<div class="gmail_quote">On Wed, Mar 16, 2011 at 11:56 PM, Douglas Peale <span dir="ltr"><<a href="mailto:Douglas_Peale@comcast.net">Douglas_Peale@comcast.net</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
I have been having picture breakups in my recordings lately. Looking in my backend logs I noticed that they seemed to correlate<br>
to "TFW, Error:Write() -- IOBOUND" errors. Most of them are short, on the order of seconds, but this one:<br>
<br>
2011-02-21 22:56:51.280 TFW, Error: Write() -- IOBOUND begin remaining(4865) free(0) size(4194304) cnt(1)<br>
2011-02-21 22:57:00.738 Player(0): Timed out waiting for free video buffers.<br>
2011-02-21 23:05:02.603 TFW, Error: Write() -- IOBOUND end<br>
<br>
is over 8 minutes long, and corresponded with the missing ending of a recording.<br>
<br>
The system is an i7 processor with 6 GB of ram and a 1.5 TB SATA drive. The only slightly unusual thing I am doing on my system<br>
is starting the commercial flagging as soon as the recording starts. The assumption being this should reduce disk accesses since<br>
the data should still be in the disk cache when the commercial flagger tries to read it.<br>
<br>
I assume the normal behavior would be worse since it would start commercial flagging the moment the recording ended, the<br>
recording would be bigger than the disk cache, so the beginning of the file would already have been dumped from the cache, and<br>
if a second recording had started, the second recording would be fighting with the commercial flagger for disk bandwidth.<br>
<br>
Am I making a bad assumption here? Should I just go out and get a SSD for the system, and use the existing drive only for<br>
recordings?<br>
<br>
Is there any way to get MythTV to buffer to RAM when the disk drive gets bottlenecked? 6GB of ram should be good enough for a 30<br>
minute ring buffer.<br></blockquote><div><br></div><div>I had no end of problems when I had the OS, swap, database, and recordings all on the same spindle. SATA is not the best for heavier random access loads and I found that having recordings separate from OS/database was good enough but having the DB on it's own disk also was best (3 devices).</div>
<div><br></div><div>Kevin</div></div>