kern.timecounter.hardware change on the fly

[Adam Vande More]

On Tue, Nov 29, 2011 at 12:34 PM, Ireneusz Pluta <ipluta at wp.pl> wrote:

> is it safe to change kern.timecounter.hardware on a busy production system
> on the fly? Or I better schedule a downtime to do that?
>
> It seems I got a bad current selection, somehow, probably as an effect of
> leaving /etc/sysctl.conf contents of early experimental setup stage of this
> machine.
>
> $ sysctl -a | grep timecounter
> kern.timecounter.tick: 1
> kern.timecounter.choice: TSC(-100) HPET(900) ACPI-fast(1000) i8254(0)
> dummy(-1000000)
> kern.timecounter.hardware: i8254
> kern.timecounter.stepwarnings: 0
> kern.timecounter.tc.i8254.**mask: 65535
> kern.timecounter.tc.i8254.**counter: 23813
> kern.timecounter.tc.i8254.**frequency: 1193182
> kern.timecounter.tc.i8254.**quality: 0
> kern.timecounter.tc.ACPI-fast.**mask: 16777215
> kern.timecounter.tc.ACPI-fast.**counter: 4619803
> kern.timecounter.tc.ACPI-fast.**frequency: 3579545
> kern.timecounter.tc.ACPI-fast.**quality: 1000
> kern.timecounter.tc.HPET.mask: 4294967295
> kern.timecounter.tc.HPET.**counter: 1703441786
> kern.timecounter.tc.HPET.**frequency: 14318180
> kern.timecounter.tc.HPET.**quality: 900
> kern.timecounter.tc.TSC.mask: 4294967295
> kern.timecounter.tc.TSC.**counter: 2720513805
> kern.timecounter.tc.TSC.**frequency: 2266656152
> kern.timecounter.tc.TSC.**quality: -100
> kern.timecounter.smp_tsc: 0
> kern.timecounter.invariant_**tsc: 1
>
> This setting seems to affect what is generally described here:
> http://archives.postgresql.**org/pgsql-admin/2011-01/**msg00059.php<http://archives.postgresql.org/pgsql-admin/2011-01/msg00059.php>and this is the primary reason I want to change.
>
> Should I switch to HPET or ACPI-fast?
>

I don't know of any reason it would be unsafe to switch.

There have been conflicting reports vs the speed of HPET and ACPI-fast
probably at least in part due to hardware differences.  You should test
each if performance is a concern for you.
-- 
Adam Vande More