[mythtv] Logo detection and TFW buffer exceeded

Craig Treleaven ctreleaven at cogeco.ca
Fri Oct 12 13:49:11 UTC 2012


Since upgrading to 0.25, I've experienced occasional failed 
recordings which never occurred with prior versions.  I have two 
HDHomeruns: one receiving North American OTA, virtually all HD, and 
the other receiving standard def cable.  Commercial detection runs 
while the program is recording.  The following message is always in 
the log at point of failure:

>2012-10-12 08:29:20.468370 E [3173/37379] HDHRStreamHandler 
>ThreadedFileWriter.cpp:216 (Write) - 
>TFW(/Volumes/Theatre/Mythical/1571_20121012081500.mpg:118): Maximum 
>buffer size exceeded.
>			file will be truncated, no further writing 
>will be done.
>			This generally indicates your disk performance
>			is insufficient to deal with the number of on-going
>			recordings, or you have a disk failure.

The utility programs I have indicate no problems with the disk and it 
didn't have a performance problem in the past.

Recently, I noticed that the recordings always failed 12 to 14 
minutes after they started recording.  Reviewing the logs more 
closely, it appears that the failures occur within a few minutes of 
logo detection starting.  Eg:

>2012-10-12 08:27:25.633142 I [3240/12291] CoreContext 
>ClassicCommDetector.cpp:359 (go) - Finding Logo

The job queue is set to run with CPU usage Low and only one job at a 
time.  The backend is an older Core2Duo machine with the recording 
volume attached via Firewire 400 and running OS X.

It seems to me that logo detection is now hogging the disk I/O for a 
few minutes--to the point that ThreadedFileWriter gives up.

I know I can set my job queue to avoid the prime recording hours.  Is 
there some other setting that may alleviate this issue?

Craig


More information about the mythtv-dev mailing list