So some questions:

What hardware is this? Are they 16 “real” cores or hyper threaded cores?
Do you have a test case setup so you can readily measure the impact of change?
Many tunings that involve numa will only show substantial results ion specific app
What does cat /proc/cpuinfo | tail -28 return?
When you say maxed out do you literally mean that cores 6,7 show 100% CPU utilization

> On Jan 12, 2018, at 9:21 AM, Raffael Vogler <raffael.vogler@yieldlove.com> wrote:
>
> I have 16 cores (0 to 15) and the most active processes are:
>
> - nginx
> - php-fpm
> - eth0-TxRx-0 (paired Tx and Rx queues #1)
> - eth0-TxRx-1 (paired Tx and Rx queues #2)
>
> eth0-TxRx-0 is already dedicated to core #5 and eth0-TxRx-1 to core #6.
>
> But on those two cores are also nginx and php-fpm processes being executed.
>
> Now I would like o restrict nginx and php-fpm to cores 0-4 and 7-15.
>
> I would like to confirm that this can be correctly achieved with:
>
> taskset -c 0-4,7-15 nginx
> taskset -c 0-4,7-15 php-fpm
>
> Is that a safe and sound approach?
> _______________________________________________
> nginx mailing list
> nginx@nginx.org
> http://mailman.nginx.org/mailman/listinfo/nginx

_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx