第 34 章 Virtualization

When installing Red Hat Enterprise Linux 7 from sources other than the network, the network devices are not specified by default in the interface configuration files. As a consequence, creating a bridge by using the iface-bridge command in the virsh utility fails with an error message. To work around the problem, add the DEVICE= lines in the /etc/sysconfig/network-scripts/ifcfg-* files.

Nesting more than 7 PCI bridges is known to cause segmentation fault errors. It is not recommended to create more than 7 nested PCI bridges.

The kernel sym53c8xx module is not supported in Red Hat Enterprise Linux 7. Therefore, it is not possible to use an emulated Small Computer System Interface (SCSI) disk when Red Hat Enterprise Linux is running as a guest on top of the Xen hypervisor or Amazon Web Services (AWS) Elastic Compute Cloud (EC2). Red Hat recommends to use paravirtualized devices instead.

When the xen_emulated_unplug=never or xen_emulated_unplug=unnecessary options are passed to the guest kernel command line, an attempt to hot plug a new device to the Xen guest does not work. Running the xl command in the host succeeds but no devices appear in the guest. To work around this issue, remove the aforementioned options from the guest kernel command line and use paravirtualized drivers to allow hot plugging. Note that xen_emulated_unplug=never and xen_emulated_unplug=unnecessary are supposed to be used for debugging purposes only.

After multiple hot plugs and hot unplugs of a SCSI disk in the Hyper-V environment, the disk in some cases logs an error, becomes unusable for several minutes, and displays incorrect information when explored with the partprobe command.

A prior Intel microcode update removed the Hardware Lock Elision (HLE) and Restricted Transactional Memory (RTM) features from 4th Generation Intel Core Processors, Intel Xeon v3 Processors, and some 5th Generation Intel Core Processors. However, after performing a live migration of a KVM guest from a host containing a CPU without the microcode update to a host containing a CPU with the update, the guest may attempt to continue using HLE and RTM. This can lead to applications on the guest terminating unexpectedly with an Illegal Instruction error. To work around this problem, shut down the guest and perform a non-live migration if moving from a CPU with HLE and RTM to a CPU without the features. This ensures that HLE and RTM are unavailable on the guest after the migration, and thus prevents the described crashes.

Due to an unintended incompatibility between QEMU and the pSeries platform, the systemd-detect-virt and virt-what commands cannot properly detect PowerKVM virtualization on IBM Power Systems. There is currently no known workaround.

When the kernel shared memory (KSM) feature is enabled with the merge_across_nodes=1 parameter, KSM ignores memory policies set by the mbind() function, and may merge pages from some memory areas to Non-Uniform Memory Access (NUMA) nodes that do not match the policies. To work around this issue, disable KSM or set the merge_across_nodes parameter to 0 if using NUMA memory binding with QEMU, as this leads to NUMA memory policies configured for the KVM VM working as expected.