks executed by VDSM as SPM using external tools, for example, dd could have been able to survive a VDSM restart or crash, possibly leading to corruptions. Now the tasks are prepared making sure that external processes are terminated if the parent (VDSM) dies unexpectedly, preventing the possible corruptions.

The log collector tool now collects /var/log/sanlock.log from hypervisors.

Users can now use the "ksmEnabled" parameter on vdsClient to enable or disable Kernel Samepage Merging (KSM) control by the Memory Overcommit Manager (MOM). By default, KSM is enabled.

When running a sosreport on a host or log collector on the manager, an unhandled exception with dumpStorageTable occurred when the poollist field was empty, causing sosreport to crash. This was fixed by proper handling of the exception when poollist is empty, so this error no longer appears.

When attaching a NIC to a virtual machine, port mirroring was part of the process regardless of whether it was a newly created virtual machine, or if the NIC was attached to the virtual machine after VDSM restart. Consequently, port mirroring was reapplied as many times as VDSM was restarted. Now, port mirroring is only set when attaching the NIC to a virtual machine that is not in a recovery process, so virtual machine port mirroring is only set as many times as intended by the user.

When vdsmd was down the first getConnectedStoragePoolsList call resulted in a 'Connection Refused' trace being shown. To avoid this trace being picked up by the automatic bug reporting tool, the connection is now tested first, catching any errors before continuing.

The v3ResetMetaVolSize argument was changed from volume object to volume UUID, which triggered attribute errors during the execution. Consequently, upgrading a storage pool or domain from 3.0 compatibility mode to 3.1 failed. This update reintroduces volume objects as accepted arguments and correctly handles such exceptions.

Trying to hot unplug a disk from a virtual machine failed with the error "'Vm' object has no attribute 'addVmToMonitoredDomains'". This is fixed by maintaining a list of the virtual machine's domains inside the virtual machine object, so hot unplugging a disk now works as expected.

Previously VDSM was stuck in a recovery phase and unable to initialize after a libvirt crash, and an exception on the _recoverExistingVms thread. Now, after this happens, VDSM attempts to restart until it succeeds.

An alignment scan failed on a disk that had been partitioned because VDSM failed to deactivate the volume. This was because the engine expected a dictionary word instead of an integer for the status of getDiskAlignment. This has now been fixed, and the alignment scan succeeds as expected.

The client IP address used to be logged when SSL errors occurred, but this stopped working with the introduction of cached SSL sessions. This update modifies the SSLServerSocket class that was introduced to implement cached SSL sessions. Now when an SSL error is detected while attempting a connection, the IP address is added to the message of the exception and logged.

Upon receiving an error from libvirt VDSM now pings it
using getLibVersion, and treats libvirt as being down only when the ping failes. This feature is useful because libvirt error codes do not distinguish recoverable from non-recoverable libvirt errors.

Previously, the getImagesList command on an NFS domain returned an ordered hash with each letter on a new line. This problem has been fixed, now getImagesList retrieves a correctly rendered list of images.

Previously, libvirt-reconfigure automatically restarted the libvirtd and supervdsm services after the reconfiguration. Now, instead of restarting both services, users receive a message requesting them to manually perform the restart by using the libvirt-configure-services-restart command. For backward compatibility with systemd, this also updates the systemd-vdsmd script to restart the libvirtd and supervdsmd in previous versions.

After a failure in creating a snapshot of a virtual machine with several disks, VDSM could not roll back the unfinished tasks. This update adds code to handle this failure, and enables parent volume rollback.

If the vdsmd service was restarted simultaneously more then once, it would respawn and have several VDSM processes running. The earlier VDSM process was not restarted using the init script, and it was bound to port 54321, preventing the newly spawned VDSM from binding to that port. Now, the init script is locked when vdsmd starts up until it is stopped, so the port binding error no longer occurs.

After a failed live storage migration, attempting to merge the snapshots caused logical volume metadata corruption. This problem has been fixed, so the corruption no longer occurs, and the disks can be moved after an unsuccessful live migration.

When processing a non-existing network, VDSM assumed that it was for creation and unexpectedly fail when it could not find the network definition. Now VDSM checks if the non-existing network is marked for removal. As a result, trying to remove a non-existing network will properly notify users that the network is missing.

For various reasons, such as memory over-commitment usage, VDSM could end up reporting negative memory values, which could affect scheduling decisions. The new memFree option reports a proper value, so scheduling now gets the proper information.

Previously supervdsm, which is responsible for all privileged options, was managed by the unprivileged vdsm process. This led to races between new and old instances of the processes. Supervdsm now runs as a service external to vdsm, additionally updates have been made to handle crashes and re-establish communication between vdsm and supervdsm after failures.

Previously, the HotPlugDiskToVmCommand simultaneously updated the 'isPlugged' and 'bootOrder' properties of all devices attached to virtual machines. This could cause a race with another thread which handled the hot plug of another disk for the same virtual machine, for example it was not possible to hotplug a disk if the previous hotplug failed. Now, HotPlugDiskToVmCommand only updates 'isPlugged' for the device that was plugged by this command, so this race does not occur.

Virtual machine migration failed when the destination host had connectivity problems to storage domains in the data center, even irrelevant ones. This caused the host to be stuck in a 'preparing for maintenance' state. Now, a destination host with connection problems to irrelevant storage domains can successfully run migrated virtual machines.

Failed to connect to Super VDSM via getDeviceList command.

This happened because each call to supervdsm checks the ctime of the process, which is used to verify that the pid of supervdsm, is the original pid that VDSM starts when executing supervdsm, otherwise the pid was taken by another process and supervdsm needs to be restarted. When running on fake kvm the ctime value changes during the process run. When ctime is different from the original that VDSM saved, VDSM relaunches new instance of supervdsm. 

This has been fixed by using proc/pid/stat field 21(starttime) to define the time the process started.

When the guest agent socket could not be created, VDSM could prevent the release of the virtual machine, so the virtual machine is stuck in a powering down and unresponsive state. Now, when the guest agent socket is not created, VDSM will kill the virtual machine process and restart the VDSM service, and log the occurrence.

A bond defined as a hidden nic, was not hidden and could be selected. Customers may now define the hidden_bonds entry in vdsm.conf. If a bonding device is defined there, it is not reported by VDSM.

The VDSM tool can now be used offline for tasks that do not require VDSM to be configured or up. This tool can be used to return the VDSM ID, check its status, start and stop VDSM, and validate certificates.

After the Storage Pool Manager (SPM) lost connection to the storage and became non-operational, the expected behavior was for another host to take its place as the SPM, but this did not happen. This was because SuperVdsm was not passing kwargs to the fuser. This has now been fixed, so when the SPM becomes non-operational a failover takes place.

VDSM now uses the reload option for multipath rescan in order to view a new LUN that was added to an already connected target.

prepareVolumePath looked for the 'path' key in specParams without looking at the path in the device itself, causing empty paths in specParams to override the device path. Consequently, the RHEV-toolsSetup ISO which provides virtio-drivers was not automatically attached to Windows virtual machines. Now, device.path is added to the checking mechanism to avoid overriding a valid path. The RHEV-toolsSetup ISO is now automatically attached to Windows virtual machines.

Previously, removing a template with dependent virtual machines failed because of a typo in the code. This has now been corrected, and fake parameters are also removed before the template is deleted.

After deleting the iscsi storage, ISCSIADM operations got stuck and wouldn't end.This caused connectStoragePool/GetDeviceList  to fail on timeout. This has been fixed by removing forceIscsiScan() as iscsiadm already does that when refreshing sessions, and rewriting the logic with a simpler implementation.

Attaching a storage domain to the data center failed with an AcquireLockFailure error caused by an incorrect lease parameter value. The storage domain now uses the correct default lease value, so attaching the storage domain works as expected.

After deleting a snapshot that was created during a failed live storage migration, the virtual machine failed to run with a "Meta Data key not found" error. This was caused by merging snapshots on the SPM and then trying to run the virtual machine on a host without refreshing its knowledge of the changes, rendering the LVM cache out of sync. This has now been fixed, so virtual machines can run after snapshot deletion.

A malformed error was produced if a failure occurred while creating a volume: 
"Drive image file %s could not be found" 

This error could occur in any flow that creates a volume, and is not Live Storage Migration specific.

The error text now reads:
"Drive image file could not be found"

libvirt did not load the SELinux driver by default as it was disabled in the configuration file. This caused virtual machine migration to fail. With this update, the policy has been modified so VDSM reconfigures the SELinux driver upon startup, and virtual machines can be migrated.

When VDSM's connection to libvirt was broken, VDSM did not initiate the shutting down of the hosts when there were no running virtual machines. This update removes the clientIF instance dependency from libvirtconnection, and sends a SIGTERM to the VDSM process which triggers the host shutdown.

Attempting to delete a snapshot based on a preallocated disk failed. Instead of deleting the snapshot, the source volume was deleted, so the virtual machine could not run. Now, during the snapshot process it creates a new volume which shrinks after the snapshot is merged back to the virtual machine. As a result, deleting a snapshot based on a preallocated disk works properly.

Previously, VDSM failed to refresh the list of images for a local ISO domain. Now, the local ISO domain refreshes after files are deleted from it or copied to it.

When the vdsm process came up, the VDSM log did not contain the new PID of vdsmd. Now when VDSM starts it now outputs to the VDSM log:
"MainThread::INFO::2013-05-05 11:57:38,609::vdsm::89::vds::(run)
(PID: 14576) I am the actual vdsm...".

When a new logical volume was created, it was accessible by hosts but links to it were not created. QEMU attempted to access the volume through a non-existent link, and failed. The workaround was to restart vdsmd, which could take a long time in large environments. Now, the symlinks are properly created on the hosts, so virtual machines with new volumes can be started on any host when the SPM is made aware of something new in the environment.

VdSM was failing to perform the "getStorageDomainStats" command and failed with a "Failed to GetStorageDomainStatsVDS, error = Domain is either partially accessible or entirely inaccessible". The problem was that the storage domain was cached (together with the path to reach it), so getStorageDomainStats was always executed using the old path. This has been fixed by removing pre-existing domains when connecting to a storage server.

Transient disk support has been added to VDSM. A transient disk is a temporary qcow layer based on a volume that is attached to a virtual machine in read-write mode. All changes stored in the transient disk is lost when the disk is hot-unplugged or when the virtual machine shuts down. Its primary use is to access a disk snapshot (a read-only volume) in read-write mode.

VDSM can now monitor domains when not connected to a pool using the startMonitoringDomain and stopMonitoringDomain options. Monitoring results can be obtained with repoStats().

During hypervisor registration, errors sometimes appeared on the console. This made for an unpleasant user experience. The errors are now hidden.

The domain monitoring thread assumed that the list of monitored domains was the same as the list of attached domains, however this was not valid when any storage domain gets detached. In such cases, the thread was not stopped, and the host ID was still locked or in use. Consequently, users could not attach a previously-detached storage domain to a data center. This update adds a tag mechanism for pool-monitored domains, so now host IDs are correctly released, and unattached storage domains can be attached to data centers.

Exporting a template that had a virtual machine based on itself failed due to an attribute, safeUnlink, which had been replaced by rmFile. The reference to the safeUnlink has now been removed, so the template export succeeds.

To prevent issues with the limitation on the length of paths to UNIX domain sockets, VDSM now uses the UUID of virtual machines as a unique identifier for socket names.

Creating a new directory in the storage server caused clients to refresh the view of the mount, but the files were still stale. This caused live snapshot creation to fail. The stale NFS client view is now properly refreshed, and live snapshots succeed.

When deleting a volume and it failed to delete, the partially deleted image could still be seen in the images list. VDSM now ignores partially deleted images when retrieving a list of images.

The new feature updateDevice is a VDSM verb called by the Engine whenever the connectivity of a VM is changed, for example a change in QoS, mirroring, or target network. Three new hook points:

- before_update_device, 
- after_update_device,  a
- after_update_device_fail 

allow customers to tweak the host behaviour when the connectivity is changed.

Previously the size of a newly merged volume was the sum of the original volumes' virtual sizes. With qcow volumes on block devices the whole size of the new volume is not used, so now VDSM allocates only the actual size needed by the new volume.

After live resizing a virtual disk, the size change was not reflected on the virtual machine. Now, the volume is refreshed to update the volume segments, and to make the new size visible to the QEMU process before live extension.

A previous commit contained the 'verbose' option for the restorecon command for /var/log/core, which overloaded the console with messages during installation. This fix removes the verbose option.

Previously, VDSM failed to start after running libvirt-reconfigure, which is performed in cases of fresh installation. Now, it is possible to automatically restart VDSM and related services after reconfiguration of libvirt.

Domain monitoring stopped after disconnecting and reconnecting the storage pool, which yielded misleading "No space left" errors. This was caused by an anonymous thread running at unexpected times, and silently stopping all domain monitors. Now, the monitoring is stopped when appropriate, so users do not receive incorrect storage space limitation errors.

A disk that was live migrated between storage domains could not be hotplugged or hotunplugged to or from virtual machines. Now, VDSM updates the domains list for disks attached to virtual machines after a live migration, so the disks can be successfully hotplugged or hotunplugged.

An invalid argument error was displayed when creating the domain and VDSM tried to read the metadata.

This was fixed by checking that underlying file system supports direct IO:

- checking when creating a domain
- checking when creating a domain object

This is done so that posixfs domains are not allowed on top of targets that don't support direct IO. Direct IO is needed in order to read the metadata reliably in a clustered environment. It is also a required feature for Sanlock.

When trying to mount a file system and it failed in VDSM, incomprehensible error message were generated. The appropriate exceptions in the case of a mount error or permissions error to a mount point are now displayed.

The SASL downstream user name has been changed from vdsm@ovirt to vdsm@rhevh to prevent a postrun scriptlet error.

VDSM imports the BrokerLink lib with ovirt_hosted_engine_ha, which generates a lot of debug output in the default configuration. This update changes the log level to ERROR to limit the amount of logs.

There is now a dependency between vdsmd and supervdsmd. The supervdsmd service needs to be up before vdsmd starts up.

The udev cache was used for obtaining a device list, but this sometimes returned an incomplete list which led to missing volume errors. Therefore, creating several disks in parallel could fail.

LVM disabled the udev cache in the in RHEL 6.5, eliminating this error. However, on machines upgraded from earlier versions, lvm does not change the current configuration after an upgrade. It is the responsiblity of the system adminstrator to check the new lvm configuration and update the old confguration with new values. 

The udev cache is now disabled in vdsm lvm configuration, which ensures proper configuration also on upgraded machines.

Previously, virtual machines running on a host could not be shut down if they were migrated to a different host and the migration failed. With this update, the process for confirming completion of migration was revised to add a null check that ensures the destination VDS is valid. This addition makes it possible to shut down virtual machines on a source host even if migration fails.

Previously, it was not possible to relocate guests on a host if that host experienced complete power failure, including fence devices, until 19 minutes after the outage. This was caused by being forced to wait until three successive three-minute timeouts had elapsed and an additional 10 minutes for the transaction reappear. With this update, the method for determining whether a VDS action can be performed has been revised to allow execution of ClearNonResponsiveVdsVmsCommand while a host is in a reboot state, thereby shortening the time until virtual machines can be relocated.

Previously, the job and step tables would not be cleaned after the failure or completion of some tasks. This would result in numerous tasks being marked as still running in the tasks tab. With this update, periodic job deletion has now been updated so that jobs causing database errors will be correctly cleared.

Previously, Red Hat Enterprise Virtualization required rhevm-branding-rhev-1.4 rather than the latest version. This version contained a bug that would cause issues when logging in to the Administration Portal. With this update, the version of this package required has now been updated to rhevm-branding-rhev-3.3.1.

Previously, migrating a host designated as an SPM from one data center to another would successfully migrate the host, but leave the origin data center and any storage domains in an up state even when there were no available hosts. With this update, the method for moving hosts to maintenance was revised to correctly identify and handle SPM hosts.

Previously, apache-sshd would not flush streams or close streams after a command had been executed. This would result in data not being written to stream readers. With this update, the flush action has been made explicit, making it possible to correctly write data.

Previously, attempting to import a virtual machine with a memory snapshot into a storage domain in which no disk was set would result in an NPE error. This was caused by the import action assuming that the storage domain would exist. With this update, the existence of the storage domain is now confirmed and subsequent action handled accordingly, making it possible to successfully import virtual machines with memory snapshtos under these circumstances.

Previously, any changes to the CD on a running virtual machine made via the REST API would be discarded after the virtual machine was stopped. With this update, changes to the CD used by a virtual machine made via the REST API are persistent.

Previously, resource limits were not set to their hard limits for ovirt-engine. This would result in denial of service if multiple users performed numerous login and logout actions in a short space of time. With this update, resource limits have been set to their hard limits, preventing over-consumption of resources under such circumstances.

Previously, engine-setup would not detect if shared_buffers for PostgreSQL were below the active kernel.shmmax. This would result in engine-setup failing if kernel.shmmax was below that required to start the postgresql service. With this update, engine-setup detects if shared_buffers for PostgreSQL is below the active kernel.shmmax and returns a runtime error if that is the case.

Previously, running engine-setup with the option to create an answer file enabled would successfully create an answer file, but users would still be prompted to select whether the engine database was hosted locally or remotely when using that answer file in subsequent setups. With this update, the answer file now correctly records whether the user elected to use a local or remote database during the setup process.

Previously, the installation dates of packages listed in var/lib/ovirt-engine/setup-history.txt were locale-dependent. Now the dates are always written in the C locale.

Previously, the instructions for creating a database user provided when running engine-setup, rhevm-reports-setup and rhevm-dwh-setup were different. With this update, all instructions have now been standardized so that they refer to creating a role with an encrypted password.

Previously, a new window would be opened for downloading virt viewer when a user connected to the console of a virtual machine from the User Portal and virt viewer had to be downloaded. With this update, virt viewer is now downloaded in the same window or tab without creating any additional pop-ups or windows.

Previously, it was not possible to change the alias of a virtual disk while a virtual machine was running, even if that disk was deactivated. With this update, it is now possible to edit the alias of virtual disks while virtual machines are running.

Previously, attempting to add a host in an environment with Chinese language support would fail. This was caused by the user name 'root' being translated into Chinese, preventing authentication. With this update, the user name is no longer translated, making it possible to add hosts without issue.

Previously, users would not be prompted to install the SPICE ActiveX plug-in when attempting to open the console of a virtual machine when using Internet Explorer 11 on Windows 8. With this update, the invocation of the plug-in has been revised so that it appears correctly.

Previously, Windows users could only install Virt Viewer and USB Clerk via the ActiveX plug-in. With this update, users can now manually download installation files for these utilities from links on the web portal. However, users must update the rhevm-branding-rhev package to version 3.3.1 for these links to become visible.

Previously, clicking the OK button in the Edit Virtual Machine window on a running machine would produce an error when the VirtIO-SCSI Enabled check box was disabled, even when no options had been changed. With this update, validation of the VirtIO-SCSI device has been updated to prevent this error from occurring.

Technical Notes

Technical Notes for Red Hat Enterprise Virtualization 3.3 and Associated Packages

PressGang CCMS Build System

Legal Notice

Preface

1. Document Conventions

1.1. Typographic Conventions

1.2. Pull-quote Conventions

1.3. Notes and Warnings

2. Getting Help and Giving Feedback

2.1. Do You Need Help?

2.2. We Need Feedback!

Introduction

Chapter 1. RHBA-2014:0033 - ovirt-node

Chapter 2. RHBA-2014:0034 - mingw-virt-viewer

Chapter 3. RHBA-2014:0035 - rhevm-reports

Chapter 4. RHBA-2014:0036 - rhevm-dwh

Chapter 5. RHSA-2014:0037 - jasperreports

Chapter 6. RHSA-2014:0038 - rhevm

Chapter 7. RHBA-2014:0039 - rhevm-log-collector

Chapter 8. RHBA-2014:0040 - vdsm

Chapter 9. RHBA-2014:0227-01 - rhevm

Revision History