<br><br><div class="gmail_quote">On Thu, Mar 24, 2011 at 10:03 AM, Kenneth Emerson <span dir="ltr"><<a href="mailto:kenneth.emerson@gmail.com">kenneth.emerson@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div>The specific problem I see during playback is a random, short pause with either missing frames or missing audio; hard to tell which. The pause is very short (less than 2 seconds) but long enough to miss 1-3 words of the dialog. The video is not damaged since I can back up and replay the same segment without any interruption. These pauses are random and occur anywhere from none in an hour-long show to four or five times in that period. It is not related to the scheduler starting/stopping other recordings as I check my watch every time one occurs. I do see a slight stutter, occasionally, when I know there are other recordings stopping/starting and that is a different and less annoying symptom. I also limit the number of simultaneous jobs to two and I hold off commflagging until after the recordings have finished. There appears to be nothing of interest in the frontend log. I occasionally see "ALSA, Error: WriteAudio: buffer underrun" messages, but these do not seem to correlate to the pauses.</div>
=========================================<br></blockquote></div>I have little to add as a follow-up. While watching several recordings over the weekend, i had a laptop next to me with a tail of the frontend log, top, and iotop running. There were several times when the 'pause' occurred but I still have no distinct correlation with some other activity other than when other recordings are starting/stopping. It doesn't appear to be a CPU utilization problem as there were no pauses when a transcode job was running putting the CPU utilization at 270% (four cores combined in top) and I/O reads > 2Mb/sec. Most of the time the CPU utilization was sitting at around 90->93% idle.<br>
<br>When a recording is starting/stopping, there is definitely a noticeable 'glitch'. The two times that a pause occurred that wasn't on the hour, I saw the "ALSA, Error: WriteAudio: buffer underrun" message pop up on the frontend log. The next pause there was nothing in the log. The underrun error could easily be the result of the problem instead of the cause of the problem; plus the fact that it didn't occur the second time would lead me to believe this to be true. The total I/O read/write never exceeded 3Mb/sec (that I could see) using iotop.<br>
<br>I can only guess that what I need now is some way to get a constant stream (data collection) of I/O and CPU activity that has a small enough granularity that it might catch some anomaly when I see the pause in the playback. Any suggestions gratefully accepted.<br>
<br>--Ken E.<br>