On Mon, Jul 3, 2017 at 12:27 AM, Darrell Budic
> It seems vdsmd under 4.1.x (or something under it’s control) changes the
> disk schedulers when it starts or a host node is activated, and I’d like to
> avoid this. Is it preventable? Or configurable anywhere? This was probably
> happening under earlier version, but I just noticed it while upgrading some
> converged boxes today.
> It likes to set deadline, which I understand is the RHEL default for
> centos 7 on non SATA disks. But I’d rather have NOOP on my SSDs because
> SSDs, and NOOP on my SATA spinning platters because ZFS does it’s own
> scheduling, and running anything other than NOOP can cause increased CPU
> utilization for no gain. It’s also fighting ZFS, which tires to set NOOP on
> whole disks it controls, and my kernel command line setting.
We've stopped doing it in 4.1.1 (
> Users mailing list
Users mailing list