VMs freezing when host is running 4.14

From: Marc Haber
Date: Tue Nov 21 2017 - 11:18:27 EST


Hi,

I am running Debian stable with home-built kernels on a number of KVM
hosts and a bigger number of KVM VMs. With 4.14, I have an interesting
phenomenon on _one_ of my hosts, while all other hosts run fine. All
systems are reasonably similar to each other.

On the affected host, VMs freeze at a rate about two or three per day.
They just stop dead in their tracks, console and serial console become
unresponsive, ping stops, they don't react to virsh shutdown, only to
virsh destroy. They do, however, take a noticeable part of CPU resources
when they're in this state, up to a full CPU core. What's left in
syslog of a VM is unsuspicious, the host logs don't have anything
uncommon. When I start a VM that allocates a lot of memory (like the 8
GB Windows VM that I use for bookkeeping and taxes), it happens that two
to five of the Linux VMs freeze in the same second.

The affected host is a Thinkpad T520 with 16 Gig of RAM and an Intel(R)
Core(TM) i5-2520M that is neither under noticeable load nor under memory
pressure (it can afford 8 Gig of disk cache).

When I boot the host back to 4.13.11, things are just fine and the
machine is chugging away painlessly for days. When I boot 4.14, the VM
freeze phenomenon usually appears in the first 24 hours.

The other VM hosts (ranging from a small PC Engines APU to a bigger 48
Gig Server in Housing) run just fine with 4.14 as well.

Any ideas?

Greetings
Marc

--
-----------------------------------------------------------------------------
Marc Haber | "I don't trust Computers. They | Mailadresse im Header
Leimen, Germany | lose things." Winona Ryder | Fon: *49 6224 1600402
Nordisch by Nature | How to make an American Quilt | Fax: *49 6224 1600421