solo6x10: all interrupts for all cards handled by CPU0, no balancing - why?

solo6x10: all interrupts for all cards handled by CPU0, no balancing - why?

[Andrey Utkin]

Hi, having another "card freeze" issue with linux-next (tag
next-20150128) on a server running 3 solo6110 cards. The freeze
happens after 3 days or so. Much better than 30 minutes, which was the
case before the recent enhancement by Krzysztof Halasa.
This is Ubuntu Trusty. There's /usr/sbin/irqbalance process running.
See interrupt stats here:
https://gist.github.com/krieger-od/f8d99080d6fc30dad3d2
I wonder why all solo6x10 interrupts happen on CPU0, while there are 3
more cores.

However, I have got an idea reading this:
Irqbalance is a daemon to help balance the cpu load generated by interrupts
across all of a systems cpus.  Irqbalance identifies the highest volume
interrupt sources, and isolates them to a single unique cpu, so that load is
spread as much as possible over an entire processor set, while minimizing cache
hit rates for irq handlers.

Disabled irqbalance launch on boot. Rebooted, and the host got down up to now :(
Any comments, except laugh? :)

-- 
Andrey Utkin

Re: solo6x10: all interrupts for all cards handled by CPU0, no balancing - why?

[Andrey Utkin]

The host was rebooted and got back online.
Without irqbalance daemon, all solo6x10 interrupts are still on CPU0.
See https://gist.github.com/krieger-od/d1686243c67fbe3e14a5
Any ideas are strongly appreciated.


-- 
Andrey Utkin

Re: solo6x10: all interrupts for all cards handled by CPU0, no balancing - why?

[Ismael Luceno]

On Fri, 30 Jan 2015 00:29:13 +0200
Andrey Utkin <andrey.krieger.utkin@gmail.com> wrote:
> The host was rebooted and got back online.
> Without irqbalance daemon, all solo6x10 interrupts are still on CPU0.
> See https://gist.github.com/krieger-od/d1686243c67fbe3e14a5
> Any ideas are strongly appreciated.
> 

My understanding was that balancing happened by package, not by core,
but my knowledge might be outdated...