[mythtv-users] mythbackend 0.15 crash

[Nick Craig-Wood]

On Tue, Jun 01, 2004 at 10:11:20PM +0100, Mark wrote:
> How long have you been running tv_grab_dvb? The reason I ask is that I have
> a very similar setup - 2 nova-t's and using the dvb grabber - and went
> through the same problem.
> 
> My setup was stable for about 2 months until I started working with the
> tv_grab_dvb grabber a few weeks back. I hadn't changed anything else in
> those 2 months and couldn't remember the last time I'd had a crash - 2
> remote frontends and 1 local. Suddenly I started to get fairly frequent
> backend crashes - usually when recordings ended and when I stopped watching
> recordings (and deleted them). I assumed I had corrupted the database
> somehow - I hadn't touched the main installation.

I don't think its related to tv_grab_dvb.  The system was perfectly
stable with 0.14 and tv_grab_dvb (I ran it for about 10 days before
upgrading to 0.15).

My backend has crashed 5 times in the last 3 days with 0.15.

I'm working the scheduling code harder (once per hour rather than once
per day) but the crashes haven't happened at the right time to be
related.

The recording it crashed after in my original post was set to auto
expire after two episodes... (I record the news every day at 6pm but
only keep 2 episodes).  The schedule update happened while the
recording was active - I wonder if this is relevant...

Also note that the last log line mentions "BBC News".  Not quite sure
what it means though!

2004-06-01 17:57:01 Started recording "BBC News" on channel: 2008 on cardid: 7, 
sourceid 2
2004-06-01 17:57:01 45      BACKEND_MESSAGE[]:[]SCHEDULE_CHANGE[]:[]empty
2004-06-01 17:57:01 51      BACKEND_MESSAGE[]:[]RECORDING_LIST_CHANGE[]:[]e...
2004-06-01 17:57:01 Changing from None to RecordingOnly
2004-06-01 17:57:01 22162   1[]:[]45[]:[]BBC News[]:[] []:[]National and in...
2004-06-01 17:57:01 Using profile 'Default' to record
2004-06-01 17:57:01 DVB#0 Recorder: Card opened successfully.
2004-06-01 17:57:01 DVB#0 DVB Quality monitor is starting at 60s for card 0
2004-06-01 17:57:01 DVB#0 Adding pid 601, type 0
2004-06-01 17:57:01 DVB#0 Adding pid 600, type 1
Audiostream: Layer: 2  BRate: 256 kb/s  Freq: 48.0 kHz
Videostream: ASPECT: 16:9  Size = 720x576  FRate: 25 fps  BRate: 15.00 Mbit/s
2004-06-01 18:06:32 Expiring "BBC News" from 2004-05-26T17:57:00, too many episo
des.
2004-06-01 18:06:38 51      BACKEND_MESSAGE[]:[]RECORDING_LIST_CHANGE[]:[]e...
2004-06-01 18:09:24 Found changes in the todo list.
[reschedule caused by tv_grab_dvb finishing snipped]
2004-06-01 18:09:26 Scheduled 45 items in 1.59996 seconds.
2004-06-01 18:09:27 45      BACKEND_MESSAGE[]:[]SCHEDULE_CHANGE[]:[]empty
2004-06-01 18:09:27 22150   1[]:[]45[]:[]BBC News[]:[] []:[]National and in...
2004-06-01 18:30:00 switching to overrecord for 300 more seconds
2004-06-01 18:35:00 Changing from RecordingOnly to None
2004-06-01 18:35:00 51      BACKEND_MESSAGE[]:[]DONE_RECORDING 7 1926[]:[]e...
2004-06-01 18:35:00 Closing DVB recorder
2004-06-01 18:35:00 Found changes in the todo list.
2004-06-01 18:35:00 PruneOldRecords...
2004-06-01 18:35:00 AddNewRecords...
2004-06-01 18:35:00  |-- Start DB Query...
2004-06-01 18:35:01  |-- Processing 168 results...
2004-06-01 18:35:01  +-- Cleanup...
2004-06-01 18:35:01 Sort by priority...
2004-06-01 18:35:01 BuildListMaps...
2004-06-01 18:35:01 SchedNewRecords...
2004-06-01 18:35:01 Scheduling:
[snip reschedule]
2004-06-01 18:35:06 Scheduled 45 items in 4.4405 seconds.
2004-06-01 18:35:03 DVB#0 Expiring DVB quality data older than 7 day(s)
2004-06-01 18:35:06 45      BACKEND_MESSAGE[]:[]SCHEDULE_CHANGE[]:[]empty
2004-06-01 18:35:06 51      BACKEND_MESSAGE[]:[]RECORDING_LIST_CHANGE[]:[]e...
2004-06-01 18:35:06 21985   1[]:[]45[]:[]BBC News[]:[] []:[]National and in...
[crash!]

> By the time I got around to investigating (about a week), it stopped and
> hasn't happened again since. I still can't come up with any rational
> explanation for what caused it - but maybe, just maybe, give it a few
> days...

I left mythbackend running under the debugger last night and of course
it didn't go wrong!

> (P.S. This was using CVS from early March, now upgraded to 0.15)

Interesting and supports my theory above...

Hopefully I will capture a crash in the debugger and we'll have some
concrete info to go on.  Glad I'm not alone in seeing this ;-)

-- 
Nick Craig-Wood
ncw1 at axis.demon.co.uk