Chapter 6 Known Issues

In addition to the known issues for Oracle Linux Virtualization Manager Release 4.3.6 listed here, you should also check the following documents:

6.1 Oracle Linux Virtualization Manager Issues

Unable to Log in to Oracle Linux Virtualization Manager Using the Host IP Address

If you access Oracle Linux Virtualization Manager using the host IP address, you see the following message and you are not able to log in:

The FQDN used to access the system is not a valid engine FQDN. 
You must access the system using the engine FQDN or one of the engine alternate FQDNs.

Solution: Only use the fully qualified domain name to access Oracle Linux Virtualization Manager. Proper hostname resolution must be in place.

Bug: 29062264

The Log Collector Tool Is Not Available on the Engine Host

The log collector tool (ovirt-log-collector) is not installed on the Manager host when you install Oracle Linux Virtualization Manager. You must install it separately.

Solution: Install the log collector tool on the Manager host:

# yum install ovirt-log-collector

Bug: 29167472

Setup Scripts for Active Domain Restrict Users to Verify for Root AD Forest and Not a Specific Subdomain

When attempting to configure authentication using the ovirt-engine-extension-aaa-ldap extension, you can add only the forest domain for authentication. When attempting to use the subdomain, you see the following warning and error messages:

[ INFO  ] Resolving Global Catalog SRV record for subdomain1.mydomain.domain.local
[WARNING] Cannot resolve Global Catalog SRV record for subdomain1.mydomain.domain.local. 
Please check you have entered correct Active Directory forest name and check that forest 
is resolvable by your system DNS servers
[ ERROR ] Failed to execute stage 'Environment customization': Active Directory forest is 
not resolvable, please make sure you've entered correct forest name. If for some reason you 
can't use forest and you need some special configuration instead, please refer to examples 
directory provided by ovirt-engine-extension-aaa-ldap package.

Solution: Issue also occurs in the upstream oVirt release. There is no workaround for this behavior.

Bug: 29182338

Active Directory Profile Missing After Restoring the Engine

When restoring from a backup to a fresh installation of the Oracle Linux Virtualization Manager, the Active Directory profile is missing.

Solution: Perform the following steps:

  1. Reinstall the ovirt-engine-extension-aaa-ldap-setup package:

    # yum install ovirt-engine-extension-aaa-ldap-setup
  2. Restart the Manager.

    # service ovirt-engine restart
  3. Add the IP address of the Active Directory to the /etc/host and /etc/resolve.conf files, respectively.

Bug: 29410228

Removed Storage Domains Are Still Shown in the Dashboard

After you remove a storage domain, it is still shown in the Dashboard.

Solution: There is no workaround for this behavior.

Bug: 29494264

Unable to Set up a Connection to an OpenLDAP Directory

When you use ovirt-engine-extension-aaa-ldap extension to set up a connection to an OpenLDAP server (option 9 - OpenLDAP Standard Schema), the setup fails because you are unable to authenticate to the directory server.

Solution: There is no workaround for this behavior. Issue also occurs in the upstream oVirt release.

Bug: 29525988

6.2 Global Configuration Issues

Virtual Machine Started on KVM Host Whose Virtual Machine Count Exceeds the HighVirtual MachineCount Property Set for an Evenly_Distributed Scheduling Policy

A virtual machine started on a KVM host whose virtual machine count exceeded the number of virtual machines set in the HighVMCount property for an Evenly_Distributed scheduling policy. Based on the scheduling policy configured in this scenario, load balancing should have been triggered and this virtual machine should have started on another KVM host in the cluster.

Solution: There is no workaround for this behavior.

Bug: 29168788

Virtual Machine Can Be Started on a KVM Host Whose CPU Utilization Exceeds the HighUtilization Property Set for an Evenly_Distributed Scheduling Policy

In a 3-host cluster where only one host is active (the other two hosts are in Maintenance mode), 5 virtual machines are created by importing OVA files. An Evenly_Distributed scheduling policy is configured with the HighUtlilization property set to 50. When the CPU utilization exceed 50% on the KVM host and a virtual machine is started, the virtual machine should fail to startup; however, the virtual machine is starting up on the KVM host in this scenario.

Solution: There is no workaround for this behavior.

Bug: 29171712

CPU Load Not Evenly Load Balanced for an Evenly_Distributed Scheduling Policy

In a 3-host cluster where an Evenly_Distributed scheduling policy is configured with the HighUtlilization property set to 50 and the CPUOverCommitDuration set to 1, CPU load did not evenly distribute across the KVM hosts in the cluster. In this scenario, virtual machines did not migrate that should have been migrated due to load balancing based on the configured scheduling policy.

Solution: There is no workaround for this behavior.

Bug: 29172270

Power_Saving Scheduling Policy Not Shutting Down Any of the KVM Hosts in a Cluster with CPU Utilization Less Than 20%

In a cluster with 3 running KVM hosts and with 4 running virtual machines, a Power_Savings scheduling policy is configured with the EnableAutomaticHostPowerManagement property set to true when the CPU and memory is found to be low on the KVM hosts. After this policy is set, the KVM hosts are not being shutdown and the virtual machines are not being migrated, even though the CPU utilization is less than 20%. Given the configured Power_Savings scheduling policy for this scenario, some of the hosts should have been shutdown.

Solution: There is no workaround for this behavior.

Bug: 29418541

Virtual Machine Not Migrating After Exceeding the MaxFreeMemoryForOverUtilized Property Value for a Power_Savings Scheduling Policy

A virtual machine is observed not migrating to another KVM host in a cluster that has enough free memory when the virtual machine exceeds the value set for the MaxFreeMemoryForOverUtilized property of a Power_Savings scheduling policy.

Solution: There is no workaround for this behavior.

Bug: 29419399

MinFreeMemoryForUnderUtilized Property Not Working for Evenly_Distributed and Power_Savings Scheduling Policies

In a cluster with 3 active hosts where there are 4 running virtual machines (3 virtual machines running on one of the hosts and 1 virtual machine running on another one of the hosts), an Evenly_Distributed policy is configured with a value set for the MinFreeMemoryForUnderUtilized property. The virtual machines in this environment then exceed the MinFreeMemoryForUnderUtilized property value set for the policy, but neither the KVM hosts are shutdown nor are the virtual machines migrated in this scenario.

The policy is then changed to a Power_Savings scheduling policy and the MinFreeMemoryForUnderUtilized property is changed to the same value as previously set for the Evenly_Distributed scheduling policy, and again it is observed that neither the KVM hosts are shutdown nor the virtual machines are migrated when this property value is exceeded.

Solution: There is no workaround for this behavior.

Bug: 29425062

MacPoolAdmin Role Is Available Only for System-Level Users

Although the MacPoolAdmin role can be assigned to users of different levels (for example, System, Data Center, Cluster, and so on), only users who are given this role at the System level are actually able to perform MacPoolAdmin tasks on the Oracle Linux Virtualization Manager, such as creating, editing, or deleting MAC address pools.

Solution: If a user requires MacPoolAdmin privileges, ensure that the user is assigned the MacPoolAdmin role at the System level on the Manager.

Bug: 29534106

6.3 Compute Issues

KVM Host Under the Control of a Engine Host Can Be Accidentally Added to Another Manager Host Without Validation of Its Current State

When an KVM host is already deployed on a Manager host, you can add this KVM host to another Manager host, causing this Manager host to take the KVM host away from the original owning Manager host. Adding a KVM host that is already under the control of a Manager host is highly not recommended, but it can be done by accident. In this event, the KVM host that resides on the original owning Manager host changes to a status of Nonreponsive and all virtual machines running on it change to a status of Unknown. When you attempt to put the nonresponsive host into Maintenance mode, the following error message is generated:

Error while executing action: Cannot switch Host to Maintenance mode.
Host still has running VMs on it and is in Non Responsive state.

Solution: For information about the workaround for this issue, refer to the related issue Removing a Stolen, Defunct, Nonresponsive, or Destroyed KVM Host .

Bug: 29127349

Duplicate KVM Host Cannot Be Moved into Maintenance Mode After Host Addition Fails

If you add a KVM host that is running virtual machines to the Manager host and you attempt to add the same KVM host using a different host name (either IP address or FQDN), the Manager host fails to add this KVM host. When you try moving the failed KVM host into Maintenance mode to remove it from the Manager, it causes the KVM host to be stuck in Preparing for Maintenance mode.

Solution: Migrate all virtual machines that are UP on the running KVM host to a different KVM host in the cluster. Doing that changes the status on the failed KVM host to Maintenance mode. You can them remove the KVM host.

Bug: 29127707

SSH Connection Hangs When Adding a New KVM Host with Firewalld Disabled to the Engine

SSH Connection hangs when adding a new KVM host with firewalld disabled to the Manager.

Solution: There is no workaround for this behavior. The SSH connection is eventually restored after some delay.

Bug: 29135914

Bridge Is Not Cleaned When Undeploying and Removing a KVM Host

After a KVM host is undeployed and removed from the Manager host, the previous bridges are not cleaned on the KVM host.

Solution: Networking resources should be manually removed or cleaned after removing the KVM host from the Manager host, and the default management network, ovirtmgmt, should not be modified.

A tool is available upstream that you can use to remove the networks that are configured by VDSM from a KVM host. For more information, refer to https://gerrit.ovirt.org/#/c/79495/.

Bug: 29167000

Creating a New Cluster in a Non-Default Data Center Without a Management Network Causes Dialog Box to Hang and Generates Uncaught Exception in UI Log

The steps to reproduce this issue are as follows:

  1. Create a new data center with the default settings.

  2. Click Configure Later when prompted on the Data Center - Guide Me dialog box.

  3. Create a new cluster by adding the cluster to the new data center, giving the cluster a new name but leaving the Management Network blank.

  4. Click OK.

    Note

    By default, the Manager adds the ovirtmgmt management network as the Management Network. This field can only be left blank if this management network has been removed and no other management networks have been created. This issue occurs only if the Management Network field is left blank.

    The dialog box then hangs, but there are no error messages that are generated in the engine.log file; however, an uncaught exception is generated in the ui.log file.

    After this issue is encountered, each time the new data center is clicked, a UI exception is generated.

    It is also observed that the new data center could be removed, even though the new cluster and a host in this data center still reside on this Manager.

Solution: If you encounter this issue, there are two possible workarounds:

  • Close the dialog box, delete the data center, and create a new data center. Creating a new data center brings back the ovirtmgmt management network.

    Or:

  • Close the dialog box, create a new network in this data center, and when creating the new cluster, select this new network as the management network.

Bug: 29385759

Incomplete Error Message Stopping a KVM Host

This issue is seen with a KVM host whose Status is Unassigned. The steps to reproduce the issue are as follows:

  1. Go to Compute and then click Hosts.

  2. On the Hosts pane, click Management and from the drop-down list select Stop.

    The following incomplete Operation Canceled error message is generated:

    Error while executing action

    Although this error message is incomplete, it is accurate; however, the message does not provide a workaround.

Solution: The Status of the KVM host must be UP before you can stop the KVM host. Try setting the KVM host to Maintenance mode and then stopping it. If this does not work, you must log in to the KVM host, resolve the issue, and bring it back up.

Bug: 29298704

Host Console Indicates KVM Host Is Registered on the Engine Host After Its Removal

The Host Console indicates that an KVM host is still registered on the Manager host after it has been removed from the Manager host. This issue is observed in the following scenario.

  1. Remove the KVM host by using the Manager as follows:

    1. On the Hosts pane, put the KVM host in Maintenance mode by clicking Management and then selecting Maintenance from the drop-down list.

    2. Click Installation and then select Reinstall from the drop-down list.

    3. Click Hosted Engine and then select Undeploy from the drop-down list.

    4. Click OK.

    5. Then set the KVM host to Maintenance mode again and click Remove to open the Remove Host(s) confirmation window and click OK.

  2. Open the Host Console.

  3. Go to Virtualization and then click Hosted Engine.

    The Host Engine Setup screen indicates that the system is already registered to the removed KVM host.

Solution: There is no workaround for this behavior.

Bug: 29444179

Removing a Stolen, Defunct, Nonresponsive, or Destroyed KVM Host

A KVM host that is under the control of a Manager host can be taken over by another Manager host. See also KVM Host Under the Control of a Engine Host Can Be Accidentally Added to Another Manager Host Without Validation of Its Current State.

For example, a KVM host that is already in use by a Manager host can be added as a New Host by a different Manager host. In this scenario, the original Manager is unable to communicate with its KVM host. Because the KVM host has a running a virtual machine, attempts to put the host into Maintenance mode are rejected with the following error message:

Host still has running VMs on it, and is Non responsive state

Solution: There are two available methods for handling this situation:

  • Method to resolve the issue of a host being taken over by another Manager host.

  • Method to remove a KVM host that has become permanently nonfunctional.

Method to Resolve the Issue of a Host Being Taking Over by Another Engine Host

When the new Manager host takes over the KVM host, the following symptoms are observed:

  • On the original owning Manager host, the KVM host shows a status of Connecting and the virtual machines show up as running.

  • On the new Manager host, copies of the virtual machines show up as external_vm-name and they are running.

Perform the following steps:

  1. Power off the KVM host.

    After powering off the KVM host, the following events occur:

    • The KVM host eventually goes into a status of NonResponsive on both Manager hosts .

    • The virtual machines go into a status of Unknown on both Manager hosts.

    • The following events are triggered for this KVM host on both Manager hosts: Handling non responsive Host host-name.

  2. Wait for these events to fail.

On both Manager hosts, perform the following steps:

  1. Click More Actions and select Confirm Host has been Rebooted from the drop-down list.

  2. Select the Confirm Operation checkbox and click OK.

    This action causes the virtual machines to be marked as Down on the Managers and the host to display a value of 0 under the Virtual Machines column. The virtual machines may now be migrated by the old manager to another host if that is permitted by the Migration Mode settings of the virtual machines.

  3. Set the KVM host to Maintenance mode.

Then, on the new Manager host, perform the following steps:

  1. Remove the virtual machine copies.

  2. Remove the KVM host.

When the KVM host comes back up, SSH to it and remove all authorized keys.

# rm /root/.ssh/authorized_keys*

Then, on the original owning Manager host, reinstall the KVM host using password authentication (for simplicity).

Note

If the installation fails on the setup due to network issues causing the KVM host to go into a NonResponsive status on the Manager host, set the KVM host back to Maintenance mode and reinstall it with an SSH key.

Method to Remove a KVM Host That Has Become Permanently Nonfunctional

Assuming the Manager host shows the KVM host status as Nonresponsive, the following symptoms are observed:

  • The virtual machines that were running on the KVM host show a status of Unknown.

  • The following events are generated on the Manager host for this KVM host:

    Handling non responsive Host host-name

    These events eventually fail.

Perform the following steps:

  1. Click More Actions and select Confirm Host has been Rebooted from the drop-down list.

  2. Select the Confirm Operation checkbox and click OK.

    This action causes the virtual machines to be shutdown and the host to display a value of 0 under the Virtual Machines column.

  3. Set the KVM host to Maintenance mode.

  4. Remove the KVM host.

For more information about these methods, refer to Doc ID 2540819.1 in the Oracle Support Knowledge Base.

Bug: 29685904

6.4 Network Issues

MAC Address Displayed on the Engine for the Non-Primary Subordinate Port of a Bond Not Synchronized With KVM Host

After a bond is created on an KVM host with 2 subordinate ports, the MAC address of the non-primary subordinate port is changed to be the same as the primary subordinate port on the KVM host, but the Manager continues to display the previous MAC address of the subordinate port.

Solution: There is no workaround for this behavior.

Bug: 29049447

Engine Does Not Update the IP Address of a Virtual Machine When Switching the Virtual Machine Network for a Running Virtual Machine

In a scenario where a virtual machine is running with one VNIC interface connected to a VM network, it is observed that the Manager allows the user to change the VNIC interface for the running virtual machine to a different VM network. This action causes the virtual machine to lose its VM network and to become inaccessible (that is, its IP address is no longer pingable). The Manager, however, still displays its old IP address and does not generate any error messages.

Solution: There is no workaround for this behavior.

Bug: 29060999

Default Route Still Set to Yes After Default Route Role Is Removed from Connected Network

Changing the default route on a network that is attached to a host and that has virtual machines connected to that network is illegal and is rejected by VDSM. The Manager host gets notified about the request failure and logs the error, but the Manager is unaware of the failure; that is, the Manage Networks pane shows that default route change was successful.

Solution: When you modify the network configuration, make sure the interface configuration accurately reflects the most recent changes.

Bug: 29133844

Exception Generated When VLAN Network Is Added with Interface Name Longer Than 15 Characters

Adding an VLAN interface whose interface exceeded 15 characters generates the following failure message:

VDSErrorException: Failed to HostSetupNetworksVDS, error = Internal JSON-RPC error: 
{'reason': '[Errno 8] Input data out of range'}, code = -32603 

The VLAN ID ranges from 0 to 4096, which is a maximum of four characters and the total interface name has a limit of 15 characters. This is a Linux kernel limitation. The maximum interface name length is defined by the kernel headers and is a global limit, affecting all applications.

Solution: Limit interface names to 15 characters (including periods and VLAN IDs).

Bug: 29221397

Adding KVM Host Connected to VLAN Network Fails When VLAN Network Is Not Configured on Management Network

Adding a KVM host that is connected to a VLAN network fails when the ovirtmgmt management network is not configured on the VLAN network.

Solution: All servers that are part of a single cluster must be on the same VLAN management network.

Perform the following steps:

  1. Go to Network and then click Networks.

  2. Select the ovirtmgmt network under the Name column.

  3. Click Edit.

  4. Select the Enable VLAN tagging checkbox and enter the VLAN tag ID in the text entry field.

    Make sure that the VLAN tag ID is the same as the VLAN ID set on the host interface to be connected to the ovirtmgmt management network.

  5. Go to Compute and then click Hosts.

  6. Click New.

  7. On the Hosts pane, click on the KVM host.

  8. Click on the Network Interfaces tab.

  9. Click Setup Host Networks and add the ovirtmgmt management network to the appropriate interface.

Bug: 29245869

Network Synchronization Fails but Is Reported as Succeeding

After creating a VLAN network with a MTU value and adding the network to a host interface, the network is not synchronized because the host retains the default MTU value of 1500. When clicking Sync All Networks, the Manager reports the following message, indicating that the network synchronization is completed:

Finished Synchronizing networks on host

The network, however, is still not synchronized.

Solution: Try clicking Sync All Networks again.

Bug: 29311422

KVM Host Becomes Non-Operational When a Bondport Connected to Virtual Machine Network Changed from an Unsupported Mode to a Supported Mode

When a bondport is created with a mode that is not supported for a VM network, the network can be connected to a non-VM network successfully; however, if the non-VM network is changed to VM network, the network becomes out of sync. Furthermore, if the user follows the instructions provided in the error message to fix the network issue, the KVM host becomes non-operational. This issue is caused by the network change (non-VM to VM network) combined with the bond mode change.

Note

No issues occur when performing these actions in reverse; that is, changing a VM network to a non-VM network and then changing the bond mode accordingly.

The steps to reproduce this issue are as follows:

  1. Create a non-VM network and connect the network to a bondport mode in 0, 5, or 6.

  2. Change the network mode to a VM Network by editing the network in the Manager and selecting the VM Network checkbox.

    This action causes the network to become out of sync.

  3. Click Sync All Networks.

    The following error is generated: Network nameis attached to bond number. VM networks cannot be attached to bonds in mode 0, 5 or 6.

  4. Change the bond mode to a mode available for a VM network (1, 2, or 4) and click OK.

    After the network setup window closes, the Sync All Network button appears again, indicating that the network is out of sync, and the host becomes non-operational.

Solution: You can click Sync All Networks again to bring the KVM host up. Alternatively, to bring the KVM host up, you can put this KVM host in Maintenance mode and activate it after the network is back in sync.

Bug: 29312752

Engine Does Not Check Connectivity When Removing an Interface

When you connect a host interface to a network, the Manager validates their connectivity before establishing the connection. However, when removing an interface, the Manager does not check their connectivity. This behavior is observed in a network configuration where the management network is already connected to a bondport with 3 or more subordinate ports and only 1 of the ports is in UP state and is able to communicate with the management network.

In this situation, when you attempt to remove the UP port (primary port) in the Setup Host Networks dialog box, the Manager generates the following Operation Canceled error message:

Error while executing action HostSetupNetworks: Could not connect to peer host

This error message, which is generated during the removal of the subordinate port, is referring to the new connection that cannot be established, even though the removal of the primary port has already occurred.

Solution: There is no workaround for this behavior. In this situation, you need to make sure the remaining communication between the bondport and the management network are not interrupted after the subordinate port is removed, otherwise the host loses connection with the Manager and becomes non-responsive.

Bug: 29338703

Network Scale Limitation When Adding Networks to a KVM Host

In an Oracle Linux environment you can add approximately 389 networks to a KVM host, after which network communication timeout errors start appearing in the Manager log file and no further networks can be added to the host.

Solution: There is no workaround for this behavior.

Bug: 29383782

Network Label Failure Generates Illegal Network Parameters Error When Performing Other Network Operations on the KVM Host

In a scenario where two VLANs are erroneously created with duplicate VLAN numbers, a warning in the form of an orange exclamation mark appears as expected in the Manager next to the duplicate VLAN on the Setup Host dialog box with the following error message:

Cannot have networks with duplicate vlan id on same interface.

After this network configuration error, attempting to perform other network operations on the KVM host generates the following Operation Canceled error message:

Error while executing action HostSetupNetworks: Illegal Network parameters 

Solution: You must resolve the network label failure before other network operations can be performed.

Bug: 29424399

Cannot Modify a VNIC in a Running Virtual Machine if an MTU Value is Set

If a virtual machine is running, editing a VNIC that has an MTU value set fails with the following message:

 Error while executing action Edit VM Interface properties: General Exception 

The following message is displayed in the /var/log/vdsm/vdsm.log file:

 libvirtError: Operation not supported: cannot modify MTU 

Solution: Shut down the virtual machine before editing the VNIC. A known issue in the upstream oVirt release.

Bug: 29456945

VM Network Bridge Name Does Not Match VM Network Name

When you create a new VM network, VDSM creates a bridge named after the VM network and its associated configuration file (/etc/sysconfig/network-scripts/ifcfg-vm_network) on the Oracle Linux KVM hosts.

If the VM network name is longer than 15 characters, or it contains special characters or a space character, VDSM generates a name for the bridge and the configuration file in the format "onXXXXXXXXXXXXX" where XXXXXXXXXXXXX is the first 13 hexadecimal characters extracted from the UUID of the network.

This is a Linux kernel limitation, bridge names must be 15 characters or less, and must not contain any special characters.

Solution: Limit VM network names to 15 characters and avoid any special characters.

Bug: 29409851

6.5 Storage Issues

Using the Change CD Option When Changing to an ISO Attached to Block-Based Storage Domain Fails

If you try to use the Change CD option for a CD/DVD-ROM attached to a virtual machine coming from a block-based storage domain like iSCSI or fiber channel, you get the following error:

Error while executing action Change CD: Drive Image file could not be found.

Solution: This is an upstream bug, however, you do have two workaround options:

  • Change an ISO with a virtual machine booted using either a local storage domain or an NFS storage domain.

  • Stop the virtual machine before changing the ISO or mount the ISO on the virtual machine.

Bug: 31598688

Storage Must be on the Same Subnet as the KVM Hosts

In order to avoid issues with routing, a storage must be located on the same subnet as the Oracle Linux KVM hosts that will use the storage.

Bug: 29220930

Decreasing the Storage Size Used in Storage Domains Is Not Recommended

Although the size of NFS Ext 3 and 4 filesystems can be decreased and Block Storage (iSCSI and Fibre Channel) provides the capability to decrease storage size by resizing the underlying LUN, decreasing the storage size that is used for a storage domain is potentially a dangerous operation and is highly not recommended.

Solution: Do not decrease the storage size used in storage domains in your virtualization environment.

Bug: 29285337

Manager Not Updating the Size of an iSCSI-Based Virtual Disk After ZFS Storage Resized

The Oracle Linux Virtualization Manager is not updating the size of an iSCSI-based virtual disk after the ZFS storage is resized. This issue is seen after creating an LUN, attaching the virtual disk to a running virtual machine, resizing the LUN, and restarting VDSM. After the refreshing the Disks pane, the iSCSI-based virtual disk size does not update accordingly. However, when checking the iSCSI disk on the SPM host, the size of the disk is correctly updated.

Solution: There is no workaround for this behavior.

Bug: 29370809

Virtual Machine Templates Can Be Imported with MAC Addresses Not in Range for the MAC Address Pool

virtual machine templates can be imported with MAC addresses that are out of range for the MAC address pool.

Solution: If there is a MAC address range restriction for your virtualization environment, the workaround is to manually change the MAC address to an address within the range of the MAC address pool before starting the virtual machine.

Bug: 29449334

Oracle Linux Virtualization Manager Does Not Prevent Adding a Direct LUN Disk Being Used by an Active Storage Domain

The Manager prevents creating a storage data domain on LUNs that are being used by direct LUN disks. The Manager, however, does not prevent adding a direct LUN disk that is being used by an active storage domain.

Solution: There is no workaround for this behavior. If you attempt to add a direct LUN disk that is being used by an active storage domain, the Manager prompts you to approve the operation with the following warning message: This message might be unrecoverable and destructive!

Bug: 29465439

Microsoft Windows 8 (and Later) Virtual Machines Disabling Third-Party Mini-Port Drivers

Microsoft Windows 8 (and later) virtual machines are disabling third-party mini-port drivers.

Solution:To bring back the third-party mini-port drivers, perform the following steps:

  1. Start a virtual machine using the IDE interface.

  2. Open the console for the virtual machine.

  3. Open the Command Prompt in the Microsoft Windows virtual machine and set the machine to boot in Safe Mode.

    > bcdedit /set {current} safeboot minimal

  4. Shutdown the virtual machine.

  5. Change the storage type.

    The virtual machine now starts normally.

  6. Reset the bcdedit setting to boot the virtual machine in Normal Mode.

    > bcdedit /deletvalue {current} safeboot

  7. Reboot the virtual machine.

Bug: 29472477

Manager Does Not Prevent Attaching VirtIO-SCSI Disks to OL6U10_X86 Virtual Machines

Manager does not prevent attaching VirtIO-SCSI disks to OL6U10_X86 virtual machines. The VirtIO-SCSI option in the Manager should be disabled for OL6U10_X86 virtual machines in the Virtual Machine: Resource Allocation Settings.

Solution: There is no workaround for this behavior. Do not attach VirtIO-SCSI disks to OL6U10_X86 virtual machines.

Bug: 29499061

6.6 Virtual Machine Issues

Virtual Machine Does Not Boot with Allow Privileged SCSI I/O Option Enabled

If you have configured a virtual machine on Oracle Linux 7.7 with a physical LUN and the Allow Privileged SCSI I/O option enabled, the virtual machine does not boot and reports an error:

libvirtError: Requested operation is not valid: unpriv_sgio is not supported
by this kernel

Solution: Do not use the Allow Privileged SCSI I/O option if your virtual machine has access to a physical LUN.

Bug: 31021044

Virtual Machines Not Migrated As Expected Through Virtual Machine Affinity and Host Affinity Rules Set to Active and Enforcing

In a 2-host cluster where 4 virtual machines are running on one of the hosts, a Vm Affinity Rule and a Host Affinity Rule are set on the Manager to Active and Enforcing. These rules, however, are not applied correctly because the virtual machines defined in the affinity group should have migrated to the other host; instead, the other KVM host is filtered out in this scenario.

Solution: Enabling the VM Affinity Rule or the Host Affinity Rule Enforcing option imposes strict limits on whether virtual machines can start or be migrated. With VM Affinity Rule Enforcing enabled (shown as Hard in the list of Affinity Groups), the system does not migrate a virtual machine to a host different from where the other virtual machines in its affinity group are running, even if attempting to satisfy a Host Affinity Rule that only permits the virtual machines to run on a host other than where they are presently running.

The workaround for both VM and Host affinity rules is to not enable enforcing unless all scenarios including migration scenarios are understood. The rules are not disabled if you do not select the Enforcing option; it changes the enforcement from Hard to Soft, permitting the system to temporarily break the rule as adjustments in virtual machine placement are made.

Bug: 29190112

Cannot Start or Edit Virtual Machine Imported from OVA with More Than Three Disks

When a Microsoft Windows virtual machine is imported from an OVA with more than 3 disks, the virtual machine cannot be started or edited.

Solution: If you want to import a Microsoft Windows virtual machine from an OVA with more than 3 disks, the workaround is to set the disks to inactive after the import and before you reboot. Then, change the disk connection protocol from IDE to virtio-scsi.

Bug: 30776581

Changing Watchdog Actions on a Running Virtual Machine Requires a Virtual Machine Restart for New Attribute to Take Effect

After enabling a watchdog device and setting the watchdog action attribute to reset, the virtual machine is restarted. Changing the watchdog action to a different attribute, such as poweroff, causes the Pending Virtual Machine Changes window to open with the following message:

Changes that require Virtual Machine restart: watchdog. The VM is then restarted. 

The issue is that even though watchdog action is changed to poweroff, the attribute value did not take effect immediately. The virtual machine must first be restarted before the new attribute takes effect.

Solution: There is no workaround for this behavior. If a watchdog action is changed on a running virtual machine, you must restart the virtual machine for the attribute changes to take effect.

Bug: 29213956

Guest Agent Hooks Do Not Work If a Migration Policy Is Set to Legacy

The guest agent hooks do not work for migration policies that are set to Legacy.

Solution:This behavior is expected, as the guest hooks are enabled depending on the migration policy that is configured, and the guest agent hook mechanism is disabled for migration policies set to Legacy. To use guest agent hooks, you must use one of the other available migration policy types.

Bug: 29261746

Importing Several OVA Files Resulted in Only One Virtual Machine Being Imported

When importing virtual machines, multiple OVA files can be selected, but only one virtual machine is actually imported. If there is a limitation, the Manager should block attempts to import multiple OVA files.

Solution: There is no workaround for this behavior.

Bug: 29373795

High-Availability Virtual Machines Fail to Restart After Restarting KVM Host

High-availability virtual machines fail to restart after the KVM host is restarted on the Manager.

Solution: There is no workaround for this behavior. The KVM host eventually starts up several minutes later.

Bug: 29393228

Migration Mode Greyed Out When Setting Pinning-to-Host Setting for a Virtual Machine

This issue is observed in a 2-host cluster. The steps to reproduce the issue are as follows:

  1. Edit a virtual machine by selecting the Highly Available checkbox on the High Availability tab, and click OK.

  2. Edit the virtual machine and update the pinning-to-host settings by selecting Specific Hosts and adding the two hosts from the drop-down list on the Host tab, and click OK.

    After doing completing these steps, the Migration Mode drop-down list is greyed out. This option should be available in this scenario.

Solution: Try changing the pinning-to-host option to Any Host in Cluster and then change the setting back to Specific Hosts again.

Bug: 29374044

Cannot Assign Virtual Machine from a Virtual Machine Pool: Manager Claims That the Virtual Machine Is Attached When It Is Not

After creating a virtual machine pool with 10 virtual machines, setting the Maximum number of VMs per user field to 10, and granting permissions to a user, the user correctly attaches some virtual machines. Later the user tries to attach more virtual machines, but the Manager generates the following error, indicating that virtual machines are already attached when they are not:

PoolName:
Cannot attach VM to VM-Pool. VM-Pool is already attached to a User. 
Cannot allocate and run VM from VM-Pool. There are no available VMs in the VM-Pool.

Solution: A known issue in the upstream oVirt release. There is no workaround for this behavior.

Bug: 29379502

Manager Does Not Generate a Warning Message to Prevent OVA Export from a Virtual Machine Whose Guest OS Is Installed on a Direct LUN Disk

The Manager does not generate an warning message that prevents OVA export from a virtual machine whose guest OS is installed on a direct LUN disk.

Solution: There is no workaround for this behavior.

Bug: 29432214

Manager Allows OVA Export from a Virtual Machine Whose Guest OS Is Installed on a Direct LUN Disk

The Manager allows OVA export from a virtual machine whose guest OS is installed on a direct LUN disk. The Manager should prevent attempts to export an OVA from a virtual machine whose guest OS is installed on a direct LUN disk image.

Solution: There is no workaround for this behavior.

Bug: 29432323

Manager Does Not Prevent Virtual Machine Shutdown When Creating a Snapshot

The Manager does not prevent a virtual machine from being shutdown when creating a snapshot, thereby causing the snapshot to hang. In this situation, the virtual machine cannot be started or removed and the creation of the snapshot cannon be stopped.

When inspecting the Manager log, the follow message is seen repeatedly:

[org.ovirt.engine.core.bll.SerialChildCommandsExecutionCallback]
(EE-ManagedThreadFactory-engineScheduled-Thread-34)
[7363ae20-1f20-451c-8318-215f122fcc15] Command 'CreateSnapshotForVm' (id:
'b50fee33-204b-4fc0-a8e6-d40c7f58d485') waiting on child command id:
'3c843771-8b8f-4c07-9230-822bd2892481' type:'AddDisk' to complete
2019-03-08 03:55:22,992Z INFO  

Solution: There is no workaround for this behavior.

Bug: 29457750

Engine Exception When Unplugging a NIC Connected to a Virtual Machine Network on a Running Virtual Machine

Unplugging a NIC that is connected to a VM network on a running virtual machine may fail on the Manager with the following Operation Canceled message:

Error while executing action Edit VM Interface properties: Failed to deactivate 
VM Network Interface.

When inspecting the Manager log, you may see the following EngineException:

EngineException: org.ovirt.engine.core.vdsbroker.vdsbroker.VDSErrorException:
VDSGenericException: VDSErrorException: Failed to HotUnplugNicVDS, error =
Timeout detaching <Interface name=vnet3, type=bridge, mac=00:21:f6:00:04:6b
at 0x7f6d30377868>, code = 50 (Failed with error DEACTIVATE_NIC_FAILED and
code 50)

Solution: Try unplugging the NIC again.

Bug: 29460927

Highly Available Virtual Machine Not Started on Other Host When Source Host Dies Unexpectedly

Highly available virtual machines are not started on the other host when the source host is powered off or dies unexpectedly.

Solution: If a host that has running highly available virtual machines is powered off or dies unexpectedly and the virtual machines do not have additional split-brain prevention mechanisms in place (such as Sanlock, VM leases, or VM resume), you may be required to use the Confirm 'Host has been Rebooted' option from the More Actions drop-down list on the Hosts pane before the Manager allows the highly available virtual machines on the host to be restarted on the other host.

Bug: 29491043

VNC Console Session for a Running Virtual Machine Closes During Virtual Machine Migration

If you have an VNC console open for an active virtual machine during virtual machine migration, the session closes. This behavior occurs when using either the native client or noVNC.

Solution: There is no workaround for this behavior.

Bug: 29491251

Migrate VM Dialog Box Cannot Be Closed While a Virtual Machine Is Being Started

The Migrate VM dialog box cannot be closed while a virtual machine is being started in the Manager.

Solution: Refresh the browser page.

Bug: 29498386

Unable to Perform CPU Hot Unplug with Default Machine Type (pc-i440fx-2.6) for Oracle Linux Virtual Machine

The CPU hot unplug feature on the Manager is not working with the default emulated machine type for Oracle Linux virtual machines (pc-i440fx-2.6).

Solution: To enable CPU hot unplug feature for a particular Oracle Linux virtual machine, you must perform the following steps:

  1. Shutdown the virtual machine.

  2. Change the Custom Emulated Machine to pc-i440fx-2.7 by performing the following steps in the Administration Portal:

    1. Go to Compute and click Virtual Machines.

    2. Select the Oracle Linux virtual machine and click Edit Virtual Machine.

    3. Click the System tab on the sidebar.

    4. Click Advanced Parameters to expand the menu.

    5. On the Custom Emulated Machine drop-down menu, select pc-i440fx-2.7.

    6. Click OK to save your changes.

  3. Restart the virtual machine.

Bug: 29517731

VM Portal Displays Error for User With Super User Permission When VM in Pool is Running

If a user with SuperUser permission logs into the VM Portal and a VM in a Pool is in a running state, the VM Portal displays an error message "Sorry, VM Portal is currently having some issues."

Solution: For users who require access to the VM Portal, create a new user who does not have SuperUser permission and log into the VM Portal as that new user.

Bug: 30770232

Cannot Import OVA Files That Contain More Than 1 Virtual Machine

If an Open Virtualization Appliance file contains more than one virtual machine, then the virtual machines in the file cannot be imported. The import begins and appears to import the first virtual machine, but then the import fails and no virtual machine are installed successfully.

Solution: Restrict Open Virtualization Appliance files to one virtual machine.

Bug: 30771759

Graphics Protocol Set to None When Using virt-v2v Upload

When using virt-v2v -o ovirt-upload, the graphics protocol is set to None and the Console button is grayed out.

Solution: The workaround is to click on the Edit button and then close the pop-up by clicking on OK, which sets the value to VNC.

Bug: 30807213