IO stops on a Red Hat Enterprise Linux guest on an Oracle VM hypervisor
Issue
-
One or more xv* devices appear as having a non zero avgqu-sz and 100% utilisation in the output of iostat, without any other indication of IO being pushed through them (i.e. there are no read or write requests, there is no await etc).
-
Any command that attempts to push IO to these devices hangs. Applications that use these devices also hang.
Environment
- Red Hat Enterprise Linux (The issue has so far been observed on Red Hat Enterprise Linux 7, but it is expected to appear on all versions)
- Oracle VM hypervisor (Based on the nature of the issue, it is expected to affect any xen based hypervisor)
- The issue is more likely to appear as the number of devices attached to guests on the same hypervisor increases. VMs with multiple disks are more likely to be affected.
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.
