Comment 19 for bug 1292400

This bug has been quiet for a while by the looks of it so just chipping in. I can confirm I witnessed this bug last week on an Ubuntu 14.04 x64 VM on a Windows Server 2012 R2 Hyper-V server with dynamic memory enabled on the guest. I suspect that dynamic memory (aka. memory hot-add/remove) may be related to this issue. If so, creating a VM with dynamic memory enabled and then running a workload that is memory intensive to ensure the Hyper-V server allocates more RAM to the VM may be beneficial. The tie-in with the balloon driver is interesting, and might suggest there's some level of interaction going on between these two components? Perhaps after dynamic memory adds more RAM to the system the balloon driver may trigger the issue due to no longer valid assumptions, or alternatively, the removal of memory from the system once it no longer needs it has a similar effect?

It may also be useful for others to know that we initially detected this issue due to it causing a very high load average on the affected system which can be trivially detected by monitoring systems like Nagios. Note that while the load average is high, this doesn't correspond to any given process actually consuming substantial CPU resources, so you'll want to take a look at the dmesg output which will confirm if you're experiencing the issue via the blocked systemd-udevd and kworker processes.