[mythtv-users] Slave backend death kills master

[Brad Templeton]

For a couple of months now, I have had an intermitted problem where
the master backend will hang due to slave backend problems.   For
example, if you make install on the slave without stopping it, the
master will hang.

I have not been able to pin the problem down very well, but yesterday
it got very frustrating.  I had the master recording the Oscars in HDTV.
As far as I know the slave was running happy, I had been using it earlier
in the day, and it dhows no signs of trouble in its logs.

However, as the Oscars are recording, at 7pm (1.5 hours in) the master
wants to tell the slave to record 60 minutes.

Here's abridged sections of the log

2005-02-27 17:28:43.849 /dev/video32 ringbuf avg 0.549535% max 2.53125% samples 47982
2005-02-27 17:30:02.105 Finished recording Oscar Countdown 2005 on channel: 2007
2005-02-27 17:30:02.149 Changing from RecordingOnly to None
2005-02-27 17:30:02.467 Started recording "The 77th Annual Academy Awards" on channel: 2007 on cardid: 2, sourceid 2
2005-02-27 17:30:02.512 Channel(/dev/video32)::SwitchToInput(in 0): Error -1 while setting video mode (v2), "Invalid argument", trying v4l v1
2005-02-27 17:30:02.513 Channel(/dev/video32)::SwitchToInput(in 0): Setting video mode with v4l version 1 worked
2005-02-27 17:30:02.625 Changing from None to RecordingOnly
2005-02-27 17:30:03.503 Reschedule requested for id 0.
2005-02-27 17:30:04.026 Scheduled 139 items in 0.5 = 0.00 match + 0.47 place
2005-02-27 17:30:11.371 Starting Commercial Flagging for "Oscar Countdown 2005" recorded from channel 2007 at Sun Feb 27 17:00:00 2005.
2005-02-27 17:30:11.941 MainServer::HandleAnnounce Playback
2005-02-27 17:30:11.970 adding: myth as a client (events: 0)
2005-02-27 17:30:12.014 MainServer::HandleAnnounce FileTransfer
2005-02-27 17:30:12.023 adding: myth as a remote file transfer
2005-02-27 17:30:12.108 MainServer::HandleAnnounce Playback
2005-02-27 17:30:12.119 adding: myth as a client (events: 0)
2005-02-27 17:32:08.683 /dev/video32 ringbuf avg 0.592545% max 5.15625% samples 33822
[mpeg2video @ 0x36c148]end mismatch left=-1
[mpeg2video @ 0x36c148]Warning MVs not available
2005-02-27 17:35:07.550 /dev/video32 ringbuf avg 0.549763% max 3.53125% samples 48063
2005-02-27 17:37:03.718 MainServer::HandleAnnounce Playback
2005-02-27 17:37:03.754 adding: myth as a client (events: 0)
2005-02-27 17:37:03.761 MainServer::HandleAnnounce FileTransfer
2005-02-27 17:37:03.763 adding: myth as a remote file transfer
2005-02-27 17:38:06.400 /dev/video32 ringbuf avg 0.550199% max 3.6875% samples 48059
2005-02-27 17:58:53.312 /dev/video32 ringbuf avg 0.54834% max 2.4375% samples 47874
2005-02-27 18:00:05.606 MainServer::HandleAnnounce Playback
2005-02-27 18:00:05.633 adding: TVWish as a client (events: 0)
2005-02-27 18:00:05.640 Reschedule requested for id -1.
2005-02-27 18:00:06.221 Scheduled 139 items in 0.5 = 0.30 match + 0.20 place
2005-02-27 18:01:51.371 /dev/video32 ringbuf avg 0.547426% max 1.65625% samples 47846
2005-02-27 18:04:49.428 /dev/video32 ringbuf avg 0.547737% max 2.375% samples 47846
[mpeg2video @ 0x36c148]invalid cbp at 34 14

adding pes stream at pid 0x31 with type 2
adding pes stream at pid 0x34 with type 129
2005-02-27 18:12:23.242 Finished, 5 break(s) found.
2005-02-27 18:13:43.768 /dev/video32 ringbuf avg 0.549919% max 3.75% samples 47870
2005-02-27 18:55:16.837 /dev/video32 ringbuf avg 0.547758% max 2% samples 47848
2005-02-27 18:58:14.910 /dev/video32 ringbuf avg 0.547955% max 2.09375% samples 47850
2005-02-27 18:59:31.836 Started recording "60 Minutes" on channel: 4005 on cardid: 1, sourceid 4
2005-02-27 18:59:32.940 Slave backend: myth has left the building
2005-02-27 18:59:32.943 Reschedule requested for id 0.
Mutex destroy failure: Device or resource busy



And then the master stops, the recording is terminated after 1.5 hours,
and the slave didn't record 60 minutes either (a stub was left in the
database though.)

The only line in the slave's log is:

2005-02-27 18:59:32.821 Changing from None to RecordingOnly



Anybody had this problem?  It was very embarassing as I had invited people
back from dinner to see the Oscars in accelerated HD.   We got back at
8, so a core hour was lost.