[mythtv-users] Backend seems to be crashing frontend -- recurring problem??

Roger Heflin rogerheflin at gmail.com
Mon Aug 18 08:21:54 UTC 2008


Allen Edwards wrote:
> On Sun, Aug 17, 2008 at 5:36 PM, Roger Heflin <rogerheflin at gmail.com> wrote:
>> Allen Edwards wrote:
>>>
>>> I am using CPU++ and the usage was much lower than before the system
>>> started having problems.  I am using nvidia driver 173.14.05
>>>
>>> Before I switch video cards, do you find it strange that two frontends
>>> act the same, either both work or both don't?  Do you think the video
>>> card could hang up myth backend so that it won't serve either its
>>> local or the remote frontend?  I don't have another PCI-express card.
>>> I probably have an old PCI card or could borrow one from one of my
>>> kids.
>>>
>> Yes, if the nvidia/X driver fails in the wrong way, it can take out the
>> machine completely.    I have had it do it on at least one of my machines
>> several times.    And it changes depending on the version of the card, it
>> can change depending on the kernel and nvidia and X driver versions.
>> There is really no easy way to validate that the combination works completely,
>> and things can change with updates.
>>
>> And I have noticed that using GL or xvmc increases the chances of software
>> issues (both appear to be less carefully tested, so more mistakes get
>> through).
>>
>> If both frontends are identical hardware (and maybe just identical video cards)
>> and they both have identical software the would both crash the same (if it were
>> a software issue, if it were a hardware issue it would be unlikely for them to
>> have the same failure.).
>>
>>                            Roger
> 
> Both frontends are completely different.  One is a AMD 5400++ and the
> other is a 2.4GHz P4.  One has Mythbuntu 8.04, the other 8.04.1.  The
> only  thing in common is that they both have 6200 video cards although
> one is PCE-e and the other AGP.
> 
> This system worked for two months without problem, now I can crash it
> at will.  I did another couple of experiments.
> 1) I started the master backend and shut down its frontend before it crashed.
> 2) I went to the other frontend and it acts just like the first one,
> crashes easily.
> 
> Here is something interesting I noticed.  It plays most channels fine
> but if I try and play anything from channel 9, it crashes.  Strange as
> 9 and 11 both are 1080i and 11 is fine.  Maybe I am imagining it but
> it sure looks like it is turn to a program from 9 and it looks like
> noise.  Remember that I can watch these channel 9 shows from mythweb.
> 
> Still stumped.
> 
> Allen

If you can find a way to see the kernel messages at the time of the crash
you might want to check those, one of my 6200's went bad (bad enough to not
consistently post)  and it was getting some odd messages in dmesg/messages
about it (usually the messages got put into the log a few seconds before
it took the machine out), the funny thing is that the replacement card
does similar things (though it does post consistently) so I think the card 
dieing (there are several bloated caps on my 6200) may have screwed up other 
things on the MB's pcie slot.

There are 9 caps on my board, 4 are bloated, with 2 of the bloated ones
having what looks like brown crap on the top of them, I would suspect that
there were earlier failure issues, just finally when the last cap had a issue
resulting in worse issues.

My card was bought in 9/05, and made by ASUS, though you can never tell
who actually made the cards so there may be other manufacturers that have the 
same issues, and from what I have seen on  cap issues are usually pretty 
consistent (all of them go and if they are all used consistently the will all 
die at close to the same time).

                                Roger


More information about the mythtv-users mailing list