temporary unresponsiveness or 'soft lockup' of KVM host and guest while a PCI device is being hot-plugged to the virtual machine
Issue
A KVM host and guest can appear to be unresponsive temporarily while a virtual PCI device such as virtio disk
, virtio net
, e1000
or rtl8139
is being hot-plugged to the virtual machine. If the duration of unresponsiveness exceeds the threshold of the 'soft lockup' watchdog, the host and guest kernel can log BUG: soft lockup ...
events similar to the examples given under Diagostic Steps. If the KVM host is a node in a High Availability Cluster, it could even be evicted/fenced from the cluster.
This issue has originally been observed while a virtio disk
was being hot-plugged to a large-sized virtual machine (16 virtual CPUs, 64 GB memory) that was running a cpu- and memory-intensive application. In that case the KVM host operating system was Red Hat Enterprise Linux 6.3 on a large-sized physical server (160 CPUs, ~1 TB memory), and the guest operating system was Red Hat Enterprise Linux 5.5.
Environment
- Red Hat Enterprise Linux 6 based KVM host/hypervisor
- Red Hat Enterprise Linux 5 or 6 guest
- large sized virtual machine on large sized physical server
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.