<br><br><div><span class="gmail_quote">On 6/11/06, <b class="gmail_sendername">Dan Milne</b> <<a href="mailto:d@nmilne.com">d@nmilne.com</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi Guys,<br> I have a problem with Mythtv version 0.19_p9163-r1 on Gentoo<br>where load suddenly spikes up to the 10 - 25 range. I'm running both<br>back and front end on the same box and it make it unusable. Top<br>
tells me the load is between 10 - 25 with 0.0 % idle time. There's<br>some 13G of disk free ( 95% used ).<br><br>CPU is AMD Athlon(tm) 64 Processor 3000+ with around 512MB of RAM.<br>I've got 3 Technisat SkyStar2 DVB card s which work fine. I'm only
<br>recording digital content.<br><br>Hdparm reports the following for the drive containing the mythtv<br>recordings:<br><br>mythbox ~ # hdparm /dev/hdb<br>/dev/hdb:<br>multcount = 16 (on)<br>IO_support = 1 (32-bit)
<br>unmaskirq = 1 (on)<br>using_dma = 1 (on)<br>keepsettings = 0 (off)<br>readonly = 0 (off)<br>readahead = 256 (on)<br>geometry = 30401/255/63, sectors = 488397168, start = 0<br><br><br>When I strace the mythbackend process, some of the time, I get this:
<br><br>mythbox ~ # strace -p 9373<br>Process 9373 attached - interrupt to quit<br>select(45, [4 13 14 40 41 42 44], [], [], {15, 456000} ) = 0 (Timeout)<br>select(45, [4 13 14 40 41 42 44], [], [], {29, 613635}) = 0 (Timeout)
<br>select(45, [4 13 14 40 41 42 44], [], [], {29, 988476}) = 0 (Timeout)<br>select(45, [4 13 14 40 41 42 44], [], [], {29, 882306}) = 0 (Timeout)<br>select(45, [4 13 14 40 41 42 44], [], [], {29, 993259}) = 0 (Timeout)<br>
select(45, [4 13 14 40 41 42 44], [], [], {29, 848225}<br><br><br>In the mythbackend log file I see this:<br>2006-06-11 22:50:48.325 TFW, Error: Write() -- IOBOUND end<br>2006-06-11 22:50:48.329 TFW, Error: Write() -- IOBOUND begin cnt
<br>(9400) free(6627)<br>2006-06-11 22:50:50.113 TFW, Error: Write() -- IOBOUND end<br>2006-06-11 22:50:50.113 TFW, Error: Write() -- IOBOUND end<br>2006-06-11 22:50:50.117 TFW, Error: Write() -- IOBOUND begin cnt<br>(9400) free(7699)
<br>2006-06-11 22:50:50.120 TFW, Error: Write() -- IOBOUND begin cnt<br>(9400) free(1683)<br>2006-06-11 22:50:50.133 TFW, Error: Write() -- IOBOUND end<br>2006-06-11 22:50:50.134 DVBRec(1): PID 0x301 discontinuity detected
<br>2006-06-11 22:50:50.135 DVBRec(1): PID 0x302 discontinuity detected<br>2006-06-11 22:50:50.135 DVBRec(1): PID 0x304 discontinuity detected<br>2006-06-11 22:50:50.139 TFW, Error: Write() -- IOBOUND begin cnt<br>(9400) free(2259)
<br>2006-06-11 22:50:51.695 TFW, Error: Write() -- IOBOUND end<br>2006-06-11 22:50:51.695 TFW, Error: Write() -- IOBOUND end<br>2006-06-11 22:50:51.698 DVBRec(0): PID 0x200 discontinuity detected<br>2006-06-11 22:50:51.701
DVBRec(2): PID 0x2d0 discontinuity detected<br>2006-06-11 22:50:51.701 DVBRec(0): PID 0x28b discontinuity detected<br>2006-06-11 22:50:51.702 DVBRec(2): PID 0x2d1 discontinuity detected<br>2006-06-11 22:50:51.703 DVBRec(0): PID 0x28a discontinuity detected
<br><br><br>And this:<br>2006-06-11 23:02:06.331 DevRdB(0) Error: Driver buffers overflowed<br>2006-06-11 23:02:06.342 TFW, Error: Write() -- IOBOUND end<br>2006-06-11 23:02:06.342 TFW, Error: Write() -- IOBOUND end<br><br>
<br>Does anyone else experience these types of loads? Obviously the<br>IOBOUND stuff is bad news - I can't see why it would be so IO bound<br>tho - just recording a couple of channels shouldn't cause that right?</blockquote>
<div><br><br>I've been looking for an answer to this, it seems to be happening to more and more people. Typically, the thread will go like this:<br><br>Someone asking about DMA settings, someone asking about what other tasks are running, someone possibly suggesting a driver upgrade. There's also the occasional mysql optimization suggestion.
<br><br>Then it fades away without ever really being resolved.<br><br>Personally, I disabled mythtranscode on the backend machine, even though there is no way my IO should be hitting bandwidth limits without there being another real problem on my machine, and I can load up the hard drives otherwise without seeing any other problems at all.
<br><br>There definitely seems to be a problem, but I don't know how to go about tracking it down. Like I said, for me, pushing the transcoding off to a slave seems to have mostly fixed it, even though it's the mythbackend process that's acting up.
<br><br>Ryan<br></div><br></div><br>