[mythtv-users] IOBOUND error over cifs (to nslu2)

[Brandon Stoll]

On 2/28/06, Steve Malenfant <smalenfant at gmail.com> wrote:
> I have a frontend and a backend.  My backend has 2 HD tuner that
> record on the local disk, that works just fine.
>
> My frontend has been transformed into slave backend, it has 1 HD tuner
> in it.  I want to use it to record an extra program.  The problem is
> that I get IOBOUND error to the backend and that my frontend has only
> a small 10G partition for linux (bad choice, I know).
>
> So I'm using my nslu2 mounted with cifs : //nslu2//public on
> /mnt/nslu2 type cifs (rw,mand)
>
> While I'm recording an HD program, top shows this :
>
> Cpu(s):  0.3% us,  3.0% sy,  0.0% ni,  0.0% id, 94.7% wa,  2.0% hi,  0.0% si
>
> I've got a few errors in my log that looks like this :
> 2006-02-28 18:54:33.218 TFW, Error: Write() -- IOBOUND begin cnt(9400) free(555)
> 2006-02-28 18:54:33.221 TFW, Error: Write() -- IOBOUND end
> 2006-02-28 18:54:33.315 TFW, Error: Write() -- IOBOUND begin cnt(9400)
> free(3123)
> 2006-02-28 18:54:33.321 TFW, Error: Write() -- IOBOUND end
>
> They lasted only for 4 seconds and the rest of the recording seems to
> have recorded just fine.
>
> I'm not sure that I would be able to use the frontend while I'm
> recording.  I know the nslu2 isn't fast, although it can handle at
> least 40 mbits/sec.
>
> Anybody knows why the "WA" is so high?
> A larger ring buffer would help here?
>
> Thanks.
>
> Steve M.


Have you tried increasing/decreasing the "HD Ringbuffer" setting in
the general settings (all backends have to be restarted if you change
this I believe)?  Also make sure the nics are at 100mb/full-duplex
(ethtool).

running `vmstat 1` should give you a little bit better information on
resource usage.  The WA is how much time is being spent waiting for
disk read/writes.  Increasing the buffer size isn't going to help you
if you are seeing high WA times for the whole recording.