I have a CentOS 6.5 server running as a host for about a dozen other VM's,
and all were running just fine. I had a mix of CentOS, Ubuntu, and FreeBSD
VM's running, no problem at all.
I then updated the Ubuntu VM's to the newer 14.x release, and that
installed a 3.13 linux kernel, and after that (which I didn't notice right
at the start) the network throughput outbound was abysmal at best. Where
before I could move an ISO image in just moments, trying to send out an
image from the VM now took forever, with constant timeouts. I tried it on
multiple Ubuntu VM's and all have the exact same issue. I thought this is
strange, the older 12.x stuff ran just fine, so I thought well the
networking is in the kernel so let's back up the kernel.
So I grabbed a 3.11 kernel and loaded it, and that was a good improvement,
but still no winner, so then I backed off to a 3.8.13 kernel which seemed to
be the newest of the 3.8's, and bingo everything started working fine. I
then tried a couple Ubuntu machines that were standalone boxes (not VM's)
with the newer 3.13 it loads by default, and running independent like that
throughput is great, so it's some interaction with the qemu/kvm host.
Has anyone run into this, or have any idea, or know of any tunable changes
I can make that would make the VM play nice with the newer recommended
kernel? I was actually stunned changing kernels made the diff between
getting hundreds of megs of throughput on the host, to getting a meg or two
if lucky, with constant pauses. At the same time, my other CentOS and
FBSD VM's seem to run fine, but then again CentOS sticks with an older
kernel it seems.
---
Howard