Chapter 1. RHEA-2018:1488 Red Hat Virtualization Manager (ovirt-engine) 4.2 GA

The bugs in this chapter are addressed by advisory RHEA-2018:1488. Further information about this advisory is available at https://access.redhat.com/errata/RHEA-2018:1488.

ovirt-engine

Previously, when the Manager was restored from a backup, it sometimes tried to start a virtual machine that was already running, believing that the virtual machine was down. This resulted in a second instance of the virtual machine being started on a different host. In the current release, the virtual machine will not be restarted automatically after restoration if it is already running somewhere else.

You can export virtual machines as named OVF/OVA files and save them to a specific path on a host or mounted NFS shared storage.

The Manager supports the ability to define logical networks, ports, and subnets that are not attached to the physical interfaces of the host. This enables the virtual network interfaces to create an isolated network within Red Hat Virtualization, allowing the virtual machines to communicate among themselves. This feature is configured through the REST API and the UI.

If an OVA file is accessible to at least one host in a data center, the OVA file can now be imported into the data center as a virtual machine.

When live or cold merge fails, snapshot disks may be left in an illegal state. If VMs with illegal snapshot disks are shut down, they will not re-start. VMs with illegal snapshot disks are now marked with an exclamation mark and a warning message not to shut them down.

Previously, if a host was placed in maintenance mode and migration was cancelled while at least 3 virtual machines were attempting to migrate to it, the host ended up in an ERROR state. In the current release, the host does not move into an ERROR state in this situation.

Previously, if a virtual machine was shut down during a live merge, an illegal snapshot disk was created, the live merge failed, and the virtual machine did not start up. In the current release, the virtual machine can be recovered with a cold merge.

Red Hat Virtualization now supports TLSv1.2.

Previously, the ForceRefreshDomainFilesListByDefault parameter used by virt-viewer was set to "True" by default, sometimes causing a flood of ISO domain refresh events in the logs if many console clients were connected. In the current release, setting this parameter to "False" reduces the number of logged refresh events.

Previously, the scheduler's Host Device filter policy unit did not filter for the host whose devices were to be used, excluding inappropriate hosts. Consequently, the Manager sometimes tried to run or to migrate a virtual machine with passthrough host devices to an inappropriate host. In the current release, the Host Device filter policy unit correctly filters for the host whose devices are to be used and the scheduler prevents the virtual machine from running on or migrating to an inappropriate host.

Previously, when creating large pools of virtual machines, the Manager assigned duplicate MAC addresses, causing network problems. In the current release, the problem has been solved by removing a problematic legacy method.

Previously, when creating a new virtual machine from a template with cloud-init configured, the virtual machine was created but the stored root password was not copied over. In the current release, the stored root password is copied to the new virtual machine.

Previously, when importing a template into Red Hat Virtualization 4.2 from a storage domain that had previously been attached to an environment earlier than version 4.1, a maximum memory error occurred. In the current release, the maximum memory limitation has been fixed.

The REST-API now supports LLDP. See www.ovirt.org/develop/release-management/features/network/lldp/ for details.

The precision of the rx_rate, tx_rate, rx_drop, and tx_drop of virtual and host network interfaces has been increased, enabling the REST API to generate more accurate network interface statistics.

Previously, live merge sometimes failed during "Merge Status", "Destroy Image", or "Destroy Image Check" commands because of network timeout, leaving the top volume in an illegal state. In the current release, the system calls are repeated until they succeed, so that network timeout does not cause live merge to fail.

A host must be reinstalled if you turn Kdump integration on or off, or change the kernel command line parameters.

All hosts in a cluster must be reinstalled if you change the firewall type of the cluster.

An exclamation mark icon in Compute > Hosts indicates that a host needs to be reinstalled. Details appear in the host's details view, in the Events tab.

iptables has been deprecated in Red Hat Virtualization 4.2 and will be completely removed in version 4.3. Administrators must switch to firewalld, which is introduced in version 4.2. Otherwise, the Manager will review the clusters every 30 days and raise warning events in the audit log. A warning message has been added to engine-config help for all iptables-related settings.

Previously, partial import of a virtual machine was only possible using the REST API. In the current release, an "Allow Partial" check box has been added to the Import Virtual Machine(s) window in the Manager.

Previously, when upgrading from Red Hat Virtualization 4.0 to 4.1, the maximum memory size allocated for virtual machines was 1TB, which sometimes resulted in a virtual machine's actual memory size exceeding the allocated maximum memory size. In the current release, the allocated maximum memory size adjusts to the actual memory size.

Previously, if a virtual machine with a strong positive affinity to a host was down, the affinity rules enforcer tried to migrate it, because it was not running on the specified host. When migration failed, the affinity rules enforcer tried repeatedly to migrate the same virtual machine, ignoring other virtual machines that violated affinity. In the current release, the affinity rules enforcer ignores virtual machines that are down.

The default value of JBoss's jboss.as.management.blocking.timeout option can be changed by creating /etc/ovirt-engine/engine.conf.d/99-jboss-blocking-timeout.conf with "ENGINE_JBOSS_BLOCKING_TIMEOUT=NNN", where "NNN" is the timeout value in seconds.

Previously, virtual machines that were paused for too long due to I/O errors were only killed when the engine tried to restart them. The current release adds a default setting, as part of high availability, that kills virtual machines that are paused too long because of I/O error, regardless of whether or when they will be resumed. This allows paused highly available virtual machines to migrate and be restarted.

Failed login attempts now appear in the audit log, with details and the user name that failed to log in.

This update prevents the user from clearing the "Allow Duplicates" check box in the Edit MAC Address Pool dialog or via a corresponding REST request if duplicate MAC addresses exist.

This release restores the ability to search Events.

This update allows you to search in Red Hat Virtualization Manager using more than one tag.

In the current release, a non-management network can be set as the default network route using the Manager, instead of selecting a custom property.

Virtual machines now stay operational when connectivity with CD-ROM images breaks. The error is reported to the guest operating system. Note that the configuration of the storage device may affect the time it takes to detect the error. During this time, the virtual machine is non-operational.

In the current release, the Manager's column widths, visibility, and order are persistent.

If a user runs 'engine-upgrade-check' without running 'engine-setup', a warning displays, informing the user that the system may not be fully up to date, even if the engine-upgrade-check says no upgrade is available.

Previously, in an emergency, users were required to shut down the hosts to preserve the data center. This caused running virtual machines to be killed by the systemd process without performing a graceful shutdown. As a result, the virtual machines' state became undefined, which led to problematic scenarios for virtual machines running databases such as Oracle and SAP.

In this release, virtual machines can be gracefully shut down by delaying the systemd process. After the virtual machines are shut down, the systemd process takes control and continues the shutdown. The VDSM is only shut down after the virtual machines have been gracefully shut down, after passing information to the Manager and waiting 5 seconds for the Manager to acknowledge the virtual machines have been shut down.

A virtual machine can now be sealed, in the Manager, in preparation for deployment as a template.

This release includes the ability to import Debian/Ubuntu virtual machines from VMware and Xen, which is available as a Technology Preview feature. From RHEL 7.4 virt-v2v can convert Debian and Ubuntu based virtual machines.

Known Issues:

1. virt-v2v cannot change the default kernel in the GRUB2 configuration and the kernel configured on the guest operating system is not changed during the conversion, even if a more optimal version of the kernel is available on the guest.

2. After converting a Debian or Ubuntu VMware guest to KVM, the name of the guest's network interface may change, and will need to be configured manually.

In the current release, the storage domain is prevented from going into maintenance if the OVF update fails. An optional check box to force maintenance mode, if desired, has been added.

The user can now decide whether a virtual machine should be warm or cold rebooted when started as "Run Once" in the Administration Portal. To facilitate this, the "Trap guest reboots" option has been renamed to "Rollback this configuration during reboots". This enables virtual machines to start on the same host when is it run as "Run Once" and then rebooted.

You can now view virtual machines that are pinned to a host, even when they are shut down.

A "Force Remove" button has been added to the Administration > Providers screen. Currently, it applies only to volume providers (Cinder). It removes the provider from the database, along with all related entities, such as storage domain, virtual machines, templates, and disks.

Previously, engine-backup always used the /tmp directory, which was problematic if the directory was full. In the current release, it is possible to change the engine-backup's temporary directory.

Previously, when creating a template from a virtual machine with the REST API, if the disk attachments were customized, the virtual machine could not be referenced by its name. In the current release, it is possible to create a template from a virtual machine referenced by its name and to customize the disk attachments in the same request.

The ovirt-engine-notifier tool exposes events that users can subscribe to by using SMTP or SNMP ovirt-engine-notifier providers.

The virtual machine events are:
- VM_PAUSED_EIO: The machine has been paused due to a storage I/O error.
- VM_PAUSED_ENOSPC: The machine has been paused due to lack of storage space.
- VM_PAUSED_EPERM: The machine has been paused due to a storage read/write permissions problem.
- VM_PAUSED_ERROR: The machine has been paused due to an unknown storage error.
- VM_RECOVERED_FROM_PAUSE_ERROR: The machine has recovered from being paused.

SPICE HTML5 support has been removed.

For REST API clients, the SSO authentication error now appears in the body of the response, for example, "<html><head><title>Error</title></head><body>access_denied: Cannot authenticate user 'admin@internal': The username or password is incorrect.</body></html>".

Previously, a bug allowed virtual machines with invalid snapshot IDs to be created in certain scenarios. The virtual machine functioned normally until the user tried to clone or commit the invalid snapshot or to run the virtual machine as 'stateless', resulting in failure or sometimes permanent disk deletion. In the current release, engine-setup checks the database for invalid image IDs. If found, engine-setup displays a warning describing the invalid image ID, the disk alias and virtual machine name assigned to the image, and advises the user to clone the virtual machine to fix the issue.

In the current release, ValidNumOfMonitors can be configured with engine-config.

In the current release, virtual machine initialization can be cleared with the REST API.

A link to the Manager's CA certificate has been added to the Welcome page.

Red Hat Virtualization now supports AMD EPYC processors for guest virtual machines.

You can now create an OVN network that is connected to a physical host network. This feature enables virtual machines on the external network to be on the same network as the virtual machines within the Data Center.

The Manager and the REST API support uploading an ISO image to a data storage domain and attaching it to a virtual machine as a CD-ROM device.

Previously, Red Hat Enterprise Linux kernels had kernel address space layout randomization enabled by default. This feature prevented trouble-shooting and analysis of the guest's memory dumps. In the current feature, "vmcoreinfo" is enabled for all Linux guests. It allows a compatible kernel to export the debugging information so that the memory image can be analyzed.

Previously, if a snapshot of a disk attached to a virtual machine was deleted and the user tried to attach the storage domain containing this virtual machine before the OVF_STORE had been updated with the change, the attachment operation would fail. Because the OVF indicated the presence of a disk with a snapshot, this disk was fetched as a potential disk to register, even though it was already part of a virtual machine.  In the current release, the disks are counted only once and the storage domain can be attached.

The table styling in the UI has been updated to use PatternFly to make text easier to read in selected table rows.

Previously, a template could only be imported as a clone, and a virtual machine based on such a template could only be imported with the "Collapse Snapshot" option set. In the current release, cloning and "Collapse Snapshot" are not enforced when importing a template or its virtual machine.

SUSE Linux Enterprise Server 11+ has been added to the list of guest operating systems in the Manager.

The Red Hat Virtualization Manager allows storage domain creation with a single Gluster brick.

Previously, unresponsive hosts with power management enabled had to be fenced manually. In the current release, the Manager, upon start-up, will automatically attempt to fence the hosts after a configurable period (5 minutes, by default) of inactivity has elapsed.

Previously, when a user attempted to move a disk with a snapshot that had been created before the disk was extended, the operation failed in storage domains whose data center was 4.0 or earlier. This occurred because "qemu-img convert" with compat=0.10 images interprets the space after the backing file as zeroes, sometimes causing the output disk to be larger than the logical volume created for it. In the current release, an attempt to move such a disk is blocked with an error message stating that the disk's snapshot must be deleted before moving the disk.

In the current release, MaxBlockSize is checked when extending a disk, to ensure that its size does not exceed the MaxBlockSize value.

Previously, the migration summary message showed the same value for 'total migration time' and 'actual migration time'. This value was calculated as the period of time from the start of execution of the migration command until the end of the entire migration process. In the current release, 'actual migration time' is calculated from the first migration progress event to the end of the entire migration process. If the migration command is run several times, 'actual migration time' reflects only the last run, while the 'total migration time' reflects the total time for all runs.

The descriptions of the self-hosted engine disks have been improved in the Manager to reduce the risk of their being accidentally deleted.

Previously, if a user tried to perform a live merge of a snapshot that included unattached disks, the live merge did not finish and the snapshot remained locked. In the current release, live merge is blocked if the snapshot includes unattached disks.

Previously, the VM32BitMaxMemorySizeInMB and VM64BitMaxMemorySizeInMB parameters were defined incorrectly as both global and version-specific parameters. In the current release, they are defined per version and behave correctly.

RFEs

The Manager now displays multipath device alerts, to enable users to troubleshoot storage issues caused by faulty paths.

Link Layer Discovery Protocol (LLDP) support is integrated into the Manager, so that LLDP information gathered by each host is available through the REST API and the UI. This feature improves and simplifies the detection of network configuration issues in large enterprise environments.

The default zeroing method has been changed from "dd" to "blkdiscard", which allows storage offloading, if supported by the storage array, and consumes much less network bandwidth. The zeroing method can be reverted by adding "zero_method = dd" to /etc/vdsm/vdsm.conf.

Additional mount options for the self-hosted engine storage domain, such as NFS mount, are available in the Manager.

collectd's swap plugin has been expanded to provide the following swap statistics: used, free, cached, swap_io - in and swap_io out.

The current default size of the pool of available MAC addresses, 133, is too small and has been increased to 1024.

Within the VM, direct LUNs are automatically assigned specific file systems and mount points so that they can be identified in /dev/disk/by-id by the 'lvm-pv-uuid' prefix.

vdsm

Previously, when a virtual machine with a thin provisioned disk was exported from an NFS data domain and imported into a block data domain, the disk's virtual and actual size appeared to be the same. Consequently, the disk's virtual size was used to allocate space for the disk. In the current release, qemu-img map is used to detect sparseness, allowing space allocation based on the actual size of the disk. This fix is only for NFS v4.2. Earlier NFS versions will continue to use pre-allocation, based on the virtual size of the disk.