Do any of you have a CHR running on vSphere that is moving more than 10 gigabit/s on a given interface?
We have one going to a vswitch is configured with more than one physical interface in an LACP configuration.
We have one VM (a CHR) on that vSwitch.
ROS reports the connection as 10 gigabit/s.
We start getting packet loss (about 70% - 80% CPU usage) at about 8.5 gigs and we peaked around 9 gigs. My first suspicion is that the bonded port isn't actually providing 20 gigs to the CHR and thus hitting an interface capacity issue. That said, I don't see anywhere I would change this, either in vSphere or in ROS.
Is that changeable somehow? If so, how?
What else might be causing me woes?
I've disabled RPS and moved to the multi-queue Ethernet queues. I got lower CPU usage, but it's not flowing more bits.
It seems like once we cross 8.5 gigabit/s through it, CPU usage increases faster than it did before.