[mythtvnz] Failed recording

[Tony Sauri]

On 27/09/12 19:02, Worik Stanton wrote:
> On 27/09/12 13:43, Stephen Worthington wrote:
>> On Thu, 27 Sep 2012 12:44:57 +1200, you wrote:
>>
>>> On 27/09/12 11:43, Tony Sauri wrote:
>>>> On 27/09/12 07:47, Worik Stanton wrote:
>>>>>
>>>>> The TV3 recording
>>>>>
>>>>> 2012-09-26 18:00:03.671 Started recording: "News (Manual Record)":"Wed
>>>>> Sep 26 18:00:00 2012": channel 1003 on cardid 1, sourceid 1
>>>>> 2012-09-26 18:00:03.909 ProgramInfo(1003_20120926180000.mpg), Error:
>>>>> GetPlaybackURL: '1003_20120926180000.mpg' should be local, but it can
>>> The recording failed.  AT 18:00:03.671 the recording started, at
>>> 18:00:03.909 it could not find the file.  But the file does exist.  But
>>> it is 4k
>>>
>>>> And One more question:
>>>>
>>>> Can you play the file that was created with another tool eg VLC,
>>>> mplayer, ffplay etc.?
>>>>
>>> It is only 4k
>> Take a look in /var/log/syslog at the same timestamp - see if there
>> was anything there.  That sort of problem can happen when the drive
>> was unable to handle all of the reading and writing that is happening
>> at the same time.  At the start of a recording, there is heavy
>> database and log access as well as the operating system access for
>> opening the recording files for writing.  So it is possible that if
>> you have too many recordings starting on the same drive as the OS and
>> database that the drive may run out of buffering and a write to it
>
> That is a good idea.  But nothing I can notice in syslog...
>
> Sep 26 17:45:04 mythdu CRON[2705]: (CRON) error (grandchild #2706 failed
> with e\
> xit status 1)
> Sep 26 18:00:01 mythdu CRON[2730]: (smmsp) CMD (test -x
> /etc/init.d/sendmail &&\
>   /usr/share/sendmail/sendmail cron-msp)
> Sep 26 18:09:01 mythdu CRON[2776]: (root) CMD (  [ -x
> /usr/lib/php5/maxlifetime\
>   ] && [ -d /var/lib/php5 ] && find /var/lib/php5/ -depth -mindepth 1
> -maxdepth \
> 1 -type f -cmin +$(/usr/lib/php5/maxlifetime) ! -execdir fuser -s {}
> 2>/dev/nul\
> l \; -delete)
>
>
>
>
>> fails.  What sort of drive is it?  You could try experimenting with
>> setting up multiple recordings starting at exactly the same time on
>> the same tuner, and then with offsetting the starting times by say one
>> minute, and see if having them start at the same time overstresses
>> things.  Modern hard drives are easily capable of writing three
>> recording files at once, after the recordings have started, but all
>> the extra head movement for accessing database and log files at the
>> start can possibly be too much.
>
> Ok.  I'll do that.  I'll start one of the 18:00 recordings early
>
>
>>
>> Also, do you have mythcommflag set up to start processing the
>> recordings as soon as they start?  Unless your processor is fast
>> enough, mythcommflag will be getting the data back out of the
>> recording files on disk and adding to the load.  My new motherboard
>> seems to be fast enough for mythcommflag to keep up in real time and
>> hence be able to use the data in the write buffers before it gets to
>> disk, but with my old motherboard mythcommflag lagged behind and
>> needed disk access.
>
> I have not noticed that setting.  Frankly I find the commercial
> detection too unreliable to use.  That is one of my jobs - to see if I
> can make that work.  In the mean time I might just turn it off.
>
>>
>> There is also a deadlock bug in 0.24+fixes, which I have run into a
>> number of times, that causes your exact error message.  That bug
>> normally causes there to be no recording file though, rather than a
>> tiny one, and I have only had it when recording three or more things
>> on multiple tuners.  And you know it has happened as there are
>> messages in the MythTV log that are very characteristic and
>> mythfrontend stops responding correctly (some things still work, but
>> many do not), until you restart mythbackend.
>
> Hmmm...
>
> Have you a reference for that bug?  Since I am running 0.24 I should
> investigate that.
>
> thanks
> Worik
>
>
Did you also check the MythBackend Log /var/log/mythtv/mythbackend.log ?

Tony