[mythtv-users] Randomly crashing Mythtv combined backend and frontend

Antony Gelberg antony.gelberg at gmail.com
Sat May 18 19:01:08 UTC 2013


On Sat, May 18, 2013 at 9:21 PM, Bill Meek <keemllib at gmail.com> wrote:

> On 05/18/2013 12:53 PM, Matthias wrote:
>
>> Am 18.05.2013 19:43, schrieb Stephen Tan:
>>
>>> Hi list
>>>
>>> I'm remotely administering a Mythtv system for some friends and we are
>>> getting some very frustrating behaviour from the system.
>>>
>>> Setup:
>>> We have 2 systems. One which is the Master mythbackend and Frontend, and
>>> a secondary backend box ( which also is a Frontend ). The main box
>>> has 2 x dvb-s cards which get Freesat  and the backend box has 2 x dvb-t
>>> cards which get Freeview.
>>> The distribution being run is Debian 7.0 Wheezy.
>>> We are running 0.26-fixes compiled from source.
>>>
>>> Behaviour: Random crashes at any time of the day which kills the
>>> frontend as well. We were not getting much meaningful logging information on
>>> the problem, so I have re-compiled with --profile to get some debugging
>>> information and we are now running debugging in order to try and get
>>> some meaningful information about the cause of the crashes
>>>
>>> So far, we haven't had a crash, but the logs are full of the following
>>> types of errors:
>>>
>>> 2013-05-18 16:46:53.704140 I [14630/14821] DVBRead
>>> mpeg/pespacket.cpp:161 (VerifyCRC) - PESPacket: Failed CRC check 0xe3967750
>>> != 0xd67980d0
>>> for StreamID = 0x82
>>> 2013-05-18 16:46:53.704150 E [14630/14821] DVBRead
>>> mpeg/mpegstreamdata.cpp:948 (HandleTSTables) - PSIP packet failed CRC
>>> check. pid(0xc0)
>>> type(0x82)
>>>
>>> Does anyone recognise these symptoms? It would help a great deal if you
>>> could share any fixes with us.
>>>
>>> One thing worth noting is that the system is now running at a high CPU
>>> load now with mythlogserver using a lot of CPU. Probably the Observer
>>> effect is in action here.
>>>
>>> thanks in advance
>>>
>>> Stephen
>>>
>>>
>>> ______________________________**_________________
>>> mythtv-users mailing list
>>> mythtv-users at mythtv.org
>>> http://www.mythtv.org/mailman/**listinfo/mythtv-users<http://www.mythtv.org/mailman/listinfo/mythtv-users>
>>>
>> Hi Stephen,
>>
>> sorry no help regarding the crashes. But regarding the high load from
>> mythlogserver, you might want to check the number of mythlogserver
>> processes - there should be only one, and there are reports from several
>> people that in certain conditions, several mythlogserver processes are
>> started, leading to 100% CPU load...
>>
>> I have had this behaviour myself, and I have seen the frontend (local and
>> remote) getting unresponsive due to the backend load. So maybe this is
>> what you are observing.
>>
>> Killall mythlogserver in my case did bring the system back, so I did not
>> have crashes in that sense.
>>
>> I have not been able to fix the mythlogserver problem, but since moving
>> from back/frontend to just a headless backend + several frontends on
>> other machines, I did not have the problem again.
>>
>> A healthy mythlogserver process should never cause much load, I would
>> think.
>>
>> HTH, good luck,
>>
>> Matthias
>>
>
> Hi;
>
> Good points ^^^. As long as you're building from source, take a look at
> this post:
>
>     http://www.gossamer-threads.**com/lists/mythtv/users/534986#**534986<http://www.gossamer-threads.com/lists/mythtv/users/534986#534986>
>
> Sounds very similar to antgel's experience on the -users channel this
> morning. The
> reason those two errors print is that -v siparser,record are set
> (respectively.) Or
> you're using -v all. My experience has been that errors that require
> action will
> print when using -v general (as opposed to debugging tools.)
>
> I'd suggest to you as well, that you define "Random crashes". If you see
> matching faults in
> /var/log/syslog especially with words like "segfault", then you can look
> for core files and use
> gdb (which sounds like the direction you're going with your rebuild.) The
> user on IRC
> reported not seeing any errors in syslog.
>
> Watching the commits mailing list *strongly* suggests that in 0.27
> mythlogserver will
> be history.
>
> --
> Bill
>
>
Hi Bill and thanks for the input.

I'm just sticking my oar in here as:
1) Stephen (the OP) is maintaining a system for me (actually my parents'
system).
2) I'm antgel on IRC who you referred to above.

Small world.  :)

I can add that at present, we have set --loglevel debug, but no -v.

What has been happening is that the backend has been crashing "randomly".
 As Stephen said, we now have a debug build, which we hope will output a
core dump - plus we are running --loglevel debug.  The high CPU utilization
of mythlogserver is (in my opinion) less of a concern than the crashes,
although if you are telling us that -v general will very likely output as
much error information as we will need to debug our crashes (stack trace
notwithstanding), I guess we could add it if it will reduce the load on
mythlogserver.

Grateful for further comments.

Antony

-- 
http://www.linkedin.com/in/antgel
http://twitter.com/antgel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.mythtv.org/pipermail/mythtv-users/attachments/20130518/4471caff/attachment.html>


More information about the mythtv-users mailing list