[mythtv-users] MySQL on SSD 99% Utilization
Mike McMullin
mwmcmlln at mnsi.net
Fri Feb 4 05:27:13 UTC 2011
On Thu, 2011-02-03 at 15:15 -0500, Brian Long wrote:
> On Thu, Feb 3, 2011 at 2:44 PM, Eric Sharkey <eric at lisaneric.org>
> wrote:
> On Thu, Feb 3, 2011 at 11:40 AM, Brian Long
> <briandlong at gmail.com> wrote:
> > Is it generally a
> > bad practice to put MySQL on a SSD? I figured it would
> provide a nice fast
> > DB and keep my DB separate from my recording disks, but I
> guess I was wrong.
>
>
> I have my backend's MySQL db on an SSD, with the recordings on
> a
> magnetic disk, and I don't have anything like the problems
> you're
> seeing. Moving the db to the SSD seems to have been a big
> boost to
> responsiveness when it comes to searching/scheduling new
> recordings or
> bringing up the guide data.
>
> One thing I have noticed with iostat is that it tends to
> report
> elevated numbers on its first invocation, which is why I
> always give
> it a repeat interval (usually 2 seconds) and watch until it
> appears to
> be at a steady state.
>
> I have comm flagging going on right now and here's what I see
> (/dev/sda is my SSD):
> <snip>
> So I think something is seriously misconfigured on your
> system.
>
> Yes, I didn't dig deep enough and it's not MySQL slamming my SSD; it's
> the mythbackend logging to /var/log which is also on my SSD. I tried
> tuning MySQL and moving /var/lib/mysql to spinning disk and noticed
> the SSD was still 99% utilized. Then I noticed my mythcommflag job
> was endlessly logging the following types of errors:
> 2011-02-03 15:10:43.547 AFD Error: Unknown decoding error
> 2011-02-03 15:10:43.564 [h264 @ 0x37412f7ec0]decode_slice_header error
> 2011-02-03 15:10:43.583 [h264 @ 0x37412f7ec0]no frame!
> 2011-02-03 15:10:43.583 [h264 @ 0x37412f7ec0]get_buffer() failed
> (stride changed)
> 2011-02-03 15:10:43.604 [h264 @ 0x37412f7ec0]decode_slice_header error
> 2011-02-03 15:10:43.653 [h264 @ 0x37412f7ec0]no frame!
>
>
> Mythbackend seems to be DOS'ing my SSD! :-) I guess I need to dig
> into removing that job from the queue since mythcommflag doesn't like
> the recording.
What about linking the log directory to one on the spinning disk?
More information about the mythtv-users
mailing list