[mythtv-users] Please help! My mythbackend started to crash sporadically

[Andrew Junev]

Tuesday, July 1, 2008, 7:58:26 PM, you wrote:

> Tuesday, July 1, 2008, 7:35:39 PM, you wrote:

>> On Tuesday 01 July 2008 09:04:28 Andrew Junev wrote:
>>> Tuesday, July 1, 2008, 6:50:40 PM, you wrote:
>>> >> *** glibc detected *** mythbackend: malloc(): memory corruption (fast):
>>> >> 0x08dec407 ***
>>> >
>>> > Not a dev but I'd reboot the machine and let memtest loose on it for a
>>> > few hours (at least one full cycle).  I'd also break open the case and
>>> > check for dust build up (along with jammed fans, buldging/leaking caps
>>> > and fouled cables).  RAM modules can go bad.
>>>
>>> Thanks a lot for the fast response! :)
>>>
>>> So you say it's likely to be a hardware problem?
>>> I thought it's something in the software, as the only application that
>>> fails is my mythbackend. I have also a frontend running on that
>>> machine and I haven't had such kind of problems with it...

>> When I initially read your post I though "hardware problem" as well, and the
>> suggestion to run memtest is a good one.

>> Then I read the "after the last upgrade" part, which would indicate a software
>> issue.

> memtest is already running. :)
> Let's see how it goes, but I have a feeling it won't show any
> problems on this machine.

> I'm using Fedora 8 and I'm not using any special software. Every
> installation or update is done via 'yum', so I don't think I screwed
> something up by doing a wrong installation. At the same time I didn't
> see any similar complaints from Fedora users which means my problem is
> probably specific to my machine or to my configuration...

>> However, checking your hardware is a good idea. Remember that just because a
>> problem only happens with a specific program does not prove it's a software
>> problem.

> I fully agree!

>>> P.S. To clean some dust inside the case is probably a good idea.
>>> I'll do some maintenance / cleaning today anyway.

>> Never a bad idea, it's astounding how many problems are related ultimately to
>> dust and other cooling problems like bad fans.

> The fans on this machine are fine. I already checked that.
> But cleaning is still on my actions list for today. :)

>> I have seen more machines brought down by bad fans than any other cause.

> This is an off-topic, but from my experience most crashes are related
> to HDD problems. At the same time many HDD failures are actually
> caused by bad cooling. :)


> As I'm writing this, the first round of memtest is finished. No
> problems detected. Of course, it doesn't mean the problem isn't
> there. I will leave it running for some hours.



Ok, here's the last update:
memtest passed 10 times with no errors detected.
The machine was switched off and cleaned completely. It probably feels
good now. :)
After that memtest still showed no errors, which is understandable. :)

My Fedora was booted up again, mythbackend started and ... it crashed
after a few hours of work. There was nothing in the log file. The last
lines were: 

# tail mythbackend.log
2008-07-02 02:21:55.497 Scheduled 1 items in 0.0 = 0.00 match + 0.01 place
2008-07-02 02:22:27.918 UPnpMedia: BuildMediaMap VIDEO scan starting in :/terabyte/video.archive/:
2008-07-02 02:22:28.810 UPnpMedia: BuildMediaMap Done. Found 563 objects
2008-07-02 02:23:15.380 AutoExpire: CalcParams(): Max required Free Space: 1.0 GB w/freq: 15 min
2008-07-02 02:24:41.394 Reschedule requested for id -1.
2008-07-02 02:24:41.409 Scheduled 1 items in 0.0 = 0.00 match + 0.01 place
2008-07-02 02:28:02.338 Reschedule requested for id -1.
2008-07-02 02:28:02.386 Scheduled 1 items in 0.0 = 0.04 match + 0.01 place
2008-07-02 02:31:02.016 Reschedule requested for id -1.
2008-07-02 02:31:02.032 Scheduled 1 items in 0.0 = 0.00 match + 0.01 place
#

And this is the relevant part of my /var/log/messages:

Jul  1 23:15:45 mythbackend smbd[3494]:   call_nt_transact_ioctl(0x9009c): Currently not implemented.
Jul  2 00:06:47 mythbackend kernel: saa7146 (0) saa7146_i2c_writeout [irq]: timed out waiting for end of xfer
Jul  2 00:43:37 mythbackend kernel: saa7146 (0) saa7146_i2c_writeout [irq]: timed out waiting for end of xfer
Jul  2 02:31:02 mythbackend kernel: mythbackend[2646]: segfault at 79 ip 080713a2 sp bff43480 error 4 in mythbackend[8039000+151000]


Generally, I think the "saa7146_i2c_writeout [irq]: timed out waiting
for end of xfer" error was there since the very beginning. I don't
know if it is important at all but it didn't cause any crashes before.
The mythbackend segfault line is, of course, unusual...

What should I do now? How do I trace the problem?

I can, of course, leave memtest running for a longer time (like a day
or two). But I suppose it won't find any problems... Is there some
other way to make sure it's a hardware issue (if it is)?

-- 
Best regards,
 Andrew