[mythtv-users] High IO activity during recording

[Karl Newman]

On Sun, Nov 30, 2014 at 8:51 AM, Thomas Börkel <thomas at boerkel.de> wrote:

> Hi,
>
> one of my systems does most of the time very heavy IO activity during
> recording, meaning that the disk is making noises and the LED is
> flickering like hell. Every now and then, activity slows down and then
> raises again shortly after.
>
> My other system can record 4 shows concurrently with very little
> visible/hearable disk activity.
>
> I am at the moment testing recording the same 2 HD shows on both systems
> (no commflagging) and the difference in activity is pretty big.
>
> But the thing is, the 2 systems are pretty similar.
>
> System 1 (low activity):
> - AsRock mainboard, Core i5 Ivy Bridge
> - SSD for system (deadline), 3 TB WD RED for recordings (cfq)
> - kernel 3.14.16
>
> System 2 (high activity):
> - AsRock mainboard, Core i5 Sandy Bridge
> - SSD for system (deadline), 4 TB WD RED for recordings (cfq)
> - kernel 3.16.5
>
> Both recording partitions are XFS with noatime,nodiratime. Both disks
> are using udma6.
>
> I see only mythbackend threads at the top when using iotop (2 MB/s total
> disk write).
>
> I also tried IO scheduler noop and deadline with no visible difference.
>
> Is there anything in MythTV or in the kernel config, I can possibly
> configure, like IO buffer size?
>
> Thanks!
>
> Regards,
> Thomas
>

Is NCQ enabled? On my motherboard the controller SATA mode is not AHCI by
default so I had to enable it in the BIOS. This post is a good reference
for checking it:
http://forums.fedoraforum.org/showpost.php?s=e03f78859973c9cd7094b67060295547&p=957135&postcount=2
Essentially, you can check it by cat /sys/block/sdX/device/queue_depth
(where X is your SATA device--a, b, c, etc.) If the depth is greater than
1, then NCQ is enabled.

Karl
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mythtv.org/pipermail/mythtv-users/attachments/20141130/bf9ccbdd/attachment.html>